Remaining Useful Life Estimation with Parallel Convolutional Neural Networks on Predictive Maintenance Applications
Abstract
Maintenance work in the industry is performed as failure based corrective maintenance and calendar based preventive maintenance strategies. These strategies cannot meet the demands of the industry in terms of maintenance costs and production efficiency. Data-based predictive maintenance strategy aim at efficiency in production and optimum point in maintenance works. This study is based on Remaining Useful Life, which is the basis of the predictive maintenance strategy. The data used in the study is the dataset of aircraft engines. The data received from many sensors of the running motor are fixed by sliding window. A new approach has been introduced in the estimation of Remaining Useful Life with the proposed Parallel Convolutional Neural Network. By defining a problem-specific asymmetric cost function, better results have been obtained in terms of sensitivity. © 2020 IEEE.