|
Resistance Spot Welding (RSW) is broadly used in automobile sheet metal joining. The quality of the welding spot nugget directly affects the safety of the car. There are two methods for monitoring the quality: destructive testing (DT) and non-destructive testing (NDT). There are several ways to perform the destructive testing, including tensile test, fatigue test, hardness test, delamination peel test, etc. Among all the options, the easiest way is to artificially destroy the welding spot and observe the welding spot by visual inspection. However, not only it is time consuming but also the cost is high. In the case of non-destructive testing, including Ultrasonic testing and thermography testing, etc., expensive equipments and well-trained inspectors is required. In order to reduce the cost and time as well as increase the convenience, Shih-Fu Ling & Lixue Wan(2000[1]) measured the signal of voltage and current during the welding, utilized the measured values to obtain the impedance curve, and used the eigenvalues of the curve to train a neural network, achieved the goal of creating an online monitoring system for welding spot. This thesis used the eigenvalues of impedance curve and impedance differential curve to train the neural network to obtain the online monitoring system, and the system is experimented under different weldment conditions and tip diameters, in order to simulate actual conditions of the production line.
|