Detecting risk of failure in a powered wheelchair sensor system using artificial neural networks

The Systems Engineering Research Group at the University of Portsmouth has been loading more and more complicated systems onto powered wheelchairs. It is important that these mixes of various complex wheelchair systems are fault tolerant. This paper presents a simple method that normalizes the electronic control sensor data of the modified powered wheelchairs to reflect their fault thresholds, and then uses the normalized data as inputs for a neural network. The network automatically analyses mixed correlations of data about the wheelchair and detects potential risks. Simulation suggests that faults in the powered wheelchair systems can be detected within the data sets.