Intelligent machine learning based total productive maintenance approach for achieving zero downtime in industrial machinery

Abstract The unpredicted breakdowns in any industrial plant paves way for huge loss to the industry in terms of production and profit. If near future breakdowns are known well ahead of time, zero downtime can be achieved, maintaining the demand-supply chain which leads to industry-4.0 standard. This paper addresses the challenges involved in such transformation and proposes a monitoring and control procedure to reduce catastrophic breakdown by 84%. The parameters and signatures related to the breakdown phenomenon which are restricting zero down time in industry-3.0 are analysed. Adaptive ARIMA model based machine learning to support adaptive error prediction model through varying windowing technique is proposed to predict the future breakdown before its occurrence by forecasting the important signature parameters in the machinery. The proposed maintenance approach is implemented in a high-pressure hydraulic sand moulding machine in an automotive grey casting manufacturing foundry. In this work, the oil contamination level is the parameter identified for analysis in the high pressure sand moulding line in foundry. The breakdown in minutes per month and the number of breakdown occurrences in a month due to various phenomena are analysed after implementing the proposed approach. The proposed system ensures promising results which increases Mean Time Between-Failure by 800% and thereby achieving zero downtime in industries.