A predictive maintenance system for wireless sensor networks: a machine learning approach

Abstract

Predictive maintenance system (PdM) is a new concept that helps system operators evaluate the current status of their systems, and it also assists in predicting the future quality of these systems and scheduling maintenance action. This paper proposes a PdM model that utilizes machine learning to predict the system’s operational status after M active steps based on L previous observations implemented by a feedforward neural network (FFNN). We use quantization and encoding schemes to reduce the complexity of the system. We apply the proposed model to build a PdM system for wireless sensors networks (WSNs), where our concern is to predict the state of the system as far as the quality of data transfer is concerned. The FFNN provides a forward prediction of the operational status of the network after M consecutive time steps in the future, based on the previous L readings of quality of service (QoS) requirements of WSN. We also demonstrate the relation between complexity and accuracy. We found that larger M leads to higher complexity and larger prediction error, where larger L entails higher complexity and smaller prediction error. We also investigate how quantization and encoding can reduce complexity to implement a real-time PdM system.