Applied Neural Network Modelling of Mobile Plant Equipment for use in Predictive Maintenance

The aim of the project is to provide a simple and effective approach and capability to enable predictive monitoring of complex plant equipment containing multiple subsystems using Neural Networks. This study is intended to advance capability of mobile plant maintenance and improve understanding on interfacing intelligent systems with maintenance management practices. It is observed in complex plant equipment that systems and subsystems are interlinked, providing the opportunity to quantify and characterise these connections for performance baselining. It is expected that this baseline imposes performance expectations that the system responds to meet. A literary review highlights existing methodology used in applying Neural Networks to complex equipment and assess applicability for maintenance, adaptability for highly dynamic systems and requires the least amount of system expertise to implement. An experiment was then designed to validate the adopted methodologies. The OBDII (On Board Diagnostics) port was used with an OBDII scanner to record non-proprietary vehicle telemetry over a series of drives. Though the project strives to identify and model system performance expectations to compare with measured performance and provide prognostic and diagnostic value for use in a predictive maintenance regime. The outcome provided working models with surprising convergence, however failed to quantify the parameters’ relationship with the other available telemetry. Alongside this, in efforts to reduce the need for expert knowledge, induced selfjustification issues due to inputting all available parameters for each model highlighting the need for sufficient cause and effect analysis