Godina / Year 2015 ZAGREB Prosinac / December Broj / No 16

Jeste li naručili? Visit our website www.hdkbr.hr

Hrvatsko Društvo za Kontrolu Bez Razaranja - punopravni član EFNDT-a i ICNDT-a

Izdavač: HDKBR Hrvatsko društvo za kontrolu bez razaranjaPublisher: CrSNDT Croatian Society of Non Destructive TestingDirektor / Director: mr.sc. Miro Džapo, dipl.ing.

Tajništvo/Secretariat:HIS, Petra Berislavića 6. 10000 Zagreb, RHTel: +385 (01) 60 40 451 Fax:+385 (01) 61 57 129E-mail: hdkbr@hdkbr.hr, tajnistvo@hdkbr.hr Website: www.hdkbr.hr

Izdavački odbor/Editorial Board:Prof. dr. sc. - Vjera Krstelj (Glavni urednik/Editor-in-Chief)Dr.sc. Dubravko Miljković (Izvršni urednik/Executive Editor)Dr.sc. Dario AlmesbergerProf.dr.sc. Nenad GucunskiMr.sc. Irena Leljak

Suradnici/Collaborators:Mag. Ivan Smiljanić (tehnička podrška/technical support)Martina Alviž, prof. (lektor hrvatskog i engleskog jezika/proofreading)Sandro Bura (priprema za tisak/Layout Editor)

Sadržaj / Contents

Poruka predsjednice Message from the President

Kien Dinh, Nenad Gucunski, Jinyoung Kim, Trung H Duong:IMPROVED GPR-BASED CONDITION ASSESSMENT of REINFORCED CONCRETE BRIDGE DECKS USING ARTIFICIAL NEURAL NETWORK

Dubravko Miljković: BRIEF REVIEW of VIBRATION BASED MACHINE CONDITION MONITORING

Fran Jarnjak, Ivan Grga:BESKONTAKTNA MAGNETOMETRIJA i NJEZINA PRIMJENA za ISPITIVANJE VISOKOTLAČNOG PLINOVODA GRADSKE PLINARE ZAGREB na TRASI od IVANJE REKE do TE-TO ZAGREB UKUPNE DULJINE 10.735 m u PROSINCU 2015.

Danko Dobranović:UPRAVLJANJE STANJEM IMOVINE POČIVA na MJERNIM METODAMA

CETRIFIKACIJA

HDKBR Centar za obrazovanje HDKBR Centar za certifikaciju

NDT- NEFORMALNO OBRAZOVANJE

HDKBR Info izlazi četiri puta godišnje/ distribucija 300 kom/broj Godišnja pretplata 300 HRK. Časopis je besplatan za članove HDKBR-a.HDKBR Info možete pratiti na www.hdkbr.hrCrSNDT journal is published four times a year/distribution: 300 copies per issue.Annual subscription - 40 Euro. The journal is free of charge for CrSNDT members An online version is available at www.hdkbr.hr

DOSTAVA PRILOGAHDKBR poziva članove i sve koji imaju materijale zanimljive čitateljima ovog časopisa da ponude priloge. Znanstveni i stručni radovi bit će recenzirani od strane međunarodno priznatih stručnjaka. Za reprodukciju publiciranih radova i izvadaka potrebno je osigurati dozvolu. Tekstovi i mišljenja autora u časopisu ne moraju biti u suglasju sa stavovima HDKBR-a i uredništva. Uredništvo ne nosi odgovornost za pogreške i propuste autora radova.

PAPER SUBMISSIONCrSNDT invites contributions that will be interesting for readers of HDKBR Info Journal. Technical papers submitted are peer-reviewed by internationaly rec-ognized experts. Permission should be obtained for reproduction of individual articles and extracts. The articles and views expressed in the publication are not necessarily in line with CrSNDT, editor and editorial. No liability is accepted for errors or omission.

OGLAŠAVANJE/ADVERTISEMENTCijena oglašavanja/The cost for advertising is:

Stranica/Page in journal Cijena za 4 broja /Cost for 4 numbers per year

Zadnja/The last page(cover page A4 size) 8000 HRK 1080 Euro or 1380 $ (US)

Unutarnja/Inside page (A4 size) 4000 HRK 540 Euro or 690 $ (US)

Unutarnja/Inside pages(A4/2 size; half page) 2000 HRK 270 Euro or 395 $ (US)

Cijena oglasa u samo jednom broju iznosi pola cijene godišnjeg oglašavanja. The price of advertisement published (in only one journal issue) is half of the yearly cost.

1 2

14-23

24-31

32-36

40

36

37-39

3-13

Poruka predsjednice

Poštovani čitatelji, kolege i kolegice, prijatelji HDKBR-a,

Svako novo izdanje našeg časopisa HDKBR INFO je poseban uspjeh, ne zato što HDKBR nema interesa i mogućnosti da časopis opisuje naše uspjehe i potrebe, već zato što sve više nedostaje potpore znanstvenom i istraživačkom radu znanstvenika i istraživača, a time je otežano prikupljanje potrebnih članaka i prezentacija rada naših stručnjaka i svih koji razvijaju opseg i primjenjivost metoda bez razaranja, osnovice sigurnosti i održavanja okoliša. Zato je ovo tek drugi broj u ovoj godini.

Obzirom na sve težu situaciju u industriji, većina tvrtki i NDT djelatnika radi izvan granica Hrvatske nastojeći se održati u ovim teškim danima recesije u kojima se gasi mnogo industrijskih postrojenja. Poznato je da u takvim situacijama najprije ponestaje sredstava za kvalitetu i održavanje. Teško stanje za naše kolege u industriji i u obrazovanju nastavlja se do te mjere da smo ove godine bili prisiljeni otkazati našu međunarodno prepoznatu konferenciju MATEST 2015 za koju je postojao veliki interes naših kolega iz inozemstva, ali nažalost premali broj sudionika iz Hrvatske nije mogao opravdati održavanje.

HDKBR je tijekom svoje 50-godišnje povijesti djelovanja doživljavao i prebrodio mnoge teškoće i prepreke pa možemo očekivati ne samo daljnje teškoće, već i uspjehe, kao i do sada. Kao predsjednica ne samo HDKBR-a, već i Hrvatskog inženjerskog saveza (HIS-a), u čijem okrilju uz više od trideset udruga djeluje i HDKBR, poznata mi je slična situacija pa i znatno teža u drugim inženjerskim udrugama, međutim nitko ne posustaje i ne samo da se traži izlaz, već se nude i naziru dobra rješenja i čak dobri rezultati.

Reindustrijalizacija je rješenje kojim Europska unija nastoji poboljšati svoju ekonomiju. Može li itko zamisliti industriju bez inženjera i tehničara, bez NDT-a i potrebne kvalitete u gradnji i održavanju?

To dalje znači da moramo sačuvati HDKBR u kojem se generiraju, osposobljavaju i potvrđuju svoju osposobljenost naši NDT stručnjaci. Svatko ovome može i mora doprinijeti.

U ovom broju možete naći popis naših članova koji su postigli certifikaciju za treći stupanj metoda nerazornog ispitivanja, a u sljedećim brojevima upoznat ćemo vas i s našim članovima i vanjskim suradnicima koji imaju certifikate iz nerazornih metoda na razini drugog stupnja.

Neka ova informacija bude na dobrobit svih naših kolega i stručnjaka za koje smo sigurni da imaju potrebna znanja i odgovarajući certifikat da ih primjenjuju ne samo u našoj industriji već i mnogo šire, diljem svijeta gdje se priznaju diplome izdane od Hrvatskog društva za nerazorna ispitivanja, udruge koja u Hrvatskoj može izdati certifikat priznat od Svjetske (ICNDT) i Europske NDT federacije (EFNDT).

S velikim interesom i nadom očekujemo vaš doprinos ovim nastojanjima. Pišite nam i predlažite što možemo još učiniti za unapređenje profesije nerazornih ispitivanja i za kvalitetniju suradnju članova unutar naše udruge. Predložite što je vama važno u sljedećoj godini.

1

Prof. dr. sc. Vjera Krstelj

2

Message from the President

Dear readers, colleagues and friends of CrSNDT,

Every new issue of our HDKBR Info journal is a great success, not because CrSNDT does not have interest and possibility to show our successes and needs in the journal, but because we are more and more lacking in support to the scientific and research work of our scientists and researchers. This makes even harder to gather necessary articles and presentations of our experts and everyone else who develop the scope of applications of NDT methods, security and environmental sustainability. Therefore, this is only the second issue of the journal this year.

Because of the difficult situation in the industry, many companies and NDT persons are working outside Croatia trying to keep pace and fight difficult days of recession in which many industrial plants are closing. It is a well-known fact that in those situations, resources for quality and maintenance are first to be reduced. Difficult conditions for our colleagues in the industry and in education are getting so serious to the extent that we were forced to cancel our internationally recognized MATEST 2015 conference. Although many foreign colleagues showed interest in it, due to the small number of Croatian participants, the conference could not take place.

During its 50-year-old history, CrSNDT has had and has overcome many difficulties and obstacles so we can expect not only more difficulties, but also many successes as it has been the case before. As the President of both CrSNDT and HIS (Croatian Engineering Association) which is the umbrella organization of CrSNDT and more than thirty other associations, I am familiar with similar and even worse situations in other associations. However, no one is giving up and everyone is looking for a way out of recession and consequently, good solutions and even good results are on the way. Reindustrialization is a solution by which the European Union is trying to improve its economics. Can anyone imagine the industry without engineers and technicians, NDT and necessary quality in building and maintenance? It further means that we have to preserve CrSNDT which gathers, qualifies and certifies our NDT experts. Everyone needs to and has to contribute to this.

In this issue of our journal you can find a list of our members who are certified for level III in NDT methods and in the next issue we will present our members and associates who are certified for level II.

Let this information be for the benefit of our colleagues and experts for whom we are sure they have knowledge and adequate certificates which they use not only in our industry, but also abroad. Certificates issued by Croatian Society of Non-Destructive Testing are valid all over the world since they are acknowledged by The World Organisation for NDT (ICNDT) and European Federation for NDT (EFNDT).

With great interest and hope, we expect your contribution to these endeavors. Feel free to write to us and suggest what else we can do to improve NDT profession and encourage better cooperation between members of our society. Tell us what you think is important for the next year.

Prof. dr. sc. Vjera Krstelj

3

IMPR

OVE

D G

PR-B

ASE

D C

ON

DIT

ION

ASS

ESSM

ENT

of R

EIN

FOR

CED

CO

NC

RET

E B

RID

GE

DEC

KS

USI

NG

A

RTI

FIC

IAL

NEU

RA

L N

ETW

OR

K

Improved GPR-based Condition Assessment of Reinforced Concrete Bridge Decks Using

Artificial Neural NetworkKien, DINH, Rutgers, The State University of New Jersey, 100 Brett Rd, Piscataway, NJ 08854,

United States, +1(848) 203-8355, kien.dinh@rutgers.edu

Nenad, GUCUNSKI, Rutgers, The State University of New Jersey, 96 Frelinguhyen Rd, Piscataway, NJ 08854, United States, +1(848) 445-2232, gucunski@rci.rutgers.edu

Jinyoung, KIM, Rutgers, The State University of New Jersey, 100 Brett Rd, Piscataway, NJ 08854, United States, +1(512) 689-1048, rutgers.no1@rutgers.edu

Trung H, DUONG, Rutgers, The State University of New Jersey, 100 Brett Rd, Piscataway, NJ 08854, United States, +1(848) 391-8451, duonghuytrung@gmail.com

ABSTRACT - Accounting for the effect of rebar depth variation is one of the most important and challenging tasks to accurately assess the condition of reinforced concrete elements using ground penetrating radar (GPR) technique. In current practices, this task is performed on the individual basis, as for each bridge deck a unique depth correction model is derived from the GPR data collected on it. It is found that such a practice has led to a limited capability of GPR in assess-ing the condition of highly-deteriorated concrete. Therefore, a generic model to account for the depth-amplitude effect is proposed in this research. Using artificial neural network (ANN) modeling, a model for depth correction was calibrated from extensive data collected for a group of bare concrete bridge decks. The obtained ANN for depth correction was then used to assess the condition of a bridge deck, and the attenuation map was compared with those using a traditional depth correction technique. Whereas the conventional approach only detected the relative difference in condition between local deck areas, the outputs using the proposed methodology clearly indicated its capability to assess deck deterioration in absolute terms.

Keywords: Nondestructive Testing (NDT), Ground Penetrating Radar (GPR), Concrete, Inspection, Condition Assessment, Artificial Neural Network (ANN).

1. INTRODUCTIONThe effect of asphalt and concrete cover thickness on the ground penetrating radar (GPR) rebar reflection amplitude, and the need for amplitude depth correction, have been investigated by many researchers/ practitioners in the evaluation of condition of concrete bridge decks. These variations are encountered as a result of deck design, inconsistent construction, deck repair or overlaying, and due to many other reasons. The main purpose of depth correction is to remove the signal loss due to depth-amplitude effects and to normalize rebar reflection ampli-tudes with respect to a common cover thickness [1, 2]. Once all rebar reflection amplitudes have been depth corrected and contour mapped,

certain amplitude threshold values will be used to delineate areas of concrete at various severity levels of deterioration. The thresh-olds may vary from one bridge to another and usually are defined from comparisons with inspection results using other NDT methods, or using a statistically-based data interpretation [1, 2, 3, and 4].

In current practices, the depth correction analysis is usually performed for each indi-vidual bridge deck based on the GPR data collected for that same deck. Although the reported results have shown that the depth correction analysis significantly improves the accuracy of condition assessment of bridge decks [1, 2], one must note that the assessment based on such

4

IMPR

OVE

D G

PR-B

ASE

D C

ON

DIT

ION

ASS

ESSM

ENT

of R

EIN

FOR

CED

CO

NC

RET

E B

RID

GE

DEC

KS

USI

NG

A

RTI

FIC

IAL

NEU

RA

L N

ETW

OR

K

depth-corrected amplitudes is still a relative evaluation technique. Specifically, a rebar in a deck having the strongest reflection would always be considered being associated with sound/good concrete, whereas the ones with amplitude below the threshold would be considered as being in a deteriorated region. As a consequence, it was reported that a GPR survey alone usually provides results with grossly underestimated concrete repair quantities [5]. The reason, as explained by Dinh et al. [6], is that the area with the strongest reflection may have already been deteriorated and attenuated itself.

2. RESEARCH OBJECTIVESThe main goal of this research was to develop an analytical approach that can correct for the effects of signal loss due to the rebar depth vari-ation, and at the same time can eliminate the relative nature of the current depth-correction approaches. In contrast to the traditional meth-ods, a generic depth correction model is being implemented for an entire network of bridg-es of the same deck type. As the signal loss may be different between various materials, GPR antennas, as well as utilized frequency ranges, it was decided that only bare (without overlays) concrete bridge decks and a 1.5 GHz ground-coupled antenna would be investigated in this research. Since the absolute amplitude measured (voltages or data units) depends on the transmit power of each antenna, and the gain set during data collection, the amplitudes need to be normalized using a common basis/scale. Given these explanations, three research objectives were identified:

(i) To explore methods for rebar reflection amplitude normalization;(ii) To better understand the impact of signal loss solely due to rebar depth variation; and(iii) To develop a generic depth correction procedure.

3. BACKGROUNDThe American Infrastructure Report Card in 2013 estimates that an annual investment of $20.5 billion would be needed to eliminate the United States’ bridge deficient backlog by 2028 [7]. As many bridges are considered structurally deficient because of the deteriora-tion of decks [8],

a major portion of this investment would be allocated to maintenance, rehabilita-tion and replacement of bridge decks alone.

Bridge decks deteriorate as a result of various factors. However, rebar corrosion has been identified as one of the most common problems [9] and, thus, numerous research efforts have been directed toward development of inspection techniques that can detect signs of this deterio-ration mechanism, and to it related damage. In that context, the GPR stands out as one of the most commonly used NDT technologies.

Exploiting the principles and phenomena of electromagnetic wave propagation, the GPR has shown to be an effective NDT technol-ogy for bridge deck condition assessment. To acquire the condition information for a particu-lar bridge deck location, a GPR antenna sends a short duration microwave into the deck and receives energy partially reflected from vari-ous interfaces. These reflections are produced as a result of difference in dielectric properties. The strength of reflections is proportional to the dielectric contrast between two adjacent media. However, in a concrete deck with a highly conductive environment caused by free chloride ions (Cl-), pore moisture, along with products (Fe2+) from rebar corrosion, the reflections tend to be attenuated and even diminish to zero. These phenomena have formed the basis for condition assessment of reinforced concrete bridge decks by GPR, and reinforced concrete elements in general. While a comprehensive review of literature regarding the application of GPR in the assessment of reinforced concrete elements is beyond the scope of this text, a detailed description of the most common practices in the GPR data analysis for bridge decks is provided in the subsequent paragraphs.

4. CURRENT PRACTICES FOR GPR DATA ANALYSISCurrent GPR data analysis practices for inspection of concrete bridge decks closely follow the guidance provided by ASTM [10]. With respect to ground-coupled antennas, it recommends the following procedure for GPR data analysis: (1) migration to focus the rebar reflection; (2) recording of rebar reflection amplitude from the migrated data; (3) conversion of reflection amplitude to

5

decibels (dB); (4) definition of the deterioration threshold and contour mapping; and (5) calcula-tion of the percentage of deteriorated deck area.

As a detailed guidance, some major assump-tions have, however, rendered the standard inappropriate for the condition assessment of many bridge decks. The two major assumptions behind the standard are: the top concrete cover (rebar depth) is uniform throughout the deck, and there is always a portion of the deck area that is in a sound/good condition. Whereas the former assumption regarding the rebar depth uniformity has been criticized [1, 2, 11], the latter has barely been discussed.

To account for the signal loss due to rebar depth variation, Barnes et al. [1] proposed a statistical method that is based on the 90th percentile linear regression. The method was conceived when they observed the scatter plots (point cloud) of normalized rebar reflection amplitude versus two-way travel time of GPR signals for some bridge decks being studied. As the upper points in the semi-log plots appeared to form a straight line, they assumed that the relationship was linear and further assumption was made regarding the 90th percentile value. The reason for choosing this value has been to obtain an appropriate statistical reference when deterioration may have affected the reflected amplitude and created outliers.

In a comprehensive study, Romero et al. [2] summarized and compared three different methods for performing depth correction. In terms of implementation, one method is done manually by GPR experts and the other two are automated through computer software. Although each method was described as involv-ing a different mathematical manipulation, they were all based on the idea previously explained. As a result, it was reported that the analysis outputs using the three approaches were similar. With respect to the deterioration threshold value, the research stated that it may vary regionally and that the values tend not to be disclosed.

5. RESEARCH METHODOLOGYThe main hypothesis in this research is that the effects of concrete cover thickness on the rebar reflection amplitude can be better studied on the network level, rather than for each bridge deck individually.

A generic relationship (if any) between the two factors can then be applied for the depth correction for all bridges in the network. If the current depth correction methods are used, highly-deteriorated bridge decks will appear to be in a better condition than they actually are, since all depth-corrected amplitudes will converge to a certain, very low value. Whereas the knowledge about true condition of the deck in such cases can be obtained by comparing the GPR results to those from other NDE techniques, it is not a favorable solution. Using a generic, network wide depth correction model can, therefore, ensure that different bridge decks on the network level be evaluated consistently by the GPR.

The research idea was enabled by the data collected within the scope of the Long-Term Bridge Performance (LTBP) Program, a research project funded by the Federal Highway Administration (FHWA). Specifically, as part of the project, a cluster of twenty-four bridge decks in the Mid-Atlantic region was surveyed by the team from the Center for Advanced Infra-structure and Transportation (CAIT) at Rutgers University, using a range of NDT technologies. All the decks were selected by the research team in coordination with the participating State Departments of Transportation (DOTs) to be representative samples of bridges of the same type. As the first cluster, untreated/bare cast-in-place concrete decks that rest either on steel or prestressed concrete girders were investigated in this study. Whereas the data collected for the cluster bridges were used to develop the depth-amplitude model, as shown in Figure 1, two independent bridge decks of the same type were used for the validation of the proposed methodology.

As can be seen in Figure 1, the depth-amplitude model was developed based on the GPR data collected from areas of sound/good concrete. These areas were identified for each bridge deck from the combined results of three NDT techniques, namely Half-Cell Potential (HCP), Electrical Resistivity (ER) and Impact Echo (IE). Specifically, the criteria for defining sound concrete from these techniques are as follows: (1) potential measurement greater than -200 mV for HCP; (2) resistivity greater than 100 kOhm•cm for ER; and (3) no signs of delamination for IE. After the sound concrete areas have been identified, the GPR data from these areas

IMPR

OVE

D G

PR-B

ASE

D C

ON

DIT

ION

ASS

ESSM

ENT

of R

EIN

FOR

CED

CO

NC

RET

E B

RID

GE

DEC

KS

USI

NG

A

RTI

FIC

IAL

NEU

RA

L N

ETW

OR

K

6

were processed to extract the rebar reflection amplitude and corresponding two way travel time. Finally, a total of 23,587 data points

(rebars) were obtained, and used to study the effects of rebar depth on reflection amplitude.

GPR DataGPR Data

Cluster Bridges

Identify Sound Concrete Areas

Half-Cell Potential

Impact Echo

Electrical Resistivity

Rebar Picking

Normalized Amplitude (dB)

Two-way Travel Time (ns)

GPR Data for Sound Concrete

Artificial Neural Networks

Depth-Amplitude Model

Figure 1 Development of Depth-Amplitude Model

5.1 Amplitude NormalizationThere are situations that require the amplitude to be normalized when evaluating GPR data for different bridge decks. For example, for the same frequency antenna, the decks may be collected using different GPR units, or different gains may be set during the data collection. The comparison, in such cases, requires these data be normalized to a common basis/background. The ideal normalization would be to have the reflection amplitude measured using a metal plate. Such a data, unfortunately, does not usu-ally exist during the most GPR data collections on bridge decks.

As a potential basis for amplitude normaliza-tion, direct-coupling is the effect in which the “air wave” and the “surface wave” merge when a GPR antenna is moved toward the surface of a bridge deck. Since having the air wave component, in comparison to the surface wave detected by an air-coupled (horn) antenna, the amplitude of this mixed reflection does not vary much with the local concrete condition. Figure 2 is the illustration of this with two GPR

waveforms collected on the same deck. As can be seen, while the rebar reflection amplitude is very sensitive to concrete deterioration, the direct-coupling amplitudes are almost identical for the two waveforms. This observation forms the basis for using direct-coupling as a normali-zation approach in this study.

For clarification, the normalization is done by dividing the amplitude from rebar reflection by the average direct-coupling amplitude measured in the corresponding profile. As can be imagined, the process will eliminate the difference in the transmit power of the antenna or gain set during the data collection, as long as the gain was set as a constant (one point gain). For the same radar unit, if a constant gain of 1 dB was used, the direct-coupling reflection amplitude would be amplified by 1 dB, as would be the reflection amplitude from a rebar. After the normalization through direct-coupling, the data for all sound concrete areas are convert-ed to decibels to be investigated further for the effects of rebar depth variation.

IMPR

OVE

D G

PR-B

ASE

D C

ON

DIT

ION

ASS

ESSM

ENT

of R

EIN

FOR

CED

CO

NC

RET

E B

RID

GE

DEC

KS

USI

NG

A

RTI

FIC

IAL

NEU

RA

L N

ETW

OR

K

7

Figure 2 Effects of deterioration on the direct-coupling amplitude.

5.2 Artificial Neural Network (ANN)Figure 3 shows a scatter plot illustrating the relationship between rebar depth and rebar reflection amplitude obtained in this study. In the literature, this relationship is assumed to be linear. The only rationale used has been the observation of the points in the upper part of

scatter plots of the GPR data [1, 2]. As the linear regression may not represent the true relationship between the two factors, artificial neural network (ANN), an information-processing technique, is employed to better investigate such a dependency.

Figure 3 Effects of deterioration on the direct-coupling amplitude.

ANNs found beneficial applications in numerous research areas. Comprised of layers of parallel processing elements, or neurons, they simulate biological nervous systems to process acquired data and provide meaningful results/informa-tion. The strength of ANN models is that they are capable to learn from examples so as to

extract essential characteristics or information without assuming the relationship between vari-ables/factors. In comparison to the regression analysis, ANNs are much more appropriate for modeling problems in which the physical nature is too complex or not well understood [12,13].

IMPR

OVE

D G

PR-B

ASE

D C

ON

DIT

ION

ASS

ESSM

ENT

of R

EIN

FOR

CED

CO

NC

RET

E B

RID

GE

DEC

KS

USI

NG

A

RTI

FIC

IAL

NEU

RA

L N

ETW

OR

K

8

Structurally, an ANN consists of an input layer, an output layer and one or several hidden layers in between. Each layer comprises one or more processing elements, also called neurons, which are connected as illustrated in Figure 4. As can be seen, each neuron in the hidden layer is connected to the neurons in neighboring layers by the so-called weighting factors. These factors are modifiable and will be adjusted during the training process when example input-output patterns are presented into the network. An ANN with this type of training algorithm is called a backpropagation (BP) neural network [14, 15].

Figure 4 Typical ANN Structure.

As the name implies, a BP neural network trains itself from examples by propagating the errors of the output backward to the neurons in previous layers of the network. This task is iteratively implemented in two phases. In the first phase, or forward pass, the input signals propagate from input through hidden layer(s) to produce output signals that are calculated based on the initial weights set randomly during the network initialization. In the second phase, the errors, i.e., the difference between the actual and the desired output (target), are propagated backward to adjust the weight-ing factors. As described in Equation 1 below [16], the purpose of the adjustment is to reduce the errors corresponding to each input-output pattern. The process is repeated for all training data until the network stops improving. In other words, the training is completed when adjust-ing weighting factors does not result in reduced errors.

(1)

Where: tpj is the target output for jth element of the output pattern p,

opj is the actual output for jth element of the output pattern p, ipi is the value of the ith element of the input pattern p, Δpwij is the change to be made to the weight from the ith to jth neuron following the presentation of pattern p, and η is the learning rate.

One of the issues with regression analysis and ANNs is problem “overfitting” [17]. It refers to the case when the regression or ANN model performs well for the training patterns, but has poor performance on new data sets present-ed to the model. In the literature, there have been several methods available to solve this problem, of which one is called “early stopping” [18]. In this technique, the available data is, basically, divided into three random data sets: training, validation, and test sets. While the training set is utilized to train the neural network and update weighting factors, the network performance (generalization) is monitored by observing the errors associated with the validation set patterns. After the training, the test set can be used to provide independ-ent evaluation of the obtained model or to compare the performance of different networks. Initially in the training process, the errors for both training and validation sets normally decrease. However, when the overfitting occurs, the errors of the validation set will increase. Therefore, by stopping the training process at this point, a properly-trained neural network can be achieved.

With the theoretical background described above, ANNs are used to investigate the effects of concrete cover thickness on rebar reflec-tion amplitude. Specifically, 23,587 data points (rebar peaks) were divided randomly into three sub sets with the following percentages: 70% for the training set plus 15% for both valida-tion and test sets. With respect to the network topology, it consists of an input layer with one neuron representing concrete cover thickness (rebar depth); one hidden layer as recom-mended by Flood and Kartam [12] for one-input neural network; and an output layer of one neuron for predicting the amplitude of rebar reflection. As for the number of neurons in the hidden layer, since there is no specific rule for determining the appropriate number [12], trials were made to find acceptable values.

IMPR

OVE

D G

PR-B

ASE

D C

ON

DIT

ION

ASS

ESSM

ENT

of R

EIN

FOR

CED

CO

NC

RET

E B

RID

GE

DEC

KS

USI

NG

A

RTI

FIC

IAL

NEU

RA

L N

ETW

OR

K

9

While the experiment has shown that a range of values can provide approximately the same accuracy, a hidden layer of 3 neurons was employed in this study.

Figure 5 depicts the performance of the obtained ANN through regression plots. Each data set in the plot shows the correlation between the target output and the one predicted by the network. As can be observed, a small difference in the correlation coefficients (R) between the three data sets

indicates that the neural network has been properly trained. In addition, since the R2 value (coefficient of determination) for the entire data set is 0.91 (0.954272), one may say that the 91% variance of the rebar reflection amplitude can be attributed to the variation of the concrete cover thickness and therefore can be well predicted by the network. The small remaining variance can be caused by variables such as measuring errors, concrete properties, different sizes of reinforcing bar, or by other random factors.

Figure 5 Regression plots for (a) training set, (b) validation set, (c) test set, and (d) all data points.

Figure 6 presents the fit function from the obtained ANN. As can be seen, the function does not exactly form a straight line, as expected by the conventional depth correction methodologies. It is especially more nonlinear in the region with a small concrete cover thickness. The reason is that, for a shallow reinforcing mat, the rebar reflection is blended with a portion of the

direct-coupling signal, so that its amplitude is affected by the configuration of this mixture. In addition, Figure 6 also reveals that the difference in rebar reflection amplitude before the depth correction for sound concrete may be up to 18 dB. Clearly, this difference is significantly larger than the threshold of -6 to -8 dB specified in the ASTM standard.

IMPR

OVE

D G

PR-B

ASE

D C

ON

DIT

ION

ASS

ESSM

ENT

of R

EIN

FOR

CED

CO

NC

RET

E B

RID

GE

DEC

KS

USI

NG

A

RTI

FIC

IAL

NEU

RA

L N

ETW

OR

K

10

Figure 6 Function fit with the artificial neural network (ANN).

5.3 Depth Correction ProcedureAs explained previously, it is clear that the relationship between concrete cover thickness and rebar reflection amplitude for sound con-crete can be used for the depth correction of GPR data along with the “absolute” condition assessment of bridge decks. The proposed procedure for doing this is depicted in Figure 7. Specifically, for a concrete bridge deck that needs to be assessed, after time-zero cor-rection, rebar locations (pixels) are picked on the profiles as in the conventional amplitude analysis. These locations are then used to extract rebar reflection amplitudes in data units and two-way travel time for implement-ing the next steps. While the purpose of the two-way travel time is to determine the reference amplitude from the ANN depth-amplitude model, the direct-coupling normalization is used to normalize the amplitude to the same background. Finally, the differences between the normalized and reference amplitudes are the depth-corrected amplitudes obtained using the proposed methodology. As can be realized, the more negative the depth-corrected amplitude, the more deteriorated the concrete.

6. CASE STUDY IMPLEMENTATIONThe Pohatcong Bridge in Warren County, New Jersey, was built in 1978. It has a bare concrete deck resting on five single-span steel girders. The bridge is 36-m long and 11-m wide, and the deck is 25-cm thick. The bridge was scanned in August of 2014 using three NDE technologies, namely: GPR, ER, and IE. For the comparison, GPR condition maps were created using both methodologies, i.e., the proposed and conventional depth correction techniques. As depicted in Figure 8, while the spatial distribution of the more deteriorated areas in the two maps appear to be at the same locations, the absolute level of deck deterioration (color spectrum) predicted by the two methods is completely different. To under-stand which method provides more reasonable results, the two GPR maps were compared to the results from the other NDT technolo-gies. The results from those are provided in Figure 9. One can clearly observe from the ER results a very highly corrosive environment in the entire deck. Whereas this condition is the same as what was suggested by the proposed methodology, the conventional depth correc-tion approach was unable to detect the global deterioration of the bridge deck.

IMPR

OVE

D G

PR-B

ASE

D C

ON

DIT

ION

ASS

ESSM

ENT

of R

EIN

FOR

CED

CO

NC

RET

E B

RID

GE

DEC

KS

USI

NG

A

RTI

FIC

IAL

NEU

RA

L N

ETW

OR

K

11

GPR DataGPR Data

GPR Profiles

Rebar Picking

Rebar Reflection Amplitude (Data units)

Two-way Travel Time (ns)

Depth-Amplitude Model

Direct-couplingNormalization

Normalized Rebar Reflection Amplitude (dB)

Reference Rebar Reflection Amplitude (dB)

Amplitude Subtraction

Depth-corrected Amplitude (dB)

Figure 7 Depth correction procedure.

Figure 8 GPR attenuation maps for the Pohatcong Bridge deck with (a) Proposed method and (b) Conven-tional depth-amplitude analysis.

(a)

(b)

IMPR

OVE

D G

PR-B

ASE

D C

ON

DIT

ION

ASS

ESSM

ENT

of R

EIN

FOR

CED

CO

NC

RET

E B

RID

GE

DEC

KS

USI

NG

A

RTI

FIC

IAL

NEU

RA

L N

ETW

OR

K

12

Figure 9 Condition maps for the Pohatcong Bridge deck with (a) ER, and (b) IE.

(a)

(b)

7. DISCUSSIONThe implementation of the proposed method for the case study has proven that the GPR is an effective, powerful technology for condition assessment of bridge decks. As has been seen, through using a generic, network wide depth-amplitude model, the GPR has a capability to assess the condition of highly-deteriorated concrete. With the proposed methodology, it is anticipated that the GPR can be deployed as a tool for managing bridge deck assets in which the condition of different decks on the network level can be compared on the same basis. Specifically, the GPR can be used to estimate concrete repair quantity for each bridge deck, to develop bridge deck condition index, and to prioritize maintenance resources.

In order to achieve the above anticipated objective, a clear roadmap for model develop-ment is proposed. First, the ANN model should be expanded to include more types of bridge decks with different types of overlays, such as latex modified concrete (LMC), bituminous (asphalt), etc.

These materials have properties different from a monolithic (bare) concrete and may respond differently to the propagating electromagnetic source waves. Second, the model should be improved/trained continuously by feeding it with the new data sets. Based on that, the effects of different factors on GPR data will be better understood, including the influence of the rebar depth investigated in this research.

8. CONCLUSIONSThe variation of rebar depth is the most visible factor affecting the condition assessment of bridge decks using the GPR technique. As has been demonstrated, a generic, network-wide depth correction model proposed in this research can minimize the effects of this variance. At the same time, the model can eliminate the relative nature of the overall deck condition when evaluating a depth-corrected data set. As a result, in comparison to the traditional depth correction techniques, the proposed model provides a better description of the absolute deterioration of bridge decks. It is anticipated that the current ANN will be

IMPR

OVE

D G

PR-B

ASE

D C

ON

DIT

ION

ASS

ESSM

ENT

of R

EIN

FOR

CED

CO

NC

RET

E B

RID

GE

DEC

KS

USI

NG

A

RTI

FIC

IAL

NEU

RA

L N

ETW

OR

K

13

expanded to include more types of bridge decks. In addition, its performance will be continuously improved with the addition of new data sets.

9. ACKNOWLEDGEMENTSThe authors sincerely acknowledge the support of the FHWA’s LTBP Program for the support in the bridge deck data collection. The authors are also grateful to Warren County Department of Public Safety for their cooperation in providing access to the bridge in this study. Finally, the authors are grateful to Kenneth Lee, Shane Mott, and Insung Hwang from Rutgers’ Center for Advanced Infrastructure and Transportation (CAIT) for their help during the data collection.

10. REFERENCES[1] Barnes, C. L., Trottier, J.-F., and Forgeron, D. (2008). “Improved concrete bridge deck evaluation using GPR by accounting for signal depth–amplitude effects.” NDT & E Int., 41(6), 427–433.[2] Romero, F. A., Barnes, C. L., Azari, H., Nazarian, S., & Rascoe, C. D. (2015). ”Validation of Benefits of Automated Depth Correction Method: Improving Accuracy of Ground-Penetrating Radar Deck Deteriora-tion Maps.” Transportation Research Record: Journal of the Transportation Research Board, (2522), 100-109. [3] Martino, N., Birken, R., Maser, K., and Wang, M. (2014). “Developing a deterioration threshold model for assessment of concrete bridge decks using ground penetrating radar.” Transportation Research Board 93rd Annual Meeting (No. 14-3861).[4] Dinh, K., Zayed, T., Moufti, S., Shami, A., Jabri, A., Abouhamad, M., & Dawood, T. (2015). “Clustering-Based Threshold Model for Condition Assessment of Concrete Bridge Decks with Ground-Penetrating Radar.” Transportation Research Record: Journal of the Transportation Research Board, (2522), 81-89.[5] Barnes, C. L., and Trottier, J.-F. (2004). “Effectiveness of ground penetrat-ing radar in predicting deck repair quanti-ties.” J. Infrastruct. Syst., 10.1061/(ASCE) 1076-0342(2004)10:2(69), 69–76.[6] Dinh, K., Zayed, T., Romero, F., and Tarussov, A. (2014). “Method for Analyzing Time-Series GPR Data of Concrete Bridge Decks.” J. Bridge Eng., 10.1061/(ASCE)BE.1943-5592.0000679, 2014, pp. 04014086.

[7] American Society of Civil Engineers. Report Card for America’s Infrastructure. 2013.http://www.infrastructurereportcard.org/bridg-es/. Accessed May 12, 2014. [8] Tsiatas, G., & Robinson, J. (2002). “Durability evaluation of concrete crack repair systems.” Transportation Research Record: Journal of the Transportation Research Board, 1795(1), 82-87.[9] Gucunski, N., Imani, A., Romero, F., Nazarian, S., Yuan, D., Wiggenhauser, H., Shokouhi, P., Taffe, A., and Kutrubes, D. (2013). “Nondestructive testing to identify concrete bridge deck deterioration.” Transportation Re-search Board, Washington D.C.[10] ASTM. (2008). “Standard test method for evaluating asphalt-covered concrete bridge decks using ground penetrating radar.” D6087-08, West Conshohocken, PA.[11] Tarussov, A., Vandry, M. and De La Haza, A (2013). “Condition assessment of con-crete structures using a new analysis method: Ground-penetrating radar computer-assisted visual interpretation.” Journal of Construction and Building Materials, Elsevier Vol. 38, pp. 1246–1254.[12] Flood, I., & Kartam, N. (1994). “Neural networks in civil engineering. I: Principles and understanding.” Journal of computing in civil en-gineering, 8(2), 131-148.[13] Bhadeshia, H. H. (1999). “Neural net-works in materials science.” ISIJ international, 39(10), 966-979.[14] Lippmann, R. P. (1987). “An introduction to computing with neural nets.” ASSP Maga-zine, IEEE, 4(2), 4-22.[15] Karnin, E. D. (1990). “A simple proce-dure for pruning back-propagation trained neu-ral networks.” Neural Networks, IEEE Transac-tions on, 1(2), 239-242.[16] Rumelhart, D. E., Hinton, G. E., & Wil-liams, R. J. (1985). “Learning internal represen-tations by error propagation (No. ICS-8506)”. CALIFORNIA UNIV SAN DIEGO LA JOLLA INST FOR COGNITIVE SCIENCE.[17] Girosi, F., Jones, M., & Poggio, T. (1995). “Regularization theory and neural net-works architectures.” Neural computation, 7(2), 219-269.[18] Yao, Y., Rosasco, L., & Caponnetto, A. (2007). “On early stopping in gradient descent learning.” Constructive Approximation, 26(2), 289-315.

IMPR

OVE

D G

PR-B

ASE

D C

ON

DIT

ION

ASS

ESSM

ENT

of R

EIN

FOR

CED

CO

NC

RET

E B

RID

GE

DEC

KS

USI

NG

A

RTI

FIC

IAL

NEU

RA

L N

ETW

OR

K

14

Brief Review of Vibration Based Machine Condition Monitoring

Dubravko, MILJKOVIĆ, Hrvatska elektroprivreda, Vukovarska 37, 10000 Zagreb, Croatiadubravko.miljkovic@hep.hr

ABSTRACT - In the process of channeling energy into job to be performed all machines vibrate. Machines rarely break down without giving some previous warning. The signs of impeding failure are generally present long before a machine totally breaks down. When faults begin to develop in the machine, some of dynamic processes in the machine are changed as well, thereby influencing machine vibration level, temporal and spectral vibration properties. Such changes can act as an indicator for early detection and identification of developing faults. This paper briefly reviews the machine condition monitoring based on vibration data analysis. After the review of major, well established and mature approaches, new unsupervised approaches based on novelty detection are also briefly mentioned.

Keywords: machine condition monitoring, condition based maintenance, vibration analysis

1. INTRODUCTIONCondition monitoring is the process of moni-toring a parameter of condition in machinery, such that a significant change is indicative of a developing failure. Machine condition monitoring can be realized by monitoring following char-acteristics: vibration, aural, visual, operational variables (state of the system), temperature and wear debris (e.g. oil analysis), [1]. Scope of this paper will be limited in consideration to machine vibration. In the process of channeling energy into the job to be performed forces are generated which will excite the individual parts of the machine directly or via the structure. During operation, machine parts are subjected to fatigue, wear, deformation and foundation settlement. When faults begin to develop some of dynamical processes in the machine are changed influencing vibrations produced by the machine (vibration magnitudes in various directions, vibration time domain recording and frequency spectrum and dynamic range). This is the basis for using vibration measurements and analysis in machine condition monitoring, [1-5]. Advantages of vibration monitoring are following: it is capable of detecting, locating and distinguishing faults, it is non-destructive technique, data can be acquired during normal machinery operation, vibration signature

contains most information, it can be applied to inaccessible components and it can be used for on-line condition monitoring.

2. VIBRATIONSVibration is oscillatory motion about a reference position. It is caused by the transfer or storage of energy within structures, resulting from the action of one or more forces. Vibrations can be categorized as free vibrations and forced vibra-tions, linear and nonlinear, and deterministic and random, [1]. If no external force acts on a system and system is left to vibrate on its own, such vibration is known as free vibration, [1]. If a system is subjected to external force, as is common in various machines, the resulting vibration is known as forced vibration, [1]. If the value of excitation is known at any momentn the excitation is called deterministic and result-ing vibration is deterministic, [1]. If excitation is repeating in a periodic way resulting vibration is periodic. Such periodic vibrations are directly linked to repetitive events in machine opera-tion, that are common in rotating machinery like motors, generators, turbines and recipro-cating engines. If excitation is nondeterminis-tic or random the resulting vibration is called random vibration. All machines during its operation vibrate. Even when machines are in

BR

IEF

REV

IEW

of V

IBR

ATIO

N B

ASE

D M

AC

HIN

E C

ON

DIT

ION

MO

NIT

OR

ING

15

BR

IEF

REV

IEW

of V

IBR

ATIO

N B

ASE

D M

AC

HIN

E C

ON

DIT

ION

MO

NIT

OR

ING

good condition, they also generate vibrations, [5]. However, most machines produce low levels of vibration when designed properly. When there are signs of impeding failures, overall vibration level, spectral content and its statistical properties change, often quite significantly.

3. MAINTENANCE SYSTEMSThe life of a machine follows the bathtub curve shown in Figure 1, according to [1]. Machine vibration level also follows the shape of the bathtub curve. Probability of failure of the machine decreases during initial running-in period, then increases very slowly during normal operating period due to normal wear.

Figure 1 Bathtub curve for the live of the machine

As time progresses, the vibration level contin-ues to increase, leading ultimately to failure or breakdown of machine. Maintenance strategies can be divided in three main categories, [3, 6]:

3.1. Run-to-Break Maintenance (breakdown maintenance)The machines are run until the break down, at which time the failed machine is replaced by a new one. This strategy can be used if machines are inexpensive to replace and breakdown doesn’t cause any additional damage.

3.2. Time Based Preventive MaintenanceMaintenance work is performed at fixed time intervals. The intervals are often determined statistically as the period in which no more than 2% (or some other percentage criterion) of the machines will fall from being in new or fully serviced condition. Although this method reduces the chances of unexpected break-downs, it has been found to be uneconomical.

3.3. On-Condition MaintenanceWhen implemented in off-line mode, vibration levels are measured at fixed predetermined intervals. Once when measurement show increased vibration level as sign of impend-ing problem, frequency of measurements increases as shown in Figure 2, according to [1]. Progress in development of vibration monitoring equipment enables implementa-tion in on-line mode at acceptable cost where integrated hardware and software measures machinery performance 24/7 on-line in real time. Condition Monitoring permits efficient maintenance with a minimum of maintenance cost and minimum of unscheduled produc-tion stops. Generally, the off-line implementa-tion is selected due to plant layout logistics against the cost of installing and maintaining an on-line system. On-line diagnostic system with continuous measurements is a choice for diagnostic of strategic high value machinery.

Figure 2 Condition based maintenance

4. TYPICAL VIBRATION MONITORING SETUPTypical setup consists of three main parts: sensors (transducers), signal conditioner with A/D interface and a computer as shown in Figure 3, inspired by [7]. There are several transducers placed on machinery bearings.

Figure 3 Typical vibration monitoring setup

16

4.1. TransducersTransducers can be accelerometers, velocity and displacement probes, [8, 9]. Choice of a transducer depends on flatness of a signal spectrum in a frequency range of interest, [6, 10]. The most reliable, versatile and accurate vibration transducer is the piezoelectric acceler-ometer, [8, 9]. Most accelerometers include an integrated circuit preamplifier (IEPE - Integrated Electronics Piezo Electric design).

Figure 4 Basic accelerometer

Frequency range of one typical accelerometer is shown in Figure 5, according to [8]. Transducers should also fulfill environmental requirements (humidity, temperature, magnetic field, radiation etc.), [9].

Figure 5 Frequency range of accelerometer

4.2. Signal Conditioner with A/D InterfaceData acquisition hardware for vibration monitoring must meet certain basic criteria. The most significant are sampling rate and dynamic range, which must be assessed to ensure compatibility with the monitoring application. For general purpose machinery monitoring, it is advised to use minimum sampling rate of 25 kHz and dynamic range of 120 dB (24 bits).

4.3. ComputerSignal derived from transducers can be analyzed in time, frequency, cepstral or some

other domain, as well as with parameters that are statistical constructs, mostly involving signal from time domain. After signal acquisi-tion, data transformation (e.g. FFT) and feature extraction is performed to extract important information in form of various indicators. Finally, the decision process is performed by classification of features, according to values, into two or three categories (e.g. normal operation, caution, and warning).

5. SIGNAL DOMAINS Signal captured by accelerometers can be analyzed in various domains, [1, 2]:

5.1. Vibration Analysis in Time DomainVibration in time domain is measured by successive sampling outputs from vibration transducer. Time domain measurement could be further represented as waveform, orbits and statistical parameters. In the waveform it is necessary to study the following symptoms: amplitude, amplitude symmetry, time sym-metry, beats/modulation, impacts (shape and amplitude). Waveforms should be compared to catalogued waveforms for particular type of failure by skillful technician/engineer or in recent time by software aided with artificial intelligence (AI) components. It is good practice to use time waveform to enhance and not to replace spectral data (use both types of analysis), [1, 2, 11, 12].

5.2. Frequency Domain TechniquesSpectrum of vibration signal is obtained by Digital Signal Processing (DSP) of vibration signal. It is generally achieved by succes-sive windows on time domain signal, followed by Fast Fourier Transform (FFT). Real and imaginary components of spectrum are used to calculate magnitude and phase shift of each and every frequency component, [1, 2, 6]

5.3. Orbital DomainThis is a special case where two signals are from time domain. Orbit diagrams are generated using two noncontact vibration displacement transducers spaced 90º apart in radial direction of the shaft, [1, 2].

5.4. Envelope DemodulationSignal form transducer first pass through low pass filter, then is rectified, after which peak-hold

BR

IEF

REV

IEW

of V

IBR

ATIO

N B

ASE

D M

AC

HIN

E C

ON

DIT

ION

MO

NIT

OR

ING

17

and smoothing is applied (Figure 6), according to [12]. Method is suitable for signals associated with impulse or impact events.

Figure 6 Simple envelope detection scheme

5.5. Quefrency domain (Cepstrum analysis)The application of cepstrum analysis is based on ability of the method to detect periodicity in the spectrum. The cepstrum can be defined as power spectrum of the logarithm of the power spectrum, (1)-(4), [2]:

( ) ( ){ }2txFSx =ω

( ){ } ( )∫−

=2

2

1T

T

ti dtetxT

txF ω

( ) ( ){ }2log ωτ xSFc =

( ) ( ){ }ωτ xSFc log1−=

(1)

(2)

(3)

(4)

can reconstruct any signal of finite energy, (6):

Cepstrum analysis can be used as a tool for the detection of families of harmonics with uniform spacing [1, 2]. Although groups of harmonics can be seen in spectrum, it is easier to detect such groups in cepstrum.

5.6. Wavelet TransformWavelet transform is the representation of a function by wavelets. Wavelets are scaled and translated copies of fast-decaying oscillating waveforms. Main difference from Fourier transform is that wavelets are localized in both time and frequency, whereas the standard Fourier transform is only localized in frequency. Wavelet transforms have advantages over traditional Fourier transforms for represent-ing functions that have discontinuities and sharp peaks, and for accurately deconstructing and reconstructing finite, non-periodic and/or non-stationary signals, [13]. Wavelet transform exists in continuous and discrete form.

Baby wavelets given in (5):

( ) ( )nbtaat mmnm −= −− ψψ 2/

, (5)

( ) ( )∑∑∈∈

=Zn

nmnmZm

txtx ,,, ψψ (6)

6. DATA PREPROCESSING AND FEATURE EXTRACTIONData captured by sensors include large amount of information. Data preprocessing, [2], include data normalization, noise floor removal, time domain averaging and windowing (Rectangu-lar, Hanning, Hamming, Blackman). Aims of feature extraction, [12], are identifying significant indicators from data, data reduction (vector dimension reduction) and achieving robustness from noise.

6.1. Time-domain Averaged WaveformsThe power of Time-Domain Averaging method, [2], lies in the possibility of relating signal changes to specific kinematic events. Synchronous time averaging uses signal from reference sensor monitoring rotating machinery (and giving ‘tacho’ signal) as shown in Figure 7. The time record length of the averaged signal corresponds to one complete revolution of rotat-ing element.

Figure 7 Time-domain averaging for gear signal processing

Among features most common are statistical parameters for interpretation of the time-domain waveform, [2, 12]. Detailed description and mathematical interpretation of a comprehensive selection of condition indicators is given in [14]. Some are listed below.

BR

IEF

REV

IEW

of V

IBR

ATIO

N B

ASE

D M

AC

HIN

E C

ON

DIT

ION

MO

NIT

OR

ING

18

Brie

f Rev

iew

of V

ibra

tion

Bas

ed M

achi

ne C

ondi

tion

Mon

itorin

g

6.2. RMS - Root Mean Square ValueThe RMS value of a vibration signal is an important measure of its amplitude, (7).

∑=

=N

nnx

NRMS

1

21 (7)

The crest factor and kurtosis function are used to describe the shape characteristics of the signal and change together with vibrations produced by machinery.

6.3. Crest FactorCrest factor, Cf, is defined as:

RMSPC L

f = (8)

where Pl is the signal peak level, (8).

The crest factor calculation tells us how much impacting is occurring in a waveform. Impacting is often associated with roller bearing wear, cavitation and gear tooth wear.

6.4. KurtosisThe fourth order moment, kurtosis, is defined in (9):

( )[ ]

( )221

4

σ

µ

N

nyk

N

n∑=

−= (9)

The kurtosis is sensitive to impulsiveness or “spikeness” of the data. The kurtosis of random signal is 3.0. The vibration signal of a rolling element in good condition is expected to have this value. The kurtosis level rises sharply to around 6 at the onset of discrete damage. The kurtosis value falls as the damage spreads.

6.5. Fractal DimensionFractal dimension D is a statistical quantity that gives an indication of how completely a fractal appears to fill space, as one zooms down to finer and finer scales. This new fea-ture is becoming popular with novelty detection approaches. A procedure to estimate the fractal dimension of waveforms is given in [15]. If there is a set of N values yi sampled from a waveform between time 0 and tmax with sampling interval, xi normalized abscissa, and yi normalized ordinate defined as, (10), (11):

max

*

xxx i

i =

minmax

max*

yyyyy i

i −−

=

(10)

(11)

xmax is the maximum xi, ymin and ymax are the minimum and maximum yi. The fractal dimension of the waveform (Φ) is then approximated by D as, (12):

)2ln()ln(1

NLD′

+=≈Φ (12)

L is the length of the curve in the unit square and

N’=N-1 (13)

6.6. Organizing Data from FFT Analysis into Frequency Beans

FFT analysis produce results for numerous frequency components. It is beneficial to organ-ize these components into smaller number of frequency beans (e.g. octave analysis) suitable for classification.

∑−

=+

+

−=

1

1

11 i

i

N

Nji

iii A

NNB bi NB ,1= (14)

Aj is the amplitude of frequency component j, Bi is value of frequency bean i, Ni and Ni+1 are beginning and end amplitude frequency compo-nent for that frequency bean and Nb is number of frequency beans.

These are some of the most traditional features used for machinery diagnostics. Some methods are more appropriate for detection of particular class of problems than other.

7. VIBRATIONAL CHARACTERISTICS OF MECHANICAL FAULTSFollowing machinery problems can be efficiently diagnosed using vibration analysis, [1-3, 6]: imbalance, misalignment, mechanical loose-ness, rubs, whirling in rotating machinery, hysteretic whirl, whirl due to fluid trapped in rotor, dry friction whip, fluid bearing whip, rolling element bearings, gears etc. Space permits only most important examples.

BR

IEF

REV

IEW

of V

IBR

ATIO

N B

ASE

D M

AC

HIN

E C

ON

DIT

ION

MO

NIT

OR

ING

19

7.1. Time WaveformIt can be used effectively to enhance spectral information in the following applications: low speed applications (less than 100 RPM), indication of true amplitude in situations where impact occurs such as assessment of rolling element bearing defect severity, gears, sleeve bearing machines with X-Y probes (2 channel orbit analysis), looseness, rubs and beats, [11].

7.2. Spectrum and CepstrumIn some situations normal spectral and phase data provide better indication as to the source of the problem without the added complex-ity of time waveform data. Examples include unbalanced and misalignment on normal speed machines. Cepstrum analysis can be used as a tool for the detection of families of harmon-ics with uniform spacing. Although groups of harmonics can be seen with spectrum, it is easier to detect such groups in cepstrum, [1, 2]. Spectrum and cepstrum of undamaged and damaged gearboxes is shown in Figure 8, from [1].

Figure 8 Spectrum and cepstrum of good and bad gearbox

7.3. Orbit DisplaysA circular orbit is obtained from two sinusoidal waveforms equal in amplitude. Waveforms of different amplitudes produce an elliptical orbit. If two waveforms contain more than one discrete frequency, orbit pattern increases in complexity. This display format shows journal bearing rela-tive motion (bearing wear, shaft misalignment, shaft unbalance, lubrication instabilities [whirl, whip], and seal rubs) extremely well (Figure 9), according to [1].

Figure 9 Shaft orbit (imbalance)

It is a powerful monitoring and diagnostic tool, especially on relatively low-speed machinery.

7.4. Envelope DemodulationSignals associated with impulse or impact events can remain dangerously hidden in a normal vibration spectrum reading (as shown in Figure 10, from [16], illustrating a bearing crack), until a catastrophic stage is reached. These “impact events” are high-energy, high-frequency events of very short duration.

Figure 10 FFT of crack in bearing outer race

BR

IEF

REV

IEW

of V

IBR

ATIO

N B

ASE

D M

AC

HIN

E C

ON

DIT

ION

MO

NIT

OR

ING

20

Brie

f Rev

iew

of V

ibra

tion

Bas

ed M

achi

ne C

ondi

tion

Mon

itorin

g

a) Original signal

b) After high-pass filter

c) After demodulation (envelope)

d) After FFT of demodulated signal

Figure 11 Demodulation analysis reveals the problem

Impact events typically occur in the early stages of rolling element bearing wear, and during gear meshing. Envelope demodulation, [16, 17], (Figure 11), [16], reveals such events.

8. DETECTION TECHNIQUESThe interpretation and correlation of vibration data is difficult task and requires experienced skilled personnel and reference vibration data (e.g. machine reference vibration levels and spectrum), [5, 18]. Automated processing and analysis methods are sometimes sought, [12].

8.1. Conventional Methods Statistical features are commonly used to provide a measure of the vibration level that can be compared to a threshold value indicative of a failed condition. The user specifies what the upper and lower reading levels are in terms of acceptable operating condition. This can be based on experience combined with values from General Machinery Criterion Chart, [5, 9].

Comparison of overall vibration levels to predetermined limits is performed using limit checkers, [19]:

Caution, (15):

CLx > (15)

Warning, (16):

wLx > (16)

x is feature value, LC caution level and LW warning level (warning is more serious). There is also choice of another three limits that can be chosen to create an alarm if variable under consideration cross it: Prior experience limits, Concern Level and Non-Operational limit. When setting alarm limits one can choose fixed alarm limits (ISO Standards), [20]. Single indicator generally gives unacceptable accuracy (60-70%). For greater accuracy more indicators should be combined. With help from previous trending data and statistics one can use calculated alarm limits, band alarms and envelope alarms, [21]. When dealing with frequency domain one should specify spectral alarm levels for each frequency bean (Figure 12).

Figure 12 Frequency band alarms

Again, for caution and warning alerts the following conditions have to be fulfilled, (17), (18), this time for each particular frequency band:

ix∃ Ci Lx >

ix∃ Wi Lx >

(17)

(18)

8.2. Supervised Classification TechniquesThese techniques are based on pattern recognition adding flexibility and adaptation to conventional decision techniques, [19, 22, 23 ].

BR

IEF

REV

IEW

of V

IBR

ATIO

N B

ASE

D M

AC

HIN

E C

ON

DIT

ION

MO

NIT

OR

ING

21

They try to mimic pattern recognition ability of human expert. Higher success rate (>90%) can be achieved. Training is accomplished with normal and faulty data. In general, the system has to be previously trained on numerous examples, usually collected by measurements on the same or similar machinery in the past. Vibration measurement data can be automatically classified in various catego-ries (no damage, moderate or heavy wear, impending damage etc.) Block diagram of AI (artificial intelligence) based vibration analysis instrument for detection of impending problem is shown in Figure 13.

Figure 13 Artificial intelligence based vibration analysis

Artificial intelligence tools can classify preprocessed measurement data. Methods used in classifications are decision functions, nearest neighbor classification (1-NN, k-NN), artificial neural networks (ANN), rule based expert system, even those employing fuzzy logic. Basic ANN design is shown in Figure 13.

Figure 14 Basic ANN architecture of interconnected neurons

Training is usually performed by some variant of backpropagation algorithm and presenting input-output (signal features - output indica-tion) sample pairs until network learns input-output mapping. Beside pattern recognition approaches, common are statistical approaches using Hidden Markov models and Gaussian Mixture Models. There also exists model based approach that use mathematical models.

9. NOVELTY DETECTIONNovelty detection is the identification of new or unknown data that a machine learning system is not aware of during training, [24, 25]. Conventional monitoring of objects’ condition relies on known cases of abnormal condition(s). The availability of these known cases is, however, not guaranteed, and in reality it is often difficult to obtain such data. Examples of abnormal engine behavior are rare, and so the novelty detection approach is taken, in which departures from a model of normality (constructed from normal data) are identified. The goal of novelty detection is maximization of detecting true novel samples and minimiz-ing false positives. To achieve this, one must distinguish normal patterns from outliers. For novelty detection, the description of normality is learnt by fitting a model to the set of normal examples, and previously unseen patterns are then tested by comparing their novelty score (as defined by the model). An outlier would be an observation that deviates so much from other observations as to arouse suspicions that it was generated by a different mechanism. Solutions are data driven and include automatic learning. Possible approaches methods include [24 - 27]:- Density Estimation: Estimate a density based on training data- Quantile Estimation: Estimate a quantile of the distribution underlying the training data- Unsupervised learning including cluster-ing of input data- Unsupervised training of artificial neural networks (Kohonen self-organizing maps and LVQ).

Once trained on ‘normal’ data artificial intelli-gence systems can automatically detect outli-ers that correspond to impending maintenance problems and interpret measurement results.

10. APPLICATIONSBasic problems that can be detected include imbalance, misalignment, journal bearings, rolling element bearings and mechanical looseness. Applications can be found in machinery from all fields [28]:

1. Engines, turbines and generatorsVibration analysis for condition assessment and fault diagnostics has a long history of application to power and mechanical equipment.

BR

IEF

REV

IEW

of V

IBR

ATIO

N B

ASE

D M

AC

HIN

E C

ON

DIT

ION

MO

NIT

OR

ING

22

This includes reciprocating engines (e.g. compressors), gas, steam, hydro and wind turbines.

2. PumpsSome of the core equipment issues include cavitation, seal failure, bearing failure and shaft imbalance.

3. FansIndustrial fans are integral components of any industrial plants, particularly automotive, air conditioning, printing, and coating, where fan failure can cause expensive unscheduled plant shutdowns.

4. Gears and gear boxesDetection of localized gear defects, such as tooth fatigue fracture, tooth flank wear (spalling), pitting fatigue, and backlash.

5. Drive beltsDrive belts are inexpensive type of power transmission, but prone to many problems. Vibration analysis can be used for early detection of wear/misalignment problems.

6. Transport vehiclesRailways, especially those operating mainline passenger and high-speed services, need to be run safely, reliably and efficiently. It is also widely used for jet engine monitoring and heli-copters - Health and Usage Monitoring System (HUMS).

7. Quality assuranceThere is an increasing need for fast, reliable and objective quality assessment of every single unit at the end of the assembly cycle. In order to meet the requirements of quality assurance automatic and in-depth quality end-tests are applied in modern manufacturing

11. CONCLUSIONVibration analysis is one of the most powerful condition based maintenance method, and is the key element of many predictive mainte-nance programs. Applying vibration analysis for monitoring the condition of machinery is convenient, inexpensive, and reliable in determining the faults in their early stages and

can avoid unscheduled shutdowns and expensive repair costs. The method is based on the fact that with development of most machine faults, there is also a corresponding change in the way machines vibrate. Machine vibrations are picked by vibration transducers (mostly accelerometers). After data preprocessing (that includes data normalization and noise removal) vibration signals can be analyzed in various signal domains (e.g. temporal, spectral, cepstral). Feature extraction is generally further applied (e.g. use of various indicators, frequency beans) to reduce data dimension-ality. Fault detection can be performed by a human expert or automatically, using conventional methods (limit checkers), pattern recognition or even novelty detection methods. Early detection of faults allows the user to initiate repairs to prevent costly maintenance and generate maximum revenues at minimum costs. Application of condition monitoring and predictive maintenance provides economic advantages in most industries. Wisely devised maintenance strategies have produced huge savings, as it is very expensive to repair broken machine, hold large reserve of spare parts or machines and reschedule production due to repair delays.

12. REFERENCES[1] Rao, S. S., ed., Mechanical Vibrations, Prentice Hall, 4th edition, March, 2003[2] Krishnappa, G., Machinery Condition Monitoring, in Encyclopedia of Acoustics, Vol. II, pp. 869-879, Wiley, 1998[3] Broch, J. T., Mechanical Vibration and Shock Measurements, Brüel & Kjær, 2nd Edition, April 1984, Denmark[4] Randall, R. B., Machinery condition monitoring, in M. J. Crocker (Ed.), Handbook of noise and vibration control, Wiley, 2007[5] Randall, R. B., Vibration-based Condition Monitoring: Industrial, Aerospace and Automotive Applications, Wiley, 2011[6] Bodre, R., Introduction to Machine Vibration, DLI Engineering Corp., 2008,[7] Bandyopadhyay, A., Mandal, S. K. D. and Pal, B., Real-time Condition Monitoring System using Vibration Analysis for Turbine Bearing, Speech and signal processing Group, Calcutta, India[8] Serridge, M. and Llcht, T. R., Piezoelectric Accelerometers and Vibration Preamplifiers, Theory and Application

BR

IEF

REV

IEW

of V

IBR

ATIO

N B

ASE

D M

AC

HIN

E C

ON

DIT

ION

MO

NIT

OR

ING

23

Handbook, Brüel & Kjær, Denmark, November 1987[9] Measuring Vibration, Brüel & Kjær, Denmark, September 1982 [10] Vibration Measurement and Analysis, Lecture Note BA 7676-12, Brüel & Kjær, 1998[11] Dunton, T. A., An Introduction to Time Waveform Analysis, Universal Technologies Inc., 1999[12] Lebold, M., McClintic, Katherine, Campbell, R., Byington, C. and Maynard, K., Review Of Vibration Analysis Methods For Gearbox Diagnostics And Prognostics, Proceedings of the 54th Meeting of the Society for Machinery Failure Prevention Technology, pp. 623-634, Virginia Beach, VA, May 1-4, 2000[13] Zabel, W. and Brehm, M., Wavelet Analysis in Structural Health Monitoring and Damage Detection, Proceedings of the SAMCO Summer Academy 2005, Sept. 2005, Zell am See, Austria[14] Zhu, J., Nostrand, T., Spiegel, C. and Morton, B., Survey of Condition Indicators for Condition Monitoring Systems, Annual Conference of the Prognostics and Health Management Society 2014, Fort Worth, Texas, September 27 - October 3, 2014[15] Sevcik, C., A procedure to Estimate the Fractal Dimension of Waveforms, Complexity International, Vol. 5, pp. 1–19, 1998[16] Demodulation Analysis Explained, Commtest, October 2005, http://www.com-mtest.com/support/faq_s/vbseries/vbclassic_demodulation_analysis [25 December 2015][17] Envelope analysis for effective rolling-element bearing fault detection – fact or fiction?, Application Note BAN0024-EN-11, Bruel & Kjær, Denmark[18] Machine-Condition Monitoring using Vibration Analysis, Brüel & Kjær, Application Note BO 0253-11, Denmark

[19] Miljković, D., Fault Detection Methods: A Literature Survey, MIPRO, Opatija 2011[20] Robichaud, J. M., Reference Standards for Vibration Monitoring and Analysis, Bretech Engineering Ltd., 70 Crown St., Saint John, NB Canada E2L 3V6[21] Utete, S. W., Clifton, D. A. and Tarassenko, L., Trending Of Performance Parameters For Aircraft Engine Condition Monitoring, World Congress on Engineering Asset Management and International Conference on Condition Monitoring 2007, 11-14 June 2007, Harrogate, UK[22] Vilakazi, C. B., Marwala, T., Mautla P. and Moloto, E., On-Line Condition Monitor-ing using Computational Intelligence, WSEAS Transactions On Power Systems, Issue 1, Vol. 1, January 2006[23] Nandi, A. K., Liu, C. and Wong, M. L. D., Intelligent Vibration Signal Processing for Condition Monitoring, Surveillance 7 Int. Conf., Oct. 29-30, 2013, Chartres, France[24] Markou, M. and Singh, S., Novelty Detection: A Review Part I: statistical approaches, Signal Processing, Vol. 83, Issue 12, December 2003[25] Miljković, D., Review of Novelty Detection Methods, MIPRO, Opatija 2010[26] Markou, M. and Singh, S., Novelty Detection: A Review Part II: neural network approaches, Signal Processing, Vol. 83, Issue 12, December 2003[27] Addison, J. F. D., Wermter, S. and MacIntyre, J., Effectiveness of Feature Extraction in Neural Network Architectures for Novelty Detection, Proceedings of the 9th ICANN, Vol. 2, 1999[28] Robichaud, J. M., Practical Appli-cations Of On-Line Vibration Monitoring, Proceedings of. the 1st Pan American Con-ference for Nondestructive Testing PACNDT ‘98, October 1998

BR

IEF

REV

IEW

of V

IBR

ATIO

N B

ASE

D M

AC

HIN

E C

ON

DIT

ION

MO

NIT

OR

ING

24

Beskontaktna magnetometrija i njezina primjena za ispitivanje visokotlačnog

plinovoda Gradske plinare Zagreb na trasi od Ivanje Reke do TE-TO Zagreb ukupne duljine

10.735 m u prosincu 2015.Fran, JARNJAK, HRID – Non destructive testing d.o.o., Zagreb, HRVATSKA,

fran.jarnjak@hrid-ndt.hr

Ivan, GRGA, HRID – Non destructive testing d.o.o., Zagreb, HRVATSKA,ivan.grga@hrid-ndt.hr

SAŽETAK – Opisana je metoda i oprema za inspekciju zakopanih cjevovoda primjenom beskontaktne magnetometerije koja omogućuje inspekciju cjevovoda gdje je nemoguća primjena pigging-a i drugih ispitnih metoda. Također prikazana je i praktična primjena metode na ispitivanju viskotlačnog plinovoda Gradske Plinare Zagreb u duljini od 10735 m obavljenog u prosincu 2015g.

Ključne riječi: beskontaktna magnetometrija, plinovod, magnetna memorija metala

1. UVODU ovom članku biti će prikazana primjena beskontaktne magnetometrijske metode za ispitivanje ukopanog visokotlačnog plinovoda Gradske Plinare Zagreb izvršenog u prosincu

Non-Contact Magnetometric Application on Gradska Plinara Zagreb High Pressure Gas Pipeline Inspection Having Total Length of 10,735 m between Ivanja Reka and TE-TO

Zagreb in December 2015

ABSTRACT - In this paper, a non-contact magnetometric method and equipment is presented that is used to perform inspections of unpiggable buried pipelines. Paper also presents method’s application to inspect a high pressure gas pipeline owned by the Gradska Plinara Zagreb in December 2015 with the total length of 10735 m.

Keywords: non-contact magnetometric inspection, metal magnetic memory, unpiggable gas pipeline

2015. godine, kao jedina mogućnost za nerazorno ispitivanja plinovoda njegovoj kompletnoj trasi, jer isti nije projektiran za druge ispitne metode. Primjena beskontaktne magnetometrijske metode je izuzetno važna kako u Hrvatskoj tako i u svijetu,B

ESK

ON

TAK

TNA

MA

GN

ETO

MET

RIJ

A i N

JEZI

NA

PRIM

JEN

A za

ISPI

TIVA

NJE

VIS

OK

OTL

AČ

NO

G P

LIN

OVO

DA

GR

AD

SKE

PL

INA

RE

ZAG

REB

na

TRA

SI o

d IV

AN

JE R

EKE

do T

E-TO

ZA

GR

EB U

KU

PNE

DU

LJIN

E 10

.735

m u

PR

OSI

NC

U 2

015.

25

BES

KO

NTA

KTN

A M

AG

NET

OM

ETR

IJA

i NJE

ZIN

A PR

IMJE

NA

za IS

PITI

VAN

JE V

ISO

KO

TLA

ČN

OG

PLI

NO

VOD

A G

RA

DSK

E

PLIN

AR

E ZA

GR

EB n

a TR

ASI

od

IVA

NJE

REK

E do

TE-

TO Z

AG

REB

UK

UPN

E D

ULJ

INE

10.7

35 m

u P

RO

SIN

CU

201

5.

jer je velika većina postojećih plinovoda sagrađena još u prošlom stoljeću, koji nisu tada bili projektirani za in-line ispitivanje (engl. pigging), tako da je beskontaktna mag-netometrija jedini način da se dobije uvid u postojeće stanje cjevovoda, prate promjene i kreira plan održavanja cjevovoda koji su već nekoliko desetljeća u eksploataciji, te je njihovo ispitivanje nužno za sigurnost kako građana tako i materijalnih dobara.

2. PRINCIPI ISPITNE METODEBeskontaktna magnetometrija (non-contact magnetometric diagnostic – NCMD) je prim-jena principa metode Magnetne Memorije Metala (MMM) na ispitivanje ukopanih čeličnih cjevovoda (plinovod, naftovod, vodovod i dr.) radi pronalaska anomalija, pogotovo na cjevovodima koji se ne mogu ispitati na neki drugi način jer nisu bili projektirani za npr. primjenu in-line inspekcijom (pigging). NCMD je moguće primijeniti za ispitivanje u urbanim kao i u ruralnim sredinama, bez neke posebne pripreme od strane naručioca inspekcije, izvođača inspekcije, a i ne zahtjeva obustavu rada cjevovoda. NCMD metodu je također moguće primijeniti za ispitivanje dostupnih cjevovoda preko izolacije npr. u industrijskim postrojen-jima, toplovodima i dr. U sljedećim poglavljima objasniti ćemo metodu MMM kao osnovu, te samu metodu NCMD.

2.1. Magnetna memorija metala (MMM)Magnetna memorija metala (MMM) pripa-da u klasu metoda koje koriste magnetske principe za provedbu bezrazornog ispitivanja sa razlikom da je MMM pasivna metoda koja koristi već postojeće magnetno polje objekta koji se ispituje jer su u njemu već formirane magnetske domene tijekom proizvodnje i tijekom same uporabe objekta. Metoda je razvijena od strane Prof. A.A. Dubova i prvi puta je kao termin MMM bila uvedena 1994. godine, te se kontinuirano razvija od strane tvrtke Energodiagnostika (Moskva, Rusija). Metoda za svoj rad koristi principe inverzne magne-torestrikcije (engl. inverse magnetorestrictive effect - Villari effect) gdje dolazi do promjene u magnetizaciji materijala

kod promjene u naprezanju, magnetoplastičnosti (povećano samo-magnetiziranje u zonama povećane plastične deformacije) i curenja magnetnog toka na nivou (jačini) prirodne magnetizacije (engl. magnetic flux leak-age). Samim time, magnetske domene u objektu su izraženije na područjima povećanog naprezanja i deformacije (engl. stress, strain) i na njima dolazi do povećane magnetizacije pod utjecajem magnetnog polja zemlje, zbog gore navedenih principa. Tijekom provedbe MMM ispitivanje bilježi intenzitet magnetskog polja, te se identificiraju zone povećane koncentracije stresa (engl. stress concentration zones - SCZ), jer je u tim zonama najveća vjerojatnost za razvoj defekata tijekom eksploatacije objekta. Moramo napomenuti da je MMM metoda, uz metodu akustičke emisije, jedna od metoda za ranu dijagnostiku stanja metala odnosno objekta još u elastičnoj zoni po krivulji plastičnog tečenja materijala, tako da se tek u plastičnoj zoni defekti mogu potvrditi primjenom klasičnih metoda kao na primjer ultrazvuk. Isto tako, MMM metoda može detektirati i mikropukotine koji zbog svoje male veličine još nije moguće detektirati primjenom ultrazvuka, tako da se preporučuje pojačano praćenje tih područja tijekom eksploatacije.

MMM je moguće primijeniti na feromagnetičnim i paramagnetičnim materijalima tijekom same proizvodnje (npr. prije i poslije toplinske obrade radi potvrde o smanjenju naprezanja, nakon zavarivanja i dr.) kao i tijekom eksploatacije. MMM metoda je primjenjiva za ispitivanje osnovnog materijala, zavarenih spojeva, cijevi, izmjenjivača topline, vijaka na prirubnicama tijekom samog rada postrojenja kao i tijekom remonta, čeličnih kablova (dizala, strojevi), lopatica turbina, statora, željezničkih tračnica i drugo. MMM metoda je standardizirana kao ISO 24497, međunarodni standard za osnovnu primjenu i za primjenu na zavarenim spojevima. Za primjenu metode nije potrebna prethodna priprema ispitnog objekta, prikupljanje podataka se vrši skeniranjem površine bez nekih poseb-nih ograničenja brzine (cca. do 300 mm/s, a i brže u nekim slučajevima), te je samim time MMM jedna od bržih metoda bezrazornog ispitivanja. Na sljedećim slikama 1 do 4 prikazati ćemo osnovne principe same metode i analize.

26

Slika 1 Usporedba MMM metode i ostalih magnetskih metoda. MMM metoda je jedina pasivna metoda i ne koristi umjetno magnetno polje za provedbu ispitivanja. MMM metoda može detektirati promjene na mikrostrukturi materijala kao i mikropukotine koje klasične metode ne mogu detektirati, čime je MMM metoda idealna i za ranu dijagnostiku (izvor slike: Energodiagnostika)

Slika 2 Primjer komponenti prirodnog magnetnog polja zavarenog spoja i shematski prikaz sonde za MMM ispitivanje zavarenog spoja – senzori ispituju zonu utjecaja topline kao i sam zavar (izvor slike: Energodiagnostika)

Slika 3 Rezultat ispitivanja koljena parovoda termoelektrane. Detektirane su dvije zone koncentracije stresa na čijem području je došlo do strukturne promjene materijala (izvor slike: Energodiagnostika)

BES

KO

NTA

KTN

A M

AG

NET

OM

ETR

IJA

i NJE

ZIN

A PR

IMJE

NA

za IS

PITI

VAN

JE V

ISO

KO

TLA

ČN

OG

PLI

NO

VOD

A G

RA

DSK

E

PLIN

AR

E ZA

GR

EB n

a TR

ASI

od

IVA

NJE

REK

E do

TE-

TO Z

AG

REB

UK

UPN

E D

ULJ

INE

10.7

35 m

u P

RO

SIN

CU

201

5.

27

Slika 4 Prikaz MMM ispitivanja - skeniranje se provodi sa senzorima u kontaktu sa objektom koji ne zahtjeva specijalnu površinsku primjenu (razmak samog senzora i površine objekta je cca 1 mm). Prikazani primjeri su ispitivanje zavara, cjevovoda, lopatica turbine, te cijevi kotla termoelektrane. Za svaki ispitni objekt postoje namjenski držači senzora radi ubrzavanja ispitivanja kao i univerzalni držači senzora. (izvor slike: HRID – Non destructive testing i Energodiagnostika)

2.2. Beskontaktna magnetomerijaBeskontaktna magnetometrija (non-contact magnetometric diagnostic – NCMD) je raz-vijena na načelima metode MMM, na način da se beskontaktnim senzorom mjeri sve tri komponente magnetnog polja zemlje (X, Z, Y) koje je iskrivljeno na mjestu ukopanog cjevovoda koji se ispituje, a samo iskrivljenje magnetnog polja je isto tako pod utjecajem stanja naprezanja/deformacije cjevovoda odnosno prisutnosti degradirajućeg procesa na cjevovodu. Analizirajući prikupljene podatke, po karakteristikama deformacija magnetnog polja detektirane anomalije se klasificiraju u 3

kategorije od kojih je prva kategorija najznačajnija. Sama karakterizacija anomalije se vrši kasnije u rovu iskopom cjevovoda na toj lokaciji. Prilikom analize podataka, analizator treba imati uvid o svim preprekama i magnetnim smetnjama tijekom inspekcije (stupovi, automo-bili, ograde, dalekovodi i dr.), što bilježi tijekom prikupljanja podataka, te je također poželjno imati nacrt cjevovoda (ventili, paralelni cjevovo-di ako ih ima, potpornji i dr.). NCMD je moguće također primijeniti na dostupnim cjevovodima za ispitivanje preko izolacije, npr. u industriji ili kod toplovoda. Na sljedećim slikama biti će prikazan osnovni princip NCMD metode kod prikupljanja podataka, Slika 5, te kod analize na Slici 6.

Slika 5 Opći prikaz beskontaktne magnetometrije. Promjena na cjevovodu utječe na magnetno područje zemlje koje se detektira i zabilježi za kasniju analizu i klasifikaciju anomalija (izvor slike: Energodiagnostika)

BES

KO

NTA

KTN

A M

AG

NET

OM

ETR

IJA

i NJE

ZIN

A PR

IMJE

NA

za IS

PITI

VAN

JE V

ISO

KO

TLA

ČN

OG

PLI

NO

VOD

A G

RA

DSK

E

PLIN

AR

E ZA

GR

EB n

a TR

ASI

od

IVA

NJE

REK

E do

TE-

TO Z

AG

REB

UK

UPN

E D

ULJ

INE

10.7

35 m

u P

RO

SIN

CU

201

5.

28

Slika 6 Primjer analize 108 m dugog segmenta ukopanog plinovoda u urbanom području. Gornji grafovi prikazuju intenzitet magnetnog polja (u A/m) a donji grafovi prikazuju gradijent polja (∆H/∆x). Pronađena je jedna anomalija, dok su drugo magnetne smetnje – željezni poklopci plinovoda na pločniku, te stupovi za sprječavanje vožnje automobila po pločniku (izvor slike: HRID - Non destructive testing)

3. PRIMJENA BESKONTAKTNE MAGNETOMETRIJE NA ISPITIVANJU CJEVOVODA GRADSKE PLINARE ZAGREBBeskontaktna magnetometrija (non-contact magnetometric diagnostic – NCMD) je bila primijenjena u prosincu 2015. godine na ispitivanje visokotlačnog (VT) plinovoda Gradske Plinare Zagreb ukupne duljine od 10.735 m. Plinovod nije bio projektiran za primjenu in-line inspekcije (pigging), a pošto je ukopan, jedina metoda za ispitivanje ukupne trase plinovoda je beskontaktna magnetometri-ja. Trasa plinovoda je podijeljena u dva dijela, i to promjera DN500 od PČ Ivanja Reka do PPMRS Zagreb Istok (Etilen) duljine 4.200 m izgrađen bešavnim cijevima 2000. godine, te drugi dio promjera DN600 od PPMRS Zagreb Istok (Etilen) do ispred ograde PMRS TE-TO Zagreb, duljine 6.535 m, izgrađen spiralno zavarenim cijevima u periodu 1984.-1986. godine. Dionica DN500 ima nominalnu debljinu materijala od 8,8 mm, dok dionica DN600 ima nominalnu debljinu od 9,31 mm do 11 mm. Prosječna dubina do tjemena cjevovoda je 2-3 m, a na nekim kraćim segmentima dubina ukopa doseže i 4-4,5 m radi raznih nasipa, korištenja zemlje i drugih sličnih uzroka na trasi pod utjecajem tijeka vremena.

3.1. Oprema za ispitivanjeLokator cjevovoda – Za točno i precizno određivanje trase plinovoda koristio se

lokator cjevovoda, kao standardni proizvod za tu primjenu kojeg koriste razne državna i privatna poduzeća na poslovima održavanja cjevovoda. Lokator cjevovoda ima vise osjetljivih detektora magnetnog polja i za primjenu ispitivanja plinovoda je podešen na traženje frekvencije aktivne katodne zaštite, koja je u ovom slučaju bila 100 Hz. Lokator sadrži grafički prikaz koji prikazuje dubinu do tjemena cjevovoda i navodi operatera grafičkim prikazom na precizno lociranje trase cjevovoda. Lokator koristi baterije velikog kapaciteta za cjelodnevni rad. U slučaju cjevovoda koji nema katodnu zaštitu ili ona nije aktivirana, postoje generatori koji se povežu na cjevovod te se tako lokator može primijeniti. Primjer takvog lokatora je prikazan na Slici 7.

GPS uređaj - GPS uređaj se koristi za točno mapiranje početnih točaka segmenta cjevo-voda koji se u danom trenutku snima kao i za mapiranje same trase cjevovoda. Na osnovu mapirane trase pronalaze se GPS koordinate detektiranih anomalija. Video kamera - video kamera se koris-ti za bilježenje glasovnih komentara kao i vizualnog stanja same trase prilikom prikupljanja, te kasnije tijekom analize poda-taka. Na primjer, u slučaju prolaska automobila prilikom prikupljanja podataka, odaziv signala na instrumentu zbog automobila će se moći kompenzirati, što će biti vidljivo na video kameri.

BES

KO

NTA

KTN

A M

AG

NET

OM

ETR

IJA

i NJE

ZIN

A PR

IMJE

NA

za IS

PITI

VAN

JE V

ISO

KO

TLA

ČN

OG

PLI

NO

VOD

A G

RA

DSK

E

PLIN

AR

E ZA

GR

EB n

a TR

ASI

od

IVA

NJE

REK

E do

TE-

TO Z

AG

REB

UK

UPN

E D

ULJ

INE

10.7

35 m

u P

RO

SIN

CU

201

5.

29

NCMD oprema - sastoji se od senzora, nosača opreme s odometrom (mjeriteljem udaljenosti) te samog instrumenta. Senzor ima dva 3-osna visoko osjetljiva flux-gate magnetometra, jedan blizu površine zemlje dok je drugi vertikalno udaljen i montiran na laganom teleskopskom aluminijskom štapu radi lakšeg transporta i cjelodnevnog baratanja. Nosač opreme se sastoji od okvira za montiranje video kamere i instrumenta, te kotača s enkoderom koji mjeri udaljenost. Nosač i odabir jednog kotača omogućuje fleksibilnost u ispitivanju u raznim uvjetima,

od čiste livade do guste šume ili šikare. Instrument ima internu memoriju i baterijsko napajanje dostatno za cjelodnevni rad, te pomoću integralnog ekrana i tipkovnice omogućuje konfiguriranje, prikupljanje i analizu podataka, te provjeru kvalitete prikupljenih podataka. Sama analiza je ugodnija i brža korištenjem software-a za analizu na prijenosnom računalu nakon preba-civanja podataka s instrumenta, te provjerom komentara i trase pregledom snimljenog video/audio materijala. Prikaz NCMD opreme je na slikama 8a, 8b i 8c. Sam instrument je prikazan na Slici 9.

Slika 7 Tipičan lokator cjevovoda - na slici proizvođača RIGID (izvor slike: HRID - Non destructive testing)

Slika 8b 3-osni senzor udaljen od površine. Osim senzora u kućištu je i elektronika za oba senzora (izvor slike: HRID - Non destructive testing)

Slika 8a 3-osni senzor blizu površine za prikupljanje NCMD podataka (izvor slike: HRID - Non destructive testing)

Slika 8c Nosač instrumenta i video kamere zajedno s enkoderskim odometrom (izvor slike: HRID - Non destructive testing)

BES

KO

NTA

KTN

A M

AG

NET

OM

ETR

IJA

i NJE

ZIN

A PR

IMJE

NA

za IS

PITI

VAN

JE V

ISO

KO

TLA

ČN

OG

PLI

NO

VOD

A G

RA

DSK

E

PLIN

AR

E ZA

GR

EB n

a TR

ASI

od

IVA

NJE

REK

E do

TE-

TO Z

AG

REB

UK

UPN

E D

ULJ

INE

10.7

35 m

u P

RO

SIN

CU

201

5.

30

Gornji graf prikazuje vrijednosti gornjeg, a donji graf prikazuje vrijednosti donjeg 3-osnog senzora. Trenutno je snimljeno 2,18 metara, a 0% je postotak koji prikazuje trenutno stanje memorije za snimanje određenog segmen-ta, gdje je 0% početno stanje, a 100% stanje kada je memorija kompletno puna. Instrument ima dovoljno memorije za segment duljine od nekoliko stotina metara, a nakon spremanja interna memorija instrumenta ima kapacitet od nekoliko km, tako da kod tipičnih ispitivanja nije potrebno nositi dodatnu opremu za presnima-vanje podataka kao npr. prijenosno računalo.

3.2. Provedba ispitivanjaZa ispitivanje je potrebno imati tim od dvoje ljudi. Osoba ispred određuje trasu cjevovoda lokatorom i bilježi je preko GPS uređaja, dok je osoba koja prikuplja podatke pomoću senzora i instrumenta slijedi. Ispituje se uvijek u smjeru protoka plina. Tipičan primjer ispitivanja je prikazan na Slici 10. Ispitivanje se provodilo u segmentima, gdje je početak i kraj segmenata uvjetovan određenim prikladnim orijentirom radi lakše orijentacije prilikom analize i prilikom kasnijih iskopa radi provjera rezultata. Kao orijentiri, najčešće su uzimana raskršće ceste, crkva, željeznička pruga, nadvožnjak, obala jezera, i drugo.

Slika 9 Prikaz NCMD instrumenta (izvor slike: HRID - Non destructive testing)

Slika 10 Tipično provođenje ispitivanja - lokacija Resnik (izvor slike: HRID - Non destructive testing)

Snimljeni podaci su vezani uz početnu točku svakog segmenta zabilježenu GPS-om, izgled početne točke putem kamere (audio komentar/video snimak) kao i sama ruta segmenta po trasi cjevovoda zabilježena preko odometra s preciznošću u centimetrima. Na primjer, kod analize podataka neka anomalija može započeti na 12,3 m od početne točke segmenta po trasi cjevovoda i završava na 18,3 m, te time ima ukupnu duljinu od 6 m. Pošto je znana početna točka, početak i kraj anomalije, kao i cijela ruta, može se preko GIS software-a naći GPS koordinata sredine anomalije.

3.3. Rezultati ispitivanjaIzvještaj s rezultatima ispitivanja sadrži tablicu pronađenih anomalija, njihovu duljinu i GPS koordinatu. Osim toga dio izvještaja je i datoteka u univerzalnom *.kmz formatu, koja se može otvoriti u GIS software-u s kartama ili

besplatnom Google Earth, gdje se može vidjeti na primjer ispitana trasa, točke segmenata ispitivanja, anomalije i drugo radi lakšeg pregleda i orijentacije pronađenih anomalija na samoj trasi odnosno u prostoru. Primjer takvog prikaza cijele ispitane trase prikazan je na Slici 11.

BES

KO

NTA

KTN

A M

AG

NET

OM

ETR

IJA

i NJE

ZIN

A PR

IMJE

NA

za IS

PITI

VAN

JE V

ISO

KO

TLA

ČN

OG

PLI

NO

VOD

A G

RA

DSK

E

PLIN

AR

E ZA

GR

EB n

a TR

ASI

od

IVA

NJE

REK

E do

TE-

TO Z

AG

REB

UK

UPN

E D

ULJ

INE

10.7

35 m

u P

RO

SIN

CU

201

5.

31

Slika 11 Prikaz ispitane trase cjevovoda

Nakon što se izvještaj preda naručiocu ispiti-vanja, uobičajeni postupak je vršenje iskopa na identificiranim mjestima anomalija prvog ranga, odnosno određenim anomalijama drugog ranga, radi provjere stanja cjevovoda primjenom klasičnih metoda bez-razornog ispitivanja kao na primjer vizualna metoda (VT) ili ultrazvuk (UT). U slučaju ispitanog cjevovo-da, anomalije prvog ranga nisu bile pronađene tako da je vršen iskop anomalija drugog ranga, prilikom čega su nađena mjesta površinske korozije ispod izolacije te lokalno stanjenje stjenke od nominalne debljine, ali i dalje u prihvatljivim okvirima za daljnju eksploataciju cjevovoda.

4. ZAKLJUČAKU praksi, općenito govoreći, prosječna brzina za ispitivanje može se računati kao prosjek 2 km dnevno ovisno o stanju na terenu i da li je pretežno urbana ili ruralna sredina gdje prolazi cjevovod. U slučaju ovog plinovoda, prikupl-janje podataka je provedeno u 4 radna dana, gdje je jedan radni dan utrošen na relativno kratkoj dionici kroz Resnik gdje plinovod prolazi kroz ograđena dvorišta obiteljskih kuća zbog kontaktiranja vlasnika radi dozvolu za ulaz u dvorište, prelaska prometne ceste i drugo. Dionice kroz poljane i šumu su bile znatno brže ispitane. Nakon prikupljanja podataka kompletne trase, pristupilo se analizi podataka i pisanju izvještaja,

te potvrdama određenih anomalija u rovu, gdje je brzina iskopa i bila uvjetovana vremenskim prilikama radi zimskog perioda inspekcije. Pronađene anomalije su bile potvrđene, te je time beskontaktna magnetometrija odličan izbor za inspekciju ukopanih cjevovoda koji se ne mogu drugačije ispitati, a isto tako ne zahtijevaju neku posebnu pripremu niti obustavu rada cjevovoda. Ovom prilikom bi se htjeli zahvaliti tvrtci Energodiagnostika (Moskva, Rusija) na pravu za korištenje grafičkih dijagrama oko objašnjenja principa metode MMM i NCMD, kao i djelatnicima Gradske Plinare Zagreb na podršci u pripremi i tijekom ispitivanja trase plinovoda.

5. LITERATURA1. Technical guideline for non-contact magnetometric inspection of gas and oil pipelines. Energodiagnostika, Moskva, Rusija

2. Dubow A. A., Physical Base of the Method of Metal Magnetic Memory, Proceed-ings of the Workshop on Non-Destructive Testing of Materials and Structures, NTM’02 Warsaw, IPPT PAN, 2002, 1-9

3. Tester of Stress Concentration (TSC) User Manual. Energodiagnostika, Moskva, Rusija

4. Scanning device Type 11-6W User Manual. Energodiagnostika, Moskva, Rusija

BES

KO

NTA

KTN

A M

AG

NET

OM

ETR

IJA

i NJE

ZIN

A PR

IMJE

NA

za IS

PITI

VAN

JE V

ISO

KO

TLA

ČN

OG

PLI

NO

VOD

A G

RA

DSK

E

PLIN

AR

E ZA

GR

EB n

a TR

ASI

od

IVA

NJE

REK

E do

TE-

TO Z

AG

REB

UK

UPN

E D

ULJ

INE

10.7

35 m

u P

RO

SIN

CU

201

5.

32

UPR

AVLJ

AN

JE S

TAN

JEM

IMO

VIN

E PO

ČIV

A na

MJE

RN

IM M

ETO

DA

MA

Upravljanje stanjem imovinepočiva na mjernim metodama

Danko, DOBRANOVIĆ, Mind Ability d.o.o., Zagreb, Hrvatska,Tel: 091 519 5178; danko@mindability.eu

SAŽETAK – Napredak tehnologije omogućava svakodnevno proširivanje naših sposobnosti, zato začuđuje spoznaja da se pojedine poslove i aktivnosti marginalizira, što dovodi do općeg urušavanja. “Jedno pričamo, drugo radimo, treće mislimo”. Svrha ovog članka je da se uskladimo sami sa sobom i da se potakne na razmišljanje o smanjenju troškova (ukidanja) uz mjerljive pokazatelje, a ne stihijski. Koristimo ultrazvučni nadzor kako bi zaustavili rasipanje energije na izvoru. Uštedimo i povećajmo pouzdanost fleksibilnim pristupom, kako bi se određene aktivnosti (podmazivanje) odradile ispravno ili prolongirale. Ultrazvučni nadzor stanja karika je koja nedostaje u Hrvatskoj, a vrijednosti koje želimo promovirati jesu poštovanje, jednostavnost, fleksibilnost.

Ključne riječi: upravljanje stanjem imovine, ultrazvuk, nadzor stanja, dijagnostika

1. UVODPrema uptime elementima životnog ciklusa imovine u područje upravljanja stanjem imovine spada 9 cjelina. Smjernice o izobrazbi ljudi za čak 6.5 područja (dokumentacija, vibracije, analiza ulja, ultrazvuk, termografija, balansiranje, podmazivanje) obuhvaćeno je kroz ISO 18436 nadzor stanja i dijagnostiku, ISO 9712 pokriva 3 područja (dokumentaciju, termografiju i kontrole bez razaranja) IEEE kroz svoje propise pokriva strujna ispitivanja motora i ostale opreme.

Asset Condition Management Foundations are Measurement Methods

ABSTRACT - Progress of technology allows expansion of our abilities on daily basis, and that emphasize surprise about marginalization of certain tasks and actions which will lead to a general collapse. “We talk one thing, do some else, think completely different” The purpose of this article is to align our self and to make us think about cost reduction actions (cuts) with measurable indicators, instead of random cuts. Use ultrasound technology and stop wasting energy at root cause. Save and increase reliability through flexible approach, by making certain tasks like lubrication properly or with prolonged interval. Ultrasonic condition monitoring is missing link in Croatia, and values we would like to promote are respect, simplicity, flexibility.

Keywords: asset condition management, ultrasound, condition monitoring, diagnostics

Slika 1 sveobuhvatan pristup imovini

33

UPR

AVLJ

AN

JE S

TAN

JEM

IMO

VIN

E PO

ČIV

A na

MJE

RN

IM M

ETO

DA

MA

Nažalost, centriranje nije pokriveno smjerni-cama jer se smatra korektivnom aktivnošću, a zanemaruje se mogućnost nadzora, pregleda strojeva koji stoje.

Zašto je važan nadzor stanja i dijagnostika?

Zato što daje uvid u stanje imovine bez zaustavljanja životnog ciklusa (proizvodnog procesa). Nadzor stanja i dijagnostika su nam urođeni iako ih možda tako nismo nazivali jer ih u našem svakodnevnom životu radimo automatizmom. Primjenom tehnologije imamo dodatne mogućnosti jer tehnologija proširuje naše sposobnosti i omogućava nam više mjerljivih informacija za donošenje zaključaka. Tako, na primjer, pomoću dokumentacije doznajemo više, vizualni pregled je evoluirao termografijom koja nam omogućuje da vidimo više, osjet dodira je evoluirao s vibracijama koje nam dozvoljavaju da osjetimo više, ultrazvuk nam omogućuje da čujemo više itd…

2. METODADozvolite da predstavim ultrazvuk za koji su kroz ISO 18436-8 propisani zahtjevi za kvalifikacijom i osposobljavanjem osoblja koje se bavi nadzorom stanja i dijagnostikom opreme. Ultrazvuk za nadzor stanja i dijag-nostiku opreme prvenstveno se bazira na pasivnom ultrazvuku, iako s ispitivanjem propusnosti zadire u područje aktivnog ultrazvuka.

Što je to ultrazvuk?

Generalno gledajući svi se slažu da je ultrazvuk mehanički val koji opisuje područje frekvencija iznad praga čujnosti ljudskog uha od 20 kHz nadalje.

Čujno područjeljudi 20Hz-20kHz

Infra zvuk <20Hz

Ultrazvuk > 20kHz

Slika 2 Frekvencijska područja zvuka

Nadalje, u ultrazvučnom području imamo razne vrste aplikacija koje koriste određene frekvencije, a pored toga se još dijeli na aktivni ili pasivni ultrazvuk. Mnogo je primjena uz svakodnevno proširivanje granica mogućeg.

Toliko da se već pomalo spotičemo u nedostatku riječi jer svatko misli na svoje kada se spomene pojam ULTRAZVUK.

Aktivni ultrazvuk: - Puls, Eho, itd.. • NDT - kontrole bez razaranja• Debljina stjenke• Nepravilnosti• Medicinske aplikacije • Ispitivanje propusnosti

Aktivni ultrazvuk: - Snaga • Čišćenja• Zavarivanja• Mehanička obrada• Rezanja

Pasivni ultrazvuk: - Bez kontaktni u prostoru • Propuštanja• Vakuum• Električne inspekcije

Pasivni ultrazvuk: - Kontaktni u strukturi materijala • Mehaničke inspekcije• Unutarnja propuštanja

U ISO 29821 je opisan općeniti postupak ispitivanja ultrazvukom za potrebe nadzora stanja i dijagnostike opreme i koje fizikalne pojave generiraju ultrazvuk, a omogućavaju nam uvid u stanje imovine:

Trenje, turbulencija, udarci (implozija) i ionizacija su fizikalne pojave koje moramo imati na umu prilikom donošenja zaključaka o stanju imovine, samostalno ili u suradnji s ostalim metodama.

Kako bi ultrazvuk mogli čuti, koristi se heterodina transformacija koja omogućava slušanje ultrazvučnih događaja na izabranoj frekvenciji.

Slika 3 Heterodina transformacija

34

Slika 4 Dinamički vremenski interval

Nadalje, tehnologija nam je omogućila snimanje vremenskog intervala događaja. S tim smo dobili mogućnost i dublje analize, uz one primarne, kao što je usporedba dvije iste pozicije ili praćenja trenda na poziciji.

Senzor nije u fizičkom kontaktu s objektom mjerenja te detektira, obrađuje i memorira signal koji se širi zrakom

Slika 5 Bez kontakta (Airborne)

Senzor je u fizičkom kontaktu s objektom mjerenja te detektira, obrađuje i memorira signal koji se širi kroz kruti materijal

U ISO 29821 razlikujemo dva pristupa:

Slika 6 Kontaktno (Structure borne)

Slika 7 Propuštanje vidljivo i potvrđeno

1. Inspekcijski pristup (Inspection) - jačina signala - Idealan za otkrivanje binarno okarakteriziranih defekata.

Propuštanja – kondenzatori pare, ventili, sistemi plinova pod pritiskom ili vakuumom, sistemi komprimiranog zraka i upravljanja, kondenzatori, bojleri, izmjenjivači topline.

2. Nadzor stanja (Condition Monitoring), jačina i analiza signala, praćenje i preporuka - Idealan kao platforma kvalitetnog pristupa održavanju po stanju.

Propuštanja – pumpe, transformatori, osigurači, turbine, generatori

Kavitacija – pumpe

Mehaničke greške – motori, pumpe, zupčanici, reduktori, ventilatori, kompresori, konvejeri, transformatori, turbine, generatori, podmazivanje, ležajevi

Električne greške – motori, rasklopišta, transformatori, izolatori, spojne kutije, osigurači, generatori.

3. REZULTATIPasivni ultrazvuk: Bez kontaktni u prostoruInspekcijski pristup - jačina signala

Propuštanja:U malom dijelu pogona (cca. 10% proizvodnog procesa): pronađeno je 29 velikih propuštanja na instalacijama komprimiranog zraka i 40 manjih. Procjena ukupnog gubitka samo na komprimiranom zraku je jednaka ukupnoj cijeni zapošljavanja 3 inženjera (godišnja plaća i svi ostali troškova - Bosna)

UPR

AVLJ

AN

JE S

TAN

JEM

IMO

VIN

E PO

ČIV

A na

MJE

RN

IM M

ETO

DA

MA

35

Slika 8 Propuštanje locirano

Pasivni ultrazvuk: Kontaktni u strukturi opreme Nadzor stanja - jačina i analiza signala

Kavitacija :Usmjerenost ultrazvuka omogućuje pregled svakog stupnja zasebno, kvalitetan uvid u stanje pumpe

Slika 9 Kavitacija u 3 stupnju pumpe

Pasivni ultrazvuk: Kontaktni u strukturi opreme Nadzor stanja - jačina i analiza signala

Mehaničke greške:Tri ista stroja, tri iste pozicije (ležaj)

Slika 10 Ležaj sporohodnog stroja

Slika 11 Indikacije nepravilnosti struje

Pasivni ultrazvuk: Kontaktni u strukturi opreme Nadzor stanja - jačina i analiza signala

Električne greške:Ionizacija, puzne struje, električni lukovi, corona, mehanička labavost

4. DISKUSIJAZašto je ponovo potrebno osvijestiti nadzor nad podmazivanjem i komprimiranim zrakom?

Zato što su inženjeri dozvolili da se podmazivanje marginalizira i prepusti ljudima koji su taj posao dobili više kao kaznu nego zato što to žele. A opće je poznato da loše podmazan stroj uzrokuje 60% problema u stroju. Mazivo je krvotok svih pokretnih dijelova. Da ne pričam o ekološkom zbrinjavanju viška masti koji se razlije po pogonima.

U današnje doba kada se gleda društveno odgovorno ponašanje neshvatljivo je kako se olako prelazi preko činjenice i gubitaka oko 30% energije za komprimiranje zraka (tijekom 5 godišnjeg ciklusa). Komprimirani zrak je jedan od skupljih medija, a pored toga postoji tehnika (ultrazvuk) koja omogućava da se gubici u instalacijama svedu na manje od 5%.

Ako ste zainteresirani za sveobuhvatni pristup pouzdanosti pogona na jednostavan način, uz mogućnost analize, slobodno se javite i organizirat ćemo demonstraciju u vašem pogonu kako bi iskoristili puni potencijal vaše imovine, kroz ultrazvučno praćenje stanja i upravljanje te upotpunili vođenje kroz pouzdanost.

UPR

AVLJ

AN

JE S

TAN

JEM

IMO

VIN

E PO

ČIV

A na

MJE

RN

IM M

ETO

DA

MA

36

5. LITERATURA[1] ISO 29821 - ISO 29821-1:2011 Condition monitoring and diagnostics of machines -- Ultrasound -- Part 1: General guidelines[2] ] Moguća područja ultrazvučne primjene u funkcionalnim ispitivanji-ma u industriji, http://www.sdt.eu/index.php?page=applications&hl=en [29 April 2016][3] ] Kako pristupiti podmazivanju po stanju, ispitivati odvajaće pare itd…, http://www.sdthearmore.com/learning-centre [29 April 2016]

[4] Uptime elements i primjena u održavanju, http://reliabilityweb.com/articles/entry/get-the-bugs-out-with-the-uptime-ele-ments [29 April 2016]

AUTOR želi zahvaliti … gospodi: André DEGRAEVE, Manager SDT International; Allan RIENSTRA, SDT International Sales Manager; Haris TROBRADOVIĆ, SDT Area Sales & Training Manager; na dostupnosti edukativnog materijala i prezentacijama iz područja nadzora stanja I dijagnostike opreme ultrazvukom. Kao I gospodinu: Terrence O’HANLON, Association of Asset Management Professionals, Reliability-web.com and Uptime Magazine

CET

RIF

IKA

CIJ

A

Hrvatsko društvo za kontrolu bez razaranja uspješno i kontinuirano osigurava usklađivanje s europskim normama za tijela koja provode certifikaciju osoblja.

HDKBR kroz djelovanje Centra za certifikaciju, tijela HDKBR-a akreditiranog za provedbu postupaka certifikacije osoba za nerazorna ispitivanja, slijedom zahtjevnih aktivnosti trajne prilagodbe ustrojstva i poslovanja, osigurava i potvrđuje akreditacijom od Hrvatske akreditacijske agencije ocjenu osposobljenosti.Posljednjom akreditacijom osigurana je valjanost do 15. rujna 2019. godine.

HDKBR je jedino hrvatsko tijelo za certifikaciju osoba za nerazorna ispitivanja koja akreditirani status neprekidno održava od mjeseca rujna 2004. godine.

HDKBR je jedina udruga u Hrvatskoj koja je članica Europske federacije za nerazorna ispitivanja (EFNDT) i Svjetske organizacije za nerazorna ispitivanja (ICNDT), te osigurava sudionicima tečajeva i seminara polaganje ispita i dobivanje certifikata koji se priznaje u Europskoj uniji i šire, kao potvrda kompetencije za provođenje ispitivanja nerazornim metodama.

Svi tečajevi na svim razinama - 1., 2. i 3. stupnja trajno se unapređuju i usklađuju s potrebama osiguravanja znanja i vještina kojima sudionici tečajeva u HDKBR-u stječu potrebne kompetencije i certifikate priznate za uključivanje u tržište rada.

S HDKBR certifikatom možete u svijet, ali vratite se, I ovdje nam treba NDT!

AKREDITACIJA

UPR

AVLJ

AN

JE S

TAN

JEM

IMO

VIN

E

POČ

IVA

na M

JER

NIM

MET

OD

AM

A

HD

KB

R

Cen

tar z

a ob

razo

vanj

e

HD

KB

R

Cen

tar z

a ce

trifi

kaci

ju

TEČAJEVI za KVALIFIKACIJU i CERTIFIKACIJU

37

CENTAR ZA OBRAZOVANJE CENTAR ZA CERTIFIKACIJU

TEČAJ stupanj datum ISPIT

MAGNETSKA KONTROLA Ispitivanje magnetskim česticama MT 1

11.01. - 12.-01. 2016. 13. 1. 2016.

21.3.– 22. 3. 2016. 23. 3. 2016.

4.7.– 5. 7. 2016. 6. 7. 2016.

28.11. – 29. 11. 2016. 30. 11. 2016.

PENETRANTSKA KONTROLA Ispitivanje penetrantima PT 1

18.1.-19.1. 2016. 20. 1. 2016.

25.4. – 26. 4. 2016. 27. 4. 2016.

5.9 – 6. 9. 2016. 7. 9. 2016.

19. – 20. 12. 2016. 21. 12. 2016.

VIZUALNA KONTROLA Vizualno ispitivanje VT 1

22.3. – 23. 3. 2016. 24. 2. 2016.

14.3. – 15. 3. 2016. 16. 3. 2016.

23.5. – 24. 5. 2016. 25. 5. 2016.

11. 7.– 12. 7. 2016. 13. 7. 2016.

3.10. – 4. 10. 2016. 5. 10. 2016.

ULTRAZVUČNA KONTROLA Ispitivanje ultrazvukom UT 1

8.2. – 18. 2. 2016. 19. 2. 2016.

30. 5. – 9. 6. 2015. 10. 6. 2015.

10.10. – 20. 10. 2016. 21. 10. 2016.

Radiografska kontrola RT 129. 2. – 10. 3. 2016. 11. 3. 2016.

14. – 24. 11. 2016. 25. 11. 2016.

MAGNETSKA KONTROLA Ispitivanje magnetskim česticama MT2

22.2. – 24. 2. 2016. 25. 2. 2016.

27.6. – 29. 6. 2016. 30. 6. 2016.

12.9. – 14. 9. 2016. 15. 9. 2016.

VIZUALNA KONTROLA Vizualno ispitivanje VT 2

25-27.1.2016. 28.1.2016.

29. 2. – 2.3. 2016. 3. 3. 2016.

7.3. – 9. 3. 2016. 10. 3. 2016.

4.4. – 6. 4. 2016. 7. 4. 2016.

13. 6.– 15. 6. 2016. 16. 6. 2016.

18. – 20. 7. 2016. 21. 7. 2016.

7.11.– 9. 11. 2016. 10. 11. 2016.

1.2.-3.2.2016 4.2.2016.

PENETRANTSKA KONTROLA Ispitivanje penetrantima PT 2

21.3. – 23. 3. 2016. 24. 3. 2016.

16.5. – 18. 5. 2016. 19. 5. 2016.

25.7. – 27. 7. 2016. 28. 7. 2016.

24.10. – 26. 10. 2016. 27. 10. 2016.

Radiografska kontrola RT211.4. – 21. 4. 2016. 22. 4. 2016.

5.12. – 15. 12. 2016. 16. 12. 2016.

ULTRAZVUČNA KONTROLA Ispitivanje ultrazvukom UT2

2.5. – 12. 5. 2016. 13. 5. 2016.

19. – 29. 9. 2016. 30. 9. 2016.

38

HD

KB

R

Cen

tar z

a ob

razo

vanj

e

HD

KB

R

Cen

tar z

a ce

trifi

kaci

ju

ŠTO TREBA ZNATI O OBRAZOVANJU

Gdje i kako se NDT uključuje u cjelovitu potrebu obrazovanja;Kako i koliko je NDT uključen u sustav formalnog, neformalnog i informalnog obrazovanja;

Odgovor na prvo pitanje i mnogo više saznat ćete u zajedničkom izvješću Vijeća i Komisije za 2015. godinu (2015/C417/04), o provedbi strateškog okvira za europsku suradnju u području obrazovanja i osposobljavanja (ET 2020; Education & Training 2020 – EC strategic framework)

Novi prioriteti za europsku suradnju u području obrazovanja i osposobljavanja proizlaze iz činjenice da je Europa suočena s nizom hitnih zadataka: ponovnom uspostavom stvaranja radnih mjesta i gospodarskog napretka, te postizanjem održivog rasta koji zahtjeva razvijanje konkurentnosti utemeljene na znanju.

Postoje snažni gospodarski argumenti u prilog tome da je obrazovanje i osposobljavanje upravo onaj sektor kojim se pogoduje razvoju. Razvijene zemlje Europske unije i svijeta, te u njima perspektivne tvrtke shvatile su da je ulaganje u ljudski kapital dobro potrošen novac. Kvalitetnim obrazovanjem i osposobljavanjem potiče se promicanje održivog gospodarskog rasta i razvoja, te se u EU ulažu napori za poboljšanjem pristupa kvalitetnom cjeloživotnom obrazovanju (LLL-Long Life Learning).

Jedan od značajnih strateških prioriteta je transparentnost i prepoznatljivost vještina i kvalifikacija. S velikim zadovoljstvom možemo istaknuti da je naša struka - ispitivanje i kontrola kvalitete nerazornim metodama - većim dijelom ispred mnogih drugih kvalifikacija i profesija jer je postigla da se uvjerenja izdana u okviru Centara za obrazovanje i Centara za certifikaciju, nacionalnih udruga članica Europske federacije za nerazorna ispitivanja (EFNDT) priznaju i prihvaćaju kao podloga za posao. Certifikacija u HDKBR-u omogućuje uključivanje naših ispitivača, koji takva uvjerenja posjeduju, u timove i ekipe za ispitivanje diljem svijeta. HDKBR, EFNDT, VT,PT, UT, RT, ET, MT su ključne riječi koje uvjerenje čine prihvatljivim u EU i šire.

Ovaj veliki korak međunarodnog priznavanja uvjerenja, kao što su npr. oni izdani u HDKBR Centru za certifikaciju, moraju učiniti još brojne struke. Međutim, pred nama je novi korak prema okviru EFNDT-a.

Seminari i tečajevi koji spadaju u neformalno obrazovanje u postupku su vrednovanja, ECVET, stimuliranog projektima Europske unije, kao npr. Erasmus+, u kojima sudjelujemo i održavamo razinu kvalitete certifikata izdanih u HDKBR-u.

U ovom broju časopisa HDKBR INFO čestitamo i objavljujemo imena NDT ispitivača koji su stekli važeća i priznata uvjerenja za metode 3. stupnja u sustavu MRA (Multilateral Recognition). U sljedećem ćemo broju objaviti popis važećih uvjerenja 2. stupnja iz metoda nerazornog ispitivanja.

U ovom broju također donosimo definicije u ovom tekstu spomenutih termina; pročitajte i zapamtite ih jer su to pojmovi o kojima će biti sve više rasprava unutar Europske unije, a tiču se nadogradnje u stimuliranju, financiranju i priznavanju obrazovanja unutar EU i šire. (za INFO pripremila prof.dr.sc. Vjera Krstelj, Voditelj HDKBR Centra za obrazovanje)

TEČAJEVI za KVALIFIKACIJU i CERTIFIKACIJU

39

HD

KB

R

Cen

tar z

a ob

razo

vanj

e

HD

KB

R

Cen

tar z

a ce

trifi

kaci

ju

VAŽEĆI CERTIFIKAT 3. stupnja za ISPITIVANJE BEZ RAZARANJA u skladu sa EN ISO 9712 u HRVATSKOJ imaju sljedeće osobe

PREZIME IME MT VT UT PT RT ETANDRIĆ Tomislav x x x x xCVITANOVIĆ Mato - x xCVITANOVIĆ Nikola xDABO Dario xEĆIMOVIĆ Marijana xFIŠER Josip xGRGA Ivan xGRUBER Davor x x xIVKOVIĆ Zvonimir x x x xJAGODIĆ Branko xJARNJAK Fran xJUKIĆ Krunoslav xKALOGJERA Leo xMALJKOVIĆ Damir xMARKEŠIĆ Zoran x x x x xNADINIĆ Berislav x xNIKOLIĆ Zoran x x x xODRČIĆ Vanja xPAPEC Andrija xPERŠIĆ Krunoslav x xPESSACH Josef x x x xPESSACH Amichai x x x x xPETROVIĆ Ivan x xRUMAC Jože xSARAJČIĆ Krešimir x xSEDMAK Florian xSELAKOVIĆ Darko xSMILJANIĆ Petar xSMOKVINA HANZA Sunčana xSOFRONIĆ Goran x xŠAFAR Dražen xŠKRTIĆ Željko xŠOLA Branko x x xŠUĆUROVIĆ Dragan xŠULJAK Anica xTEREK Verica x x x x xZAJEC Zlatko xŽUNAC Dražen xMT - ispitivanje magnetskim česticama VT - vizualno ispitivanje UT - ispitivanje ultrazvukom PT - ispitivanje penetrantima RT - radiografsko ispitivanje ET - ispitivanje vrtložnim strujama

40

ND

T- N

EFO

RM

ALN

O O

BR

AZO

VAN

JE

FORMALNO, NEFORMALNO I INFORMALNO OBRAZOVANJE:

Niže navodimo definicije prema OECD-u (The Organisation for Economic Co-operation and Development) koji promovira politiku poboljšanja ekonomskog i socijalnog stanja ljudi diljem svijeta. OECD omogućuje forume u okviru kojih će vlade djelovati skupa, izmjenjivati iskustva i tražiti rješenja zajedničkih problema. OECD mjeri produktivnost, svjetski trend investicija, analizira i uspoređuje podatke predviđajući buduće kretanje na tržištu sudjelujući također u generiranju normi u širokom području.

Formal learning Learning organized and guided by a formal curriculum in an organised and structured environment (such as in an education or training institutions/HIE) and is explicitly designated as learning process leading to a formally recognized credential such as a high school completion diploma or a degree, and is often guided and recognized by government at some level. Teachers are usually trained as professionals in some way.

U Hrvatskoj Obrazovanje u okviru Sveučilišta i visoko obrazovnih institucija, visokih škola, koledža i sl.

Non-formal learning

Education that is institutionalized, intentional and planned by an education provider. The defining characteristic of non-formal education is that it is an addition, alternative and/or a complement to formal education within the process of the lifelong learning of individuals..Learning embedded in planned activities/occupation or regulated professions activities. This type of education may be led by a qualified teacher or by professionals with more experience. Non-formal learning outcomes may be validated and may lead to certification.

U Hrvatskoj Seminari i tečajevi u koje treba uvrstiti NDT tečajeve

Informal learningIF learning

Learning resulting from daily activities related to work, family or leisure, experiential or incidental/random learning. It is not organised or structured in terms of objectives or time or learning support. No formal curriculum and no credits earned but learning outcomes may be validated and certified.

U Hrvatskoj Potrebno je omogućiti dokazivanje vještina i kompetencija stečenih izvan standardnih okvira obrazovanja - formalnog i neformalnog obrazovanja.

(za INFO pripremila prof.dr.sc. Vjera Krstelj, Voditelj HDKBR Centra za obrazovanje)

FEANI--TERMINOLOGYGLOSSARYFormal/NonFormal/InFor

41