8.1 Conclusions of the present work
The motivation for the current research work is to develop methodologies that can facilitate monitoring of FSW process with less time, reliability, and appreciable accuracy with less human intervention. With this motive for the present research work is divided in to two main groups. The first one is to develop methodologies for accurate monitoring of weld quality in FSW process and second to develop accurate strategies for identification of internal defects in friction stir welded samples with inconsiderable post processing time. In the first phase of the research work experiments are conducted over wide range of tool rotational speed, welding speed and shoulder diameter. Welding experiments are conducted on AA1100 aluminum alloy in butt joint configuration in a vertical knee type milling machine modified for FSW process. A new forces and torque measurement setup for FSW process is developed in the present study. The welded samples are prepared for mechanical testing as per ASTM E8M standards and tested for measurement of UTS, yield strength and percentage of elongation of the joints. The next phase of research work started with acquisition of main spindle motor current signal, tool rotational speed signal, welding motor current signal, vertical force signal, torque signal and temperature signal during the welding process. The process of acquisition is followed by processing of these signals for effective features extraction. In the processing of acquired signals wavelet transform, Hilbert-Huang transform, and fractal theory have been implemented. Later, extracted signal features are used for developing different methodologies for weld quality monitoring and internal defect identification in FSW process. The salient findings and conclusions of the present research work are highlighted as follows.
Chapter 8
A strain gauge based force and torque measurement setup has been developed for FSW process. The installation of the developed setup with the existing FSW machine is comparatively easy compared to other force or torque measurement systems. The developed setup can perform to a maximum tool rotational speed of ~11000 rev/min. The developed setup has maximum working limit for vertical force, transverse force and torque are 15 kN, 5kN and 40 N-m with measurement accuracies of 98.73%, 98.94% and 99.04%, respectively.
A novel method for selection of suitable mother wavelet function for WPT framework has been developed. Comparison of the developed method with the already published methods revealed that the proposed method is free from process dependencies. The method used a ratio between the energy of the signal and entropy of the wavelet packets for determination of suitable mother wavelet function.
The limitation of wide frequency band in the intrinsic mode functions computed from HHT is eliminated by combining the HHT with WPT. The combined WPT-HHT method can offer better visualization of decomposed signals over narrow frequency band which is advantageous for observing sharp events over less duration of time.
Features from main spindle motor and welding motor current signals, vertical force and torque signals and tool rotational speed signals are estimated using WPT, WPT-HHT, DWT and fractal theory, respectively. Features are fused with BPNN, RBFNN and SVR models in order to develop models for prediction of UTS of the joints. In all the cases SVR outperform prediction of UTS with maximum accuracy of 99.55% with vertical force signal features and minimum of 97.85% with current signal features. On the other hand maximum and minimum prediction accuracy of BPNN model is found to be 98.42% with vertical force signal features and 94.13% with current signal features.
Identification of internal defects in welded specimens has been successfully achieved with features of vertical force signal, torque signal, tool rotational speed signal and temperature signal. Three new indicators namely, defect index ( ), rate of change of temperature ( ) and wavelet based indicator ( ) are developed for identification of defective welds. The indicator is
Conclusions and future scopes
developed from torque signal features, and and are developed form temperature signal features. Along with these indicators, instantaneous phase and frequency of vertical force signals computed using WPT-HHT are also presented as effective features for identification of internal defects. Fractal dimensions computed from tool rotational speed signal using Higuchi’s algorithm also provided an insight for identification of internal defects.
The present study also develops a novel method based on top surface images of the welds for monitoring the UTS of the joints. Two methodologies have been proposed for processing of images to extract suitable information using fractal theory. The developed methodologies are compared with 2D wavelet transform and similar results have been obtained. The proposed work delivers a simple yet effective approach for monitoring UTS of the joints with inconsiderably less post processing time.
8.2 Future scopes of the present work
The research work presented in this thesis deals in development of different strategies for monitoring of FSW process. Real time signals are processed with developed methodologies and effective features are extracted from signals and correlated to weld qualities in terms of strength of the joints. Apart from the weld quality monitoring, the current research work also develops methodologies for accurate identification of internal defects in the friction stir welded samples using different signal features. However, the work presented can be extended further and following are the possible future scopes.
The developed weld quality monitoring methodologies only tested for UTS of friction stir welded joints. However, there are other quality attributes such as yield strength, ductility, hardness and bend strength can be tested with the developed methodologies.
One of the major contributions of the present research work is the development of strategies for identification of internal defects in friction stir welded samples.
These strategies can be extended further for characterization of internal defects for its size estimation, orientation and location of occurrence.
Chapter 8
The image processing methodology developed in the current research work can be extended to test thermographic images of welded samples for monitoring of weld quality and identification of defects in the welded samples.
Suitable hardware and software integration can be developed so that methodologies developed can be extended towards hardware realization for actual industrial implementation.