Shim Bond Coverage Analysis Using Artificial Intelligence

Shim bond coverage analysis is a common practice in brake and pad manufacturing during brake pad development. This analysis is used to assess the quality of a shim bond and quantify it in case of any quality or de-bond issues during production and warranty returns. Currently, the analysis is carried out manually in the industry using a 1:1 template printed on tracing paper, which is placed on the deboned shim to identify bad bonded regions. The bond coverage is then calculated manually based on the data obtained from the template, which is a time-consuming process taking around 15 minutes per pad/shim analysis. To minimize manual work and increase accuracy, artificial intelligence is being used to estimate the shim bonding quality and coverage. The idea is to feed the deboned shim and pad picture to the model and predict the following:

• Whether the bond coverage is good or bad.
• Identify the good/bad and unnecessary regions on the shim/pad for bond coverage analysis.
• Finally, provide a bond coverage percentage with a pass/fail criterion.

The training dataset was prepared with good and bad coverage information, and a convolutional neural network model was selected to predict the bond coverage. This paper discusses in detail how the model was trained and deployed to analyze shim bond coverage using visual AI.