Inverse-PID: A Mathematical Approach Towards Detecting Real-World Wear and Tear in Industrial Machines

Conventional predictive maintenance models can be categorised into physics-based and data-driven approaches. In this paper, we present a novel methodology that combines both approaches by inversely computing the PID (proportional, integral, and derivative) gains. Erratic behaviour of calculated gains indicates wear and tear and processing issues.This novel method exhibits several desirable properties. Firstly, it is grounded in a sound mathematical basis, assessing changes in mechanical systems through corresponding changes in PID control behaviour, thus yielding mathematically traceable results. Secondly, it has broad applicability, as PID loop logic is widely used for motion and machine control in various types of industrial machinery. Finally, it preserves data privacy since the observed gain values of the PID controller do not typically convey sensitive information about the operations of industrial equipment.We evaluated the real-world performance of this technique in a trial using a fleet of 400 large-scale industrial injection moulding machines across multiple operators.