The procedure for improving the management of the maintenance and repair process using the neural network technology

BACKGROUND: Stable significant degree of wear of on-ground vehicles of transport industry and of agricultural machinery keeps improvement of maintenance management relevant. Development of the procedure for improving the maintenance process using state-of-the-art digital technologies is a relevant technical problem. AIM: Determination of parameters for development of the procedure of making good managing decisions in the process of maintenance and repair of products in conditions of planned and preventive repair system using the neural network technology. METHODS: Simulation of operation of the proposed neural network was performed in the Deductor software. The built model of the neural network contains one hidden layer with 10 neurons. A sigmoidal function was used as an activation function in neurons of the neural network model. Tools and definitions of mathematical statistics and algorithms theory were used for solving the given problems. RESULTS: The cycle variation coefficient is proposed for revealing the necessity of managing impact on the processes of maintenance and repair of products. The proposed values of the coefficient describe stability of product maintenance and repair process. Using these values, the block diagram of the procedure of improving the maintenance process was developed. The tool of the industry 4.0 neural networks was considered. The performed simulation based on the example of vibrational diagnostics of a bearing unit showed that neural networks are capable of defining defects using amplitude-frequency response of a vibration signal that means to interpret the diagnostic information that can be crucial in conditions of expert absence. The scientific novelty of the study lies in presenting the values of the cycle variation coefficient for making managing decisions in maintenance and repair processes, as well as in obtaining the results of simulation of the neural network operation that confirms potential for their use for interpretation of diagnostic information which is presented in a form of various spectrographs. CONCLUSIONS: The practical value of the study lies in the potential of using the proposed neural network for development of the system of diagnostic information analysis in condition of expert absence and using the proposed values of the cycle variation coefficient for decision-making in management of maintenance and repair.