IIM Indore Develops Model To Predict Machine Failures With High Precision; Promises Smarter Maintenance, Reduced Downtime

Indore (Madhya Pradesh):

In a major stride toward smarter industrial maintenance, researchers from Indian Institute of Management Indore (IIM Indore) and Indian Institute of Technology Kanpur (IIT Kanpur) have developed a cutting-edge predictive model that significantly enhances the accuracy of machine failure forecasts. Published in the journal Quality Engineering, the study represents a leap forward in data-driven maintenance strategies, particularly for high-value manufacturing assets.

The research, co-authored by Prof Pritam Ranjan and Dr Arnab Koley of IIM Indore, introduces a robust statistical approach tailored for injection moulding (IM) machines—a vital component in sectors such as plastics, packaging and automotive manufacturing. The model offers a refined mechanism to predict the timing of equipment failures by analysing the progression of machine states: from “running without alert” to “running with alert” to eventual “failure.”

 “Traditional predictive models often fail to capture the real-time complexity of machine behaviour,” said Ranjan. “Our method models the time intervals between critical machine states using exponential and shifted Poisson distributions, delivering failure forecasts with greater precision than conventional tools like the Cox Proportional Hazard model,” he said.

As industries transition toward digital and automated systems under the banner of Industry 4.0, the need for intelligent maintenance solutions is growing. This model addresses that demand by enabling condition-based maintenance, replacing outdated calendar-based routines. It uses machine learning—specifically Random Forest classifiers—to identify the most influential sensors among the 72 embedded in the IM machines, filtering out noise and ensuring actionable insights for operations teams.

For manufacturing companies, the potential impact is substantial. By offering early warnings of mechanical issues and highlighting the sensors most critical to system health, the model empowers managers to proactively plan maintenance, reduce unplanned downtime, and extend the life of costly machinery. These capabilities are particularly valuable for industries operating with tight production timelines and minimal tolerance for disruptions.

The model was validated using real-world data from 45 operational epochs of IM machines, confirming a significantly stronger correlation between predicted and actual failure times compared to existing methods. Its compatibility with digital twin platforms, industrial IoT frameworks, and cloud diagnostics also makes it suitable for integration into modern smart factory environments.

“This research exemplifies how cross-disciplinary collaboration—in this case, between management science, engineering, and statistics—can solve pressing industrial challenges,” said Koley. “As India accelerates its push toward advanced manufacturing through initiatives like Make in India and Digital India, such innovations are critical to global competitiveness,” he added.

The model not only advances the technical frontier of predictive maintenance but also provides a practical roadmap for manufacturers aiming to optimise operational efficiency, reduce costs and enhance equipment reliability.

“With industries like automotive, FMCG, and pharmaceuticals depending heavily on precision and uptime, the adoption of such predictive frameworks could redefine maintenance strategies across the board,” the researchers said.