A serverless automated MLOps framework for scalable industrial predictive maintenance

Machine maintenance is a major hurdle in industries. It has become expensive and time consuming because of traditional maintenance practices, namely reactive and preventive. Industry 4.0 talks about Predictive Maintenance (PdM) which makes use of Machine Learning(ML) along with machine’s real-time sensor data to predict the occurrence of next failure of the machine and its parts. It has not been widely adapted by MSMEs because of the high cost associated with it along with the lack of machine learning expertise required. The proposed system, AUTO-MAINT, is a fully automated cloud-based platform which employs microservices and simplifies the implementation of ML models, increasing the ease of use and reducing the overall time and cost required for the setup of predictive maintenance for MSMEs. AUTO-MAINT offers automation, customization and support in the ML-Ops pipeline, and real-time predictive maintenance capabilities. Serverless architecture enabled scalable processing of heterogeneous sensor data in real-time. Automation of MLOps pipeline for full model lifecycle management without manual intervention. AUTO-MAINT democratizes predictive maintenance by lowering capital entry barriers for MSMEs. Cloud-native platform integrated continuous learning to maintain high prediction accuracy. Flexible MLOps control accommodates both novice and advanced users to maximize system usability.