SIGNATURE ANALYSIS OF ACOUSTIC EMISSION DATA OBTAINED DURING PROOF PRESSURE TEST OF 15CDV6 PRESSURE VESSEL

Size: px

Start display at page:

Download "SIGNATURE ANALYSIS OF ACOUSTIC EMISSION DATA OBTAINED DURING PROOF PRESSURE TEST OF 15CDV6 PRESSURE VESSEL"

Darren Lawrence
6 years ago
Views:

1 SIGNATURE ANALYSIS OF ACOUSTIC EMISSION DATA OBTAINED DURING PROOF PRESSURE TEST OF 15CDV6 PRESSURE VESSEL Anto Zacharias, Mechanical Engineering, AmalJyothi College of Engineering, Jeby Philip, Head EXMD, VSSC, Trivandrum, Toney Varghese, Mechanical Engineering, AmalJyothi College of Engineering. Abstract-Acoustic Emission (AE) is a phenomenon where by stress waves are released due to microstructural changes in the material. This technique is a powerful tool to investigate the micro structural changes within a material and is extensively used as a real time NDT method for the structural integrity assessment of hardware made with different types of materials.ae signals captured from a hardware test will be a combination of noise signals and genuine AE signals. Major task for the signature analysis of the AE emission from any specimen is the discrimination of noise signals like grip noises, machines noises from the genuine AE signals from the material corresponding to defect growth and impending failure. The pattern recognition analysis of acoustic emission signals is a widely accepted practice for classifying the various source mechanisms involved by which we can effectively discriminate the genuine signal from the noise. This paper gives the details on the analysis carried out on AE data obtained during the proof pressure testing of 15CDV6 steel alloy pressure vessels using the pattern recognition feature of the AE analysis software Noesis. The pattern recognition by the cluster analysis of the specimen data helped in obtaining the characteristics of AE signals. Key words: Acoustic emission (AE), 15CDV6 steel alloy, Pressure vessel, Signature analysis, Pattern recognition I. INTRODUCTION AE is a transient elastic wave generated by the rapid release of energy from a source within a material [1]. AE is a phenomenon and occurs in materials, structures and processes during loading. A typical AE test setup is illustrated in Fig.1 Fig.1 A typical AE test set-up AE sensors mounted on the structure to detect the displacement of the surface and convert this tiny movement to an electrical signal by using piezoelectric sensors. After sensing and preamplification, the signal is transmitted to the main instrument, where it is amplified and filtered. Detection of the signal is accomplished with a comparator circuit, which generates a digital output pulse for the AE signals exceed a fixed threshold voltage. The threshold level is usually set by the operator and it is the key variable that determines the test sensitivity. [2] Acoustic Emission is used as powerful structural integrity monitoring tool for the pressure tanks. It is made of three cylindrical segments with long seam TIG welded, joined together by means of circular seam TIG weld with tori spherical end domes capped on both sides. The hardware is made of 15CDV6 alloy steel with resin sealed aluminium coating on internal surface. The Acoustic Emission acquisition software AEwin used in this study is manufactured by Physical Acoustics Corporation. [3] During the first Proof pressure cycle, the tank registered number of continuous AE hits of low Amplitude signals from lower pressure onwards. Roll over phenomena [2, 10] was not observed during the pressure hold periods of pressure testing. During maximum expected operating pressure and Proof pressure hold, the tank registered more number of low duration AE hits with low amplitude. These AE signals could be due to epoxy resin sealing above the internal aluminium coating on tank. During repeat proof pressure cycle the hardware followed Kaiser effect[2, 10] up to 110bar.Beyond 110 bar tank registered AE hits of low amplitude and low duration continuous signals including repeat PPT hold. This behaviour also could be due of epoxy resin sealing inside tank. The recorded AE is more complex to analyse for resin coated 15CDV6 material compared to AE analysis of other metallic hardware. At these conditions the resin contents of the vessel can produce some AE activity which must be distinguished from material related emission for structural integrity assessment. Thus, more advanced data treatment techniques are needed. 2855

Pattern recognition (PR) techniques is used as an alternative and or complementary AE data processing technique, aiming to help operators in noise identification and filtering as well as assisting

Pressure tank is used for storing the gas injectant. These tanks are coated internally with pure aluminium and sealed with epoxy 1061 resin for resistance against corrosive action of the injectant.

3 Internal Epoxy Resin Coating The hardware is instrumented with 32 AE sensors of 150 KHz resonant frequency, mainly covering all the T-joints, circular seam and long seam weld joints.

2 Pattern recognition (PR) techniques is used as an alternative and or complementary AE data processing technique, aiming to help operators in noise identification and filtering as well as assisting the overall AE based evaluation.[4] II. EXPERIMENTAL SETUP All pressure vessel specimens used for the tests presented in this thesis were made of 15CDV6 heat treated steel. Pressure tank is used for storing the gas injectant. These tanks are coated internally with pure aluminium and sealed with epoxy 1061 resin for resistance against corrosive action of the injectant. The photograph of hardware and internal epoxy resin coating is shown in figure 2 and 3 respectively. Fig.2 Hardware used for test Fig.3 Internal Epoxy Resin Coating The hardware is instrumented with 32 AE sensors of 150 KHz resonant frequency, mainly covering all the T-joints, circular seam and long seam weld joints. Sensor layout of hardware is shown in figure 4. Data Acquisition Computer 32channels, 12bit vertical resolution 100 MHz bandwidth CPU: Pentium Dual core T4300 (2.1 GHz), 4 GB RAM, 285GB, OS: Windows 7 The tank was subjected tohydraulic proof pressure tested up to 120 bar pressure in following sequence Leak Cycle bar. Proof Pressure Cycle (MEOP) (Proof Pressure)-0bar, with pressure hold in all steps. Repeat Proof Pressure Cycle bar III.EXPERIMENTAL OBSERVATION During the first Proof pressure cycle, the tank registered AE hits of low Amplitude from lower pressure onwards. The tank registered more number of continuous AE hits of Amplitude below 80dB from 20bar onwards. Roll over phenomena was not observed during the pressure the pressure holds.during maximum expected operating pressure (MEOP) and Proof pressure hold also the tank registered more number of low duration AE hits with amplitude below 80dBshown in Fig 5. During repeat proof pressure cycle beyond 110 bar tank registered AE hits of low amplitude below with low duration continuous signals including repeat PPT hold shown in Fig 5. AE emission obtained during hydraulic proof pressure test is combination of genuine AE signals from defects or some dislocation and noise signals. Due to presence of this unwanted signals this hardware do not satisfy the AE testing criteria. AE-13 AE-14 AE-15 AE-20 AE-29 AE-4 Closure Flange AE-2 AE-3 AE-5 AE-12 AE-17 AE-23 AE-24 AE-25 AE-30 AE-31 Closure Flange Pressure IN AE-1 AE-7 AE-8 AE-9 AE-19 AE-21 AE-26 AE-28 AE-32 Fig.5 Amplitude v/s Time graph AE-6 AE-10 AE-11 AE-16 AE-18 AE-22 AE-27 AFT END Additional support provided FOR END Fig.4 Sensor layout AE equipment used in pressure test is shown in Table 1 Table 1 AE Equipment Equipment Specification kHz AE Sensor Dimension:Height:17mm, φ:18mm Pre Amplifier Gain: 40dB, 20kHz-3MHz Fig. 6 Duration v/s Time graph 2856

R,τ,R/τ International Journal of Science, Engineering and Technology Research (IJSETR), Volume 3, Issue 10, October 2014 The low amplitude AE signals are obtained from all around the hardware.

3 R,τ,R/τ International Journal of Science, Engineering and Technology Research (IJSETR), Volume 3, Issue 10, October 2014 The low amplitude AE signals are obtained from all around the hardware. For filtering noise of this type some advanced methods of noise filtering such as pattern recognition is used. Pattern recognition (PR) techniques is an alternative and complementary AE data processing technique, aiming to help operators in noise identification and filtering as well as assisting the overall AE based evaluation of pressure vessels. NOESIS is Advanced Acoustic Emission data analysis pattern Recognition & neural networks software for datasets of Acoustic Emission Applications. discrimination of different failure mechanisms and the remaining noise sources are Channel, Rise time, Counts to Peak, Energy, Amplitude, ASL, Average Frequency, Reverberation Frequency, Partial Power 1, Partial Power 2, Frequency Centroid and Peak Frequency. IV. RESULTS AND DISSCUSSION UNSUPERVISED PATTERN RECOGNITION In unsupervised pattern recognition, acoustic emission data sets are classified in to groups according to their similarity.the first step in unsupervised pattern recognition is to find the number of clusters suited to AE data set. To extract the best suited number of clusters among the AE data sets different criteria is used. The selection of number of clusters is based on the R criterion defined by Davies and Bouldin and τ criteria defined by Bow [5]. According to above criterion the cluster members separate more distinctly for low values of R and high values of τ or the ratio R/τ should be minimum[8] shown in table Number of clusters v/s R,τ,R/τ R τ R/τ Fig. 8 Single link clustering of features As features are not measured in the same range or same units, it is to normalise the data sets before any unsupervised or supervised algorithm is applied. V. THE k MEANS METHOD The k-means method is a classical tool for pattern analysis consisting in unsupervised data set segmentation, which aims at clustering n observations into k clusters, in which each observation belongs to the cluster with the nearest mean[4].the pressure vessel has total 7530 AE hits. K means clustering is done on AE hits from the hardware. The clustered image of the signals from the pressure tank are given below Fig. 7 Number of Clusters V/S R/τ, R, From the above observation it is found that for internally resin coated pressure vessel there is three acoustic emission sources. The next step before application of UPR was to select a representative set of AE features based on which thedata would be segregated into classes, exhibiting similar characteristics. [4]The degree of correlation of features in the data set viewed in the form of a dendrogram (Fig.8 from Noesis software) where the correlated features are joined together at the top. The correlation value less than 0.7 indicate presence of various sources. [7]Based on the above procedure the AE features selected for the Fig. 9 Energy v/s Time graph of coated hardware Fig. 10 Amplitude v/s Time graph of coated hardware 2857

R,τ,R/τ International Journal of Science, Engineering and Technology Research (IJSETR), Volume 3, Issue 10, October 2014 To find the source of AE hits of different classes AE features of uncoated

CLUSTERING OF UNCOATED PRESSURE TANK R criterion defined by Davies and Bouldin and τ defined by Bow are used to find the number of clusters suitable for clustering of uncoated pressure tank [5] shown

11 Number of Clusters V/S R/τ, R, of uncoated pressure tank R τ R/τ From the Figure 12 and 13 it is found that there are two classes of AE signals in which one class have low amplitude, low duration

4 R,τ,R/τ International Journal of Science, Engineering and Technology Research (IJSETR), Volume 3, Issue 10, October 2014 To find the source of AE hits of different classes AE features of uncoated hardware and compared with coated tank. VI. CLUSTERING OF UNCOATED PRESSURE TANK R criterion defined by Davies and Bouldin and τ defined by Bow are used to find the number of clusters suitable for clustering of uncoated pressure tank [5] shown in table Number of clusters v/s R,τ,R/τ Fig.11 Number of Clusters V/S R/τ, R, of uncoated pressure tank R τ R/τ From the Figure 12 and 13 it is found that there are two classes of AE signals in which one class have low amplitude, low duration and low energy and other class have high amplitude, high duration and high energy. This two AE source signals are from hardware itself so high energy and high amplitude class of signals are genuine signals and low energy low amplitude class of signals are global dilation signals of signals. But the coated hardware has three acoustic emission sources. This paper found out the third acoustic emission source and its features. To validate the acoustic emission source of clusters we compare the main features of these two classes of signals from uncoated 15CDV6 hardware and epoxy resin coated 15CDV6 hardware. By analysing AE signals of uncoated and coated hardwires it is found that AE features of two classes are similar shown in Fig 14 and 15. Genuine AE signal Resin Coated Hardware Uncoated Hardware By using above criterion we found the number of clusters provides better the clustering for the uncoated 15CDV6 hardware is two is illustrated in Fig. 11.The clustered data of signals from uncoated 15CDV6 steel tank is given below. Fig. 14 Features of genuine AE band Global Dilation Signal Fig. 12 Energy v/s Time graph of uncoated tank Resin Coated Hardware Uncoated Hardware Fig. 15 Features of global dilation Fig. 13 Amplitude v/s Time graph of uncoated tank From the figure 14 it is found that genuine signal class of uncoated tank and resin coated 15CDV6 tank have almost similar range for AE features. The 2858

duration, counts and energy of these two hardwires are in similar range. The AE features of global dilation class of two hardwires are compared are illustrated in Fig. 15.

the high energy band is due to defects in hardware weld joints and the low energy band is due to uniform dilation of hardware during external loading both datasets are from hardware.

5 duration, counts and energy of these two hardwires are in similar range. The AE features of global dilation class of two hardwires are compared are illustrated in Fig. 15. The counts, duration and energy of two hardwires are found to be in similar range. Therefore these two classes of AE signals are originating from same source and are from the hardware itself.the high energy band is due to defects in hardware weld joints and the low energy band is due to uniform dilation of hardware during external loading both datasets are from hardware. Therefore other class of signals of coated hardware is from the coating. This class of signals is may be due to activities in internally coated epoxy resin. Three clusters of signals from the coated hardware can separated as below figures. Fig. 16 Amplitude v/s Time graph based on source range AE parameter for three classes can be found.by analysing all case studies it can found that there is one class which has most of AE hits. This class have almost 90% of total AE hits and it is from resin coating. This class found to have similar band for AE features like Counts, Energy and Duration. Table 2.Mean value of AE features from resin coating Features Mean Value Amplitude 60 Energy Duration Counts 5-6 VII.SUPERVISED PATTERN RECOGNITION Supervised Pattern Recognition, which involves a learning process and where each new set of data is processed and classified to one of the previously known and predefined groups by comparing its features to a database or using rules derived from the learning process. [4]Supervised method was applied to some data files that are similar in nature to the file that used to crate method. Similar means data have been acquired from a similar physical process and settings of AE hardware were the same. The results of supervised method on other case study are given below. Fig. 17 Energy v/s Time graph based on source From the above discussion three acoustic sources of coated 15CDV6 tanks are validated.of these three classes of signals genuine AE signals can be exported and analysed to assess the structural integrity of pressure tanks. If genuine AE signals satisfy the AE criteria for pressure tanks then integrity of tanks are good. In order to obtain a range for the AE features of different case studies were compared. By comparing AE features we can obtain fixed band for the features like Counts, Energy, Duration, Amplitude, and Rise Time.AE parameters for different classes of epoxy resin coated 15CDV6 hardware over five case studies are compared. Bycompiling the features of many case studies the Fig. 18 Energy v/s Time Fig. 19 Amplitude v/s Time 2859

6 The unsupervised and supervised pattern recognition is compared to validate the clustering by supervised method. Validation of supervised pattern recognition is an important milestone for AE monitoring of pressure tanks to assess structural integrity.ae features Amplitude, Energy, Counts, Duration were selected for comparison.supervised method can effectively separate signals of different sources similar to unsupervised method. So this supervised method can effectively use to filtering of noise by online monitoring. VIII.CONCLUSION Pattern recognition based on clustering methods was applied on the AE data was found beneficial in noise filtering of internally resin coated pressure tanks.ae features of five case studies are compiled to find the range AE features form epoxy resin coating.supervised method can effectively separate signals of different sources and can be effectively used to filter the noise signals due to the micro cracking internally coated of epoxy resin to aid in real time structural integrity assessment. 7.Noesis, NOESIS Advanced Acoustic Emission Data Analysis and Pattern Recognition &Neural Networks Software for Acoustic Emission Applications, User s Manual, Markus G RSause, DanielSchulthei and Siegfried Horn. Acoustic Emission Investigation of Coating Fracture and Delamination in Hybrid Carbon Fiber Reinforced Plastic Structures. J. Acoustic Emission, 26, 1-13, Anastassopoulos, A.A, ApostolosTsimogiannis, Evaluation of Acoustic Emission Signals during Monitoring of Thick-Wall Vessels Operating at Elevated Temperatures. J. Acoustic Emission, 22, 59-70, KKPurushothaman, B Binu, Jeby Philip, S A Pillai and S AIlangovan. Acoustic Emission Monitoring and Evaluation of Nickel Hydrogen Battery Cell Cases, NDE IX. AKNOWLEDGEMENT Authors wish to acknowledge Director, VSSC and Deputy Director, Engineer in-charge, Composites Entity, VSSC for permission to publish this paper. We wish our thanks to engineers and technical assistants of Acoustic Emission Testing Lab, VSSC for their efforts put for completing testing and analysis. X.REFERENCE 1.ASTM Standard Definitions of Terms Relating to Acoustic Emission American Society for Testing and Materials, E610-82, Miller, R. K. and P. McIntire, Acoustic Emission Testing, Vol. 5, 3rd ed.ndt Handbook, PAC, DiSP with AEwin User s Manual Rev. 3. Physical Acoustic Corporation, Dimitrios Kouroussis and Athanassios Anastassopoulos, Unsupervised pattern recognition of Acoustic Emission from full scale testing of wind turbine blade, Proceedings 2nd International Conference on Emerging Technologies in NDT 24-26, Anastassopoulos A.A., PhilippidisT.P, Clustering methodologies for the evaluation of AE fromcomposites, J. Acoustic Emission, Vol. 13(1/2), 11-21, Anastassopoulos. A.A., PhilippidisT.P. Pattern Recognition Analysis of AE from Composites, Proceedings of the 23rd European Conference on AE Testing, pp15-20,

Noesis - Advanced Data Analysis, Pattern Recognition & Neural Networks Software for Acoustic Emission Applications

21. Kolloquium Schallemission Vortrag 12 More info about this article: http://www.ndt.net/?id=21578 Noesis - Advanced Data Analysis, Pattern Recognition & Neural Networks Software for Acoustic Emission