An intelligent Model for Defect Prediction in Spot Welding

Abstract

There are more than 30% defect in the spot welding of cars and randomly chosen cars are performed
ultrasound or destructive testing. This makes the process very vulnerable and unpredictable. This results
in huge reworks, productivity, monetary loss and negative impact on brand name.
This research paper presents the prediction of defect using machine learning models and as well forecasting
models in spot welding through optimized methodology. This defect prediction model is useful in
determining the defects that are likely to occur during spot welding. The forecasting model for process
parameters data pattern, trends, etc. helps to identify the link between predicted defects. This model can
evolve and improve over time by considering data from previous phases and history data of the spot welding
cycle. Predicting the defects before testing begins improves the quality of the product being delivered and
helps in planning and decision making for future spot welding.
The optimized defect prediction methodology in spot welding reduces the defects and predicted sample for
testing which reduces the rework and increase the productivity, monetary value and brand name.
The experimental result shows that the spot-welding methodology has shown improvement over existing
spot-welding method. Please see the six-sigma (Fig:13) chart for before and after improvement curve and
value.