Artificial Neural Network Modelling for Prediction of

0 downloads 0 Views 444KB Size Report
Keywords: Austenitic stainless steel welding; Ultrasonic testing ... problems in the inspection of austenitic welds is the presence of large spurious signals, ... Download Date | 5/31/18 2:29 AM ... occurred in the weld. At the end of the process, defect-free ..... Phased Array Technology to Examine Austenitic Coarse Grained.
2017; 1 (2): 122–135

Journal xyz2018; 2017; 116: (2):511–515 122–135 Open Chem.,

Research Article Memduh Kurtulmuş*

The First Decade (1964-1972) Research Article

Open Access

Max Musterman, Paul Placeholder

What Modelling Is So Different About Artificial Neural Network for Prediction Neuroenhancement? of SNR Effected by Probe Properties on Ultrasonic Was ist so anders am Neuroenhancement? Inspection of Austenitic Stainless Steel Weldments Pharmacological and Mental Self-transformation in Ethic Journal xyz 2017; 1 (2): 122–135

https://doi.org/10.1515/chem-2018-0056 The First Decade (1964-1972) received January 24, 2018; accepted March 27, 2018.

Research Article

Comparison Pharmakologische und mentale Selbstveränderung im 1 Introduction ethischen Vergleich

Abstract: Many austenitic stainless steel components are Austenitic steels are non-magnetic stainless steels that https://doi.org/10.1515/xyz-2017-0010 used in the construction of nuclear power plants. These contain high levels of chromium and nickel and low levels received February 9, 2013; accepted March 25, 2013; published online July 12, 2014 Max Musterman, Paul Placeholder components are joined by different welding processes, and of carbon. Known for high resistance to corrosion, high Abstract: In the concept of the aesthetic formationand of knowledge its as soon radiation damages occur in the welds during the service strength, high creep properties excellentand formability, as possible and success-oriented application, insights and profits without life of the plant. The plants are inspected periodically austenitic steels are the most widely used grade the of reference to the arguments developed around 1900. The main investigation also with ultrasonic test methods. Many ultrasonic inspection stainless steel [1]. Austenitic stainless steels are used includes the period between the entry into force and the presentation in its current problems arise due to the weld metal microstructure extensively at nuclear power plants as the main material version. Their function as part of the literary portrayal and narrative technique. of austenitic stainless steel weldments. The present of construction for process vessels and pipework. Stainless Keywords: transmission,are investigation, principal, period research was conducted in order to describe the affectsFunction, steel assemblies visible throughout any particular Pharmacological and Mental Self-transformation in Ethic of probe angle and probe frequency of both transversal plant. Austenitic stainless steel parts are joined by Dedicated to Paul Placeholder Comparison and longitudinal wave probes on detecting the defects of various welding processes. The atoms of austenitic steels Pharmakologische mentale austenitic stainless steel und weldments. FeedSelbstveränderung forward back have aim face centered cubic structure at all temperatures. propagation neural network (ANN) models have Therefore, the macrocrystalline structure of an austenitic ethischenartificial Vergleich 1 Studies and Investigations been developed for predicting signal to noise ratio (SNR) of weld is established when it solidifies and the austenitic transversal and longitudinal wave probes. InputThe variables phase forms long columnar dendritic grains, which grow https://doi.org/10.1515/xyz-2017-0010 main investigation also includes the period between the entry into force and received 9, 2013; March 25, published onlinealong July 12, 2014 that affectFebruary SNR output inaccepted these models are2013; welding angle, directions ofTheir maximum heat lossof during cooling. the presentation in itsthe current version. function as part the literary porprobe angle, probe frequency and sound path. Ofand thenarrative Thesetechnique. coarse dendritic grains have extreme anisotropic trayal Abstract: In the concept of the aesthetic formation of knowledge and its as soon experimental data, 80% is used for a training dataset properties [2]. as possible and success-oriented application, insights and profits without the and 20% is used for a testing dataset with 10 neurons in Ultrasonic testing is the dominant non-destructive reference to the arguments developed around 1900. The main investigation alsoNational Taiwan Ocean University, 2 Pei-Ning Musterman: Institute of Marine Biology, hidden layers in developed ANN models. Mean*Max absolute testing process utilized to detect under surface defects Road Keelung 20224, Taiwan (R.O.C), e-mail: includes the period between the entry into force and the presentation in its [email protected] error (MAE) and mean absolute percentage error (MAPE) in nuclear power plants and Paul Placeholder: Institute of Marine Biology, National [3]. TaiwanThe Oceanheterogeneous University, 2 Pei-Ning version. Their function as part of the literary portrayal and narrative technique. Road Keelung 20224, Taiwan (R.O.C), e-mail: [email protected] types are calculated as 0.0656 and 16.28%, respectively, anisotropic structures exhibited by austenitic steel toKeywords: predict performance of ANN models in a transversal welds cause difficulties in the interpretation Function, transmission, investigation, principal,multi-pass period Open Access. © 2017 Mustermann and Placeholder, published by De Gruyter. This work is licensed under the of Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 most License. serious wave probe. In addition, MAE and MAPE are calculated ultrasonic testing results [4]. One of the asDedicated 0.0478 and 18.01%, respectively, for performance in a problems in the inspection of austenitic welds is the to Paul Placeholder longitudinal wave probe. presence of large spurious signals, which have been attributed to the characteristic acoustic impedance Keywords: Austenitic stainless steel welding; Ultrasonic mismatch existing between the metals that make up the 1 Studies and Investigations testing; Signal-to-noise ratio; Artificial Neural Networks weld [4]. The characteristic large, mostly orientated and PACS – 07.05.Mh, 06.60.Vz, 81.20.Vj. dendritic weld grains lower the ultrasonic sound velocity The main investigation also includes the period between the entry into force and and distribute the sound in multiple directions, causing the presentation in its current version. Their function as part of the literary porthe sound beam to twist. [5]. The grain boundaries, the trayal and narrative technique. weld root and the weld fusion line also cause scattering of the sound [6]. These factors lead a poor signal to noise ratio and a low defect detection ability [7]. Several attempts *Max Musterman: Institute of Marine Biology, National Taiwan Ocean University, 2 Pei-Ning have been made to resolve this inspection problem. New *Corresponding author:Taiwan Memduh Kurtulmuş, University, Road Keelung 20224, (R.O.C), e-mail:Marmara [email protected] probes have been produced [8] and some improvements Applied Science High School,ofIstanbul, Turkey, E-mail: memduhk@ Paul Placeholder: Institute Marine Biology, National Taiwan Ocean University, 2 Pei-Ning marmara.edu.tr have been achieved in ultrasonic testing methods [9-12] Road Keelung 20224, Taiwan (R.O.C), e-mail: [email protected]

What Is So Different About Neuroenhancement? Was ist so anders am Neuroenhancement?

Open Access. Access. ©©2017 Mustermann Placeholder, published Gruyter. This is is licensed under the Creative Commons Open 2018 Memduh and Kurtulmuş, published by by DeDeGruyter. Thiswork work licensed under the Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 License. Unauthenticated Attribution-NonCommercial-NoDerivatives 4.0 License. Download Date | 5/31/18 2:29 AM

512 

Memduh Kurtulmuş

permitting higher signal to noise ratios and higher defect detection abilities in the inspection of austenitic stainless steel welds. The intention of this study was to reveal the effects of probe properties on ultrasonic testing of austenitic weldments. The influences of probe type, sound length and probe frequency were investigated. From the test results, we developed feed forward back propagation artificial neural network (ANN) models for the prediction of signal to noise ratio (SNR) of transversal and longitudinal wave probes.

2 Materıals And Methods A 20 mm thick SAE 304L austenitic stainless steel plate was used as the test case. The chemical properties of the plate are presented in Table 1. Four 250x150 mm welding test pieces were obtained from the plate by the laser cutting process. In the cutting operation the longitudinal direction of the piece was chosen cautiously in order to make the welding application parallel to the rolling direction of the plate. Standard 45° and 60° single-V weld groove workpieces [13] were obtained at a milling machine. The plates were welded by the shielded metal arc process (SMAW). ASP 308L electrodes were used in welding. Welding operations were done in the flat position without preheating. During welding each pass of the weld was controlled by the liquid penetrant test. The aim of this examination was to reveal any crack that might have occurred in the weld. At the end of the process, defect-free full penetration butt welds were obtained. After finishing the welding operation 1mm diameter and 20 mm deep holes were drilled in the middle of the weld metal. The welds were inspected by the ultrasonic method using a KRAUTKRAMER USM 25S detector, which is a digital/analog pulse echo flaw detector. Prior to the inspection, the workpiece surface was lightly greased. The probes used in the experiment and their properties are shown in Table 2. In the tests the reference display was chosen as 40%. The signal to noise (S/N) ratio was calculated for each test. The measured S/N ratio gives the defect detection ability of a probe [14]. A high S/N ratio shows a high detect ability. In each test the sound path length was displayed directly on the screen of the detector. The butt joint, the drilled hole and position numbers of the ultrasonic test probes are schematically shown in Figure 1. The numbers indicate the probe described in Table 2. Ethical approval: The conducted research is not related to either human or animals use.

Table 1: Chemical composition (% of mass) of the SAE 304L test plate. C

Si

Mn

Cr

Ni

S

P

0,03

0,65

1,33

19,15

10,76

0,01

0,03

Table 2: The properties of the ultrasonic test probes. Probe Position

Mode

Frequency (MHz)

Probe Angle

1

Tranversal

2

45

2

Tranversal

2

70

3

Tranversal

4

45

4

Tranversal

4

70

5

Longitudinal

1,8

45

6

Longitudinal

1,8

70

7

Longitudinal

4

45

8

Longitudinal

4

70

Figure 1: The probe positions in the tests.

3 Results And Dıscussıon The results of the ultrasonic tests using the transversal probes are shown in Table 3. The test results illustrate that S/N ratio decreases with the increase in sound path length. The scattering increases with the length of the sound path, a factor that lowers the S/N ratio [1]. The 4MHz probes at similar sound path lengths gave higher S/N ratios than the 2MHz probes. The results of ultrasonic tests using the longitudinal probes are shown in Table 4. These test results illustrate that the S/N ratio decreases with the increase in sound path length. These results are in good agreement with the transversal probe results. The 1.8MHz probes at similar sound path lengths gave higher S/N ratios than the 4MHz probes. These results are contrary to the transversal probe results. Transversal 4MHz probes showed better defect detection than longitudinal 4MHz probes. Similar results were obtained in ultrasonic testing of austenitic stainless steel welds [4, 14].

Unauthenticated Download Date | 5/31/18 2:29 AM

 513

Artificial Neural Network Modelling for Prediction of SNR Effected by Probe Properties on Ultrasonic... Table 3: Results of Ultrasonic Tests using Transversal Probes. Groove Probe Angle Position

Probe Angle

45°

1

45°

4.0 MHz Frequency Probe Sound S/N Path Ratio 15.08 16.5

45°

2

45°

15.90

16.0

15.42

12.5

60°

1

45°

17.36

15.5

16.94

12.0

60°

2

45°

18.02

15.0

17.33

11.5

45°

3

70°

25.77

15.0

23.09

11.0

45°

4

70°

26.46

14.0

23.85

11.5

60°

3

70°

28.91

14.5

28.97

10.5

60°

4

70°

29.21

13.0

29.64

10.5

45°

5

45°

41.67

11.5

42.07

9.5

45°

6

45°

43.74

12.5

43.56

9.0

60°

5

45°

44.80

11.0

44.97

8.5

60°

6

45°

46.96

10.0

45.73

9.0

45°

7

70°

81.75

9.0

82.98

7.5

45°

8

70°

81.96

9.5

83.18

8.0

60°

7

70°

83.70

9.5

86.39

7.0

60°

8

70°

84.28

8.5

88.61

6.5

2.0 MHz Frequency Probe Sound S/N Path Ratio 14.63 13.0

Layer

Probe Characteristics

Transversal

Longitudinal

Inputs

Probe Angle

45

60

45

60

Probe frequency (MHz)

2

4

1,8

4

Sound Path (mm)

15,08 – 86,61

13.64-92.34

Outputs

SNR (dB)

6.5-16.5

3.5-16.5

Table 4: Results of Ultrasonic Tests using Transversal Longitudinal Probes. Groove Probe Angle Position

Probe Angle

45°

1

45° 60°

Table 5.: Test probes and sound path length of tests.

45°

4.0 MHz Frequency Probe Sound S/N Path Ratio 15.25 15,0

1.8 MHz Frequency Probe Sound S/N Path Ratio 13,64 12.5

2

45°

15.40

15.5

13.77

12.0

1

45°

15.82

15.0

15.41

16.5

60°

2

45°

16.08

14.0

15.91

16.0

45°

3

70°

26.18

12.0

29.93

12.5

45°

4

70°

26.43

10.5

30.08

13.5

60°

3

70°

26.95

11.5

34.73

12.5

60°

4

70°

27.38

11.5

36.38

12.5

45°

5

45°

43.14

8.0

42.38

12.5

45°

6

45°

44.08

7.5

43.10

12.0

60°

5

45°

45.72

6.5

45.03

11.5

60°

6

45°

46.14

7.0

45.88

12.0

45°

7

70°

86.83

3.5

87.75

10.5

45°

8

70°

87.44

5.0

88.40

10.0

60°

7

70°

88.17

4.0

91.20

10.5

60°

8

70°

89.22

4.0

92.34

9.5

4 Developed Artificial Neural Network Models ANN models have many applications in the optimization of welding parameters and analysis of quality control specifications [16, 17]. In this study, two ANN models were developed for both transversal and longitudinal probes, These models included four input variables: Welding Angle, Probe Angle, Probe Frequency and Sound Path. Feed forward back propagation ANN models have been used to predict the SNR. Since SNR depends on transversal and longitudinal wave probes, variables were changed parametrically according to data presented in Table 5. Two ANN models for transversal and longitudinal wave probes were developed to predict SNR values in this study, and their prediction performances were compared. Both ANN models included 10 neurons in hidden layers. Thirtytwo rows of data were obtained from the experiments and 80% of this dataset (26 rows of data) was used for training data and 20% (six rows of data) is used for validation of developed models. The input-output data can be actual or normalized. It is clear that using normalize data lead to better results. Normalized training and testing datasets of laboratory experiments are calculated using Equation (1). X = (Xi-Xmin) / (Xmax-Xmin) (1) X = Normalized data Xi = Actual data Xmin = Minimum value of actual data Xmax = Maximum value of actual data In this study, the feed forward back propagation ANN model was preferred. The reasons for using the feed forward back propagation ANN model for multi-layered ANNs are that it is a global approximator and that it was the best performing ANN model under current conditions. Levenberg Marquardt was used as a training algorithm in feed forward back propagation ANN models that were developed. The Gradient Descent with Momentum (GDM) learning algorithm was applied as the learning algorithm utilizing Matlab software. Variables were normalized

Unauthenticated Download Date | 5/31/18 2:29 AM

514 

Memduh Kurtulmuş

Figure 2: Structure of a developed ANN models. Table 6: MAE and MAPE Values of 2 Developed ANN Models. Wave probes

Transversal

Longitudinal

Ratio of training dataset

80%

80%

Ratio of testing dataset

20%

20%

Number of neurons in hidden layer MAE

10

10

Figure 3: Regression Analysis Results of two Developed A

CONCLUSION

In this study, we found that an extended sound path decrea MAPE (%) 16.2855 18.0118 addition, the defect detection characteristics of transvers Figure 3: Regression Analysis Results of two Developed ANN Models. Figure 3: Regression Analysis Results of two Developed ANN Models. frequency, while the defect detection characteristics of lo between 0-1; therefore, the LOGSIG (Log-sigmoid) transferprobe frequency. Of the two probe types, the longitudina tests3. of austenitic function was preferred for the developed ANN models. ultrasonic models areinspection shown in Figure MAE, MAPE andstainless MSE (R2) steel wel As a result of tests and analysis, the optimum topologywas values show that the prediction performances of these developed. CONCLUSION 0.0656

0.0478

of a network has been obtained with a specific number of models were successful. epoch which is equal with 300. ANN models that haveConflict of interest: Authors state no conflict of interest. been developed consisted of four-neuron- input layers that represent inputs. The hidden layer is made of 10 neurons, In this study, we found that an extended sound path5decreased the detection of the discontinuities. In Conclusion and the output layer is made of one neuron. A structure addition, the defect detection characteristics of transversal wave probes increased with the probe which represents ANN’s input, output and hidden layers REFERENCES In this study, we found that an extended sound path frequency, while the defect detection characteristics of longitudinal wave probes decreased with the are displayed in Figure 2. The ANN model thus developed decreased the detection of the discontinuities. In addition, has been run with the described properties. probe frequency. Of the two probe types, the longitudinal probes were found to be superior the defect wave detection characteristics of transversal wave in As the last step of the study, two different ANN probes increased with the probe frequency, ANN while the ultrasonic inspection tests of austenitic stainless steel weldment. Finally, a successful model models are compared with the actual values after training [1] Kurtulmus M., Buyukyildirim G., Yukler A.I., Op defect detection characteristics of longitudinal wave wasprocesses developed. have been completed. During the comparison, probes decreased with the probe frequency. Of the two data MAE (Mean Absolute Error) and MAPE (MeanButt Welded SAE 304L Austenitic Stainless Steel, First As the last step of the study, two different ANN models are compared with the values after probe types, theactual longitudinal wave probes were found As the last step of the study, two different ANN models are compared with the actual values Conflict of Percentage interest: Authors state no conflict of interest. Romania, 2006.after Absolute Error) completed. were selected as the comparison, type ofTimisiora, training MAE Absolute be superior in ultrasonic inspection tests of austenitic training processes processes have have been been completed. During During the the comparison,todata data MAE (Mean (Mean Absolute Error) Error) error with the help of Eq. (2) and Eq. Error) (3) for validation of the type of error with the help of Eq. and stainless Finally, and MAPE MAPE (Mean (Mean Absolute Absolute Percentage Percentage Error) were were selected selected as as the type ofsteel errorweldment. with the help of Eq.a successful ANN model [2]are given Kemnitza P., Richtera U., Klüberb H., Nuc. Eng. D the ANN models. Resulting data are given in Table 6. data (2) in (2) and and Eq. Eq. (3) (3) for for validation validation of of the the ANN ANN models. models. Resulting Resulting data was are given in Table Table 6. 6. developed. ∑ REFERENCES ∑

∑ ∑

(2) [3] Moysan J., Apfel A., Comeloup G., Chassignole (2) (2) Conflict of interest: Authors state no conflict of interest.

B.

85.

(3) (3) (3)

[4] of samples. Erhard A., Lucht B., Schulz E., Montag H. J., W where At is actual data, Ft is forecast at tt and nn is the where Kurtulmus At FtFt isBuyukyildirim forecast at time time and isYukler theisnumber number of samples. [1] where M., G., Optimum Ultrasonic Inspection Conditions of At is is actual actualdata, data, is forecast at time t and n the A.I., References Journ., 2000, 5, 1-5. of samples. Buttnumber Welded SAE 304L Austenitic Stainless Steel, First South East European Welding Congress, Comparisons between experimental and predicted [1] Kurtulmus M., Buyukyildirim G., Yukler A.I., Optimum [5] Ultrasonic Chassignole B., Doudet L., Dupond Timisiora, Romania, 2006. Table and MAPE Values Developed ANN Inspection Conditions of Butt Welded SAE 304L O., Fouque T Table 6: 6:ofMAE MAE MAPE Values of ofof22 two Developed ANN Models. Models. values the and output variable developed ANN Ultrasonic Inspection of Austenitic Stainless Steel Welds, Transversal Wave probes probes Transversal Longitudinal [2] Wave Kemnitza P., Richtera U., KlüberbLongitudinal H., Nuc. Eng. Design, 1997, 174, 259-272. France, 2005. Ratio 80% 80% Unauthenticated Ratio of of training training dataset dataset 80% 80% Download Date | 5/31/18 AM 80, 77[3] Ratio ofMoysan J., Apfel A.,20% Comeloup G., Chassignole B., Int. J. Press. Vessels Piping,2:29 2003, 20% [6] Connolly G.D., Lowe M.J.S., Rokhlin S.I., Temple Ratio of testing testing dataset dataset 20% 20% 85.

Artificial Neural Network Modelling for Prediction of SNR Effected by Probe Properties on Ultrasonic...

[2] [3] [4] [5]

[6] [7]

[8] [9] [10] [11] [12] [13] [14]

[15] [16] [17]

 515

Austenitic Stainless Steel, First South East European Welding Congress, Timisiora, Romania, 2006. Kemnitza P., Richtera U., Klüberb H., Nuc. Eng. Design, 1997, 174, 259-272. Moysan J., Apfel A., Comeloup G., Chassignole B., Int. J. Press. Vessels Piping, 2003, 80, 77- 85. Erhard A., Lucht B., Schulz E., Montag H. J., Wüstenberg H., Beine U., Non-Destr. Test. Journ., 2000, 5, 1-5. Chassignole B., Doudet L., Dupond O., Fouque T., Richard B., New Developments for the Ultrasonic Inspection of Austenitic Stainless Steel Welds, European Nuclear Conference, Versailles, France, 2005. Connolly G.D., Lowe M.J.S., Rokhlin S.I., Temple J.A.G., AIP Conf. Proc., 2008, 975, 1026- 1033. Anderson M.T., Cumblidge S.E., Steven R., Applying Ultrasonic Phased Array Technology to Examine Austenitic Coarse Grained Structures for Light Water Reactor Piping, 3rd EPRI Phased Array Inspection Seminar, Seattle, U.S.A., 2003. Kawanami S., Kurokawa M., Taniguchi M., Tada Y., Mitsubishi Heavy Industr. Tech. Review, 2001, 38, 121-125. Praveen A., Vijayarekha K., Abraham S.T., Venkatraman B., Ultrasonics, 2013, 53, 1288-1292. Moles M., Cancre F., AIP Conf. Proc., 2002, 615, 855. Schmerr L.W., Fundamentals of Ultrasonic Phased Arrays, Elsevier, London, U.K., 2015. Chardome V., Verhagen B., Non-Dest. Test. Cond. Monitor., 2008, 50, 490-492. AWS D1.1/2015 Structural Welding Code-Steel, AWS, 2015. Kurtulmus M., Fidaner O., Yukler A.I., Testing of Repair Welded Boiler Pipes of a Thermal Power Station, 13th International Energy Conference, Istanbul, Turkey, 2007. Kurtulmus M., Yukler A.I., Sci. Res. Essays, 2011, 6, 305-312. Ozcanli Y., Kosovali C., Beken M., ACTA Physica Polonica A, 2016, 130, 444. Tekin H.O., Manici T., Altunsoy E.E., Yilancioglu K., Yilmaz B., ACTA Physica Polonica A, 2017, 132, 967.

Unauthenticated Download Date | 5/31/18 2:29 AM