ALTEN was contacted by a leading supplier of technology in the Automotive – Tier 1 industry to design and implement an AI-driven preventive maintenance system  capable of monitoring and detecting anomalies in the behaviour of machines in the Assembly Line in real time. A cloud-based solution was desired that would monitor all the configured machines and alert the relevant personnel and store data to display historical dashboards.

Manufacturing Execution Systems  – MES is a system to encompass all key manufacturing capabilities, including quality control, material synchronization, MES, focuses on shop floor execution, tracking production, work orders, and machine performance & most importantly ensures smooth daily operations, maintenance providing enterprise-wide visibility, control, and synchronization. This central system integrates both, bridging high-level planning with real-time execution.

Challenge: The Organization did not have an automatic means to monitor Production line machinery and needed human intervention to make sure that the machines are within their operational limits as per the guidelines of the Manufacturing Execution Systems (MES).

The following are the requirements of the System to be developed:

• The System needed to monitor sensor data from the machines and correlate it with the historical behavioral patterns and determine if the current behavior qualifies as an anomaly or not.
• The system also needed to provide a dashboard/visualization that displays the current status of all monitored machines.
• The Predictive Engine of the system also should be capable of estimating the RUL –Remaining Useful Life based on the Historical data.
• Finally, a Seamless Integration into the Manufacturing Execution Systems (MES) was desired.

Solutions: Alten implemented a framework that monitors machines based on one parameter (of any type – pressure, voltage, temperature, time etc.), observe trends and flag anomalies. 

The system is based on ML algorithms monitoring univariate data by unsupervised learning. Given data and time, the system “learns” the behavioral patterns of machine and uses the knowledge in flagging anomalies. User/Operator feedback was also considered that enhanced the reasoning ability of the ML Model.

The solution framework is hence extensible to several use-cases plant-wide.

The solution used a SSO mechanism for authenticating users and had a frontend and a web services working with the AI/ML core. MERN stack was used in implementing the solution. The AI/ML core was developed using Python with RabbitMQ servings as the cross-platform bridge.

Benefits:

• Early anomaly detection prevents wastage of products, components and spares
• Reduces production line downtime as defective machines can be isolated and serviced prior to a failure occurring.
• Reduces spare stock quantity due to reduced rejection rate of PCBs.