We can classify maintenance into three classes: corrective maintenance (CM), preventive maintenance (PM), condition based maintenance (CBM). CBM has more recently been outmoded by predictive maintenance (PredM). This article describes 4 scenarios where A.I. for equipment maintenance becomes meaningful.
Condition-based maintenance is maintenance when the need arises. We perform after one or more indicators show that equipment is going to fail or that equipment performance is deteriorating. Condition-based maintenance will try to maintain the equipment at the right time. CBM is based on using real-time data to prioritize and optimize maintenance resources. When we observe the state of the system, we are doing condition monitoring. So such a process will determine the equipment’s health, and act only when maintenance is actually necessary. Recent developments have allowed extensive instrumentation of equipment, and together with better tools for analyzing condition data, the maintenance personnel will be able to decide when is the right time to perform maintenance on some piece of equipment. Ideally condition-based maintenance will allow the maintenance personnel to do only the needed things, minimizing spare parts cost, system downtime and time spent on maintenance.
Sensor data integrates with machine learning
Telematics enabled sensor data will help us understand equipment degradation process or predict failure based on past failure history or pattern or circumstances in which failure took place. We can use various machine learning methods to detect varieties of failure modes.
- Trend Analysis: will show when the asset data is on an obvious and immediate downward slide toward failure. Typically a minimum of three monitoring points are recommended for arriving at a trend accurately as a reliable measure to find if the condition is deprecating linearly.
- Pattern recognition: Decodes the causal relations between certain type of events and machine failures. For example, we may find that after being used for a certain product run, a component in the asset fails due to stresses that are unique to that run
- Critical range and limits: We can run tests to verify if the data is within a critical range limit (set by professional experience). However we can adopt machine learning schemes to eliminate user intuition for setting these limits.
- Statistical process analysis: We can retrieve existing failure record data (from warranty claims, data archives and case-study histories). We then apply analytical procedures to find an accurate model for the failure curves, for example. We can finally compare against those models to identify any potential failures.
New tech in an old bottle
New mechanisms for analyzing sensors data include configuring big data platforms for data collection, management and analytics and are making inroads into sensor data analytics for PredM. Powered by machine learning and advanced analytics, this can generate several benefits. Here are a few advantages of incorporating A.I. for equipment maintenance or PredM in no particular order.
- Detecting failures in early stages and preventing them
- Finding ‘Remaining life of asset’
- Schedule predictive maintenance
- Maintaining “Right level of inventory” for spare parts
- Evaluate “What if” alternate scenarios
- Determine right warranty period for the assets at the design time
- Predict breakdown
- Notify operator at right time
- Prevent risk of collateral damage and secondary failure
- Prevent high production downtime
- Maximizing equipment life