Case study: Sensor testing for process monitoring in predictive maintenance

A vehicle manufacturer monitors the life cycle of brakes with acoustic sensors. COSMO CONSULT was commissioned to develop a solution that provides results for the individual measured values, from which conclusions can be drawn on the current maintenance requirements. Motorists can thus benefit from greater comfort and safety.

• Signal processing of raw data
• Modeling the life cycle
• Generalizing the variance of processed sensor data
• Achieving an 80 percent forecasting quality

In the face of rapidly falling costs, it has become economical to equip vehicles and vehicle parts with sensors. The idea of the automotive company was to monitor the life cycle of brakes with acoustic sensors. These sensors measure the elasticity of the material and provide values that can determine the actual wear. COSMO CONSULT was commissioned to develop a solution that would help in determining the remaining life expectancy of the brakes as well as the optimal times for maintenance.

One of the challenges of the project was that the brakes to be measured were not new and the measurement procedure was not carried out uniformly. In addition, the measurement times for the different data sources were different. Therefore, the project team started by synchronizing the data. Within the framework of signal processing, it was possible to extract the most important features from the acoustic sensor for the analysis. These included:

• Fourier transforms
• Asymmetry of peaks
• Signal statistics & impulse metrics
• Power
• Bandwidth.

The synchronization of the data was then followed by interpolation of the elasticity measurements, since the measurement frequency was lower compared to the acoustic sensor. The elasticity of some acoustic information was then modeled with a tangent to reduce all experiments into a uniform standard. Using this standard, a generalized variance pattern was used to predict when brake maintenance would be due. The relatively high error rate of about 20 percent could be explained by the inhomogeneity of the brakes in the experiments. Today, the solution provides an additional service that can be integrated into the acoustic sensor. It enables the user to process measurement data and determine the current status of the brake.

An indicator of the board computer notifies the driver about the need for maintenance as soon as the previously defined limit value is exceeded. Therefore, drivers can benefit from greater comfort and safety because wearing parts like the brakes can be individually maintained according to driving behaviors and usage.