In the quest for “Industry 4.0”, “Smart Manufacturing” and “Connected Manufacturing” many manufacturers have shortlisted predictive maintenance as their strategic initiatives. While the manufacturers target the predictive analytics, it is important to understand how ready we are to move to predictive space. While implementing internet of things (IoT) in their premise, there are few parameters manufacturers should ask themselves to be predictive maintenance ready and be able to derive the benefits from the framework of connected manufacturing.
Below are some of the key question every manufacturer should assess before onboarding the Industry 4.0 model and institutionalise predictive maintenance.
Has the organisation set the expectation from predictive maintenance?
The first and most significant step is to accurately define the management problem for which predictive maintenance can be the answer. Management problem would lead to the operational success & failures. Thus defining failure becomes critical for predictive maintenance scenarios. For instance, most of the industrial equipment comes with the fail-safe mechanism. The instances of complete system failure are rare or ruled out. So failure could be defined as equipment tripping or low/faulty output depending upon the machine. Post-failure definition, it is critical to identify the related parameter. This may be beyond machine parameters, which could include the operational parameter. Failure or anomaly limits would need to be defined for each parameter. This would give multiple permutations and combinations of multiple failure scenarios.
The questions here are: “Are these failure scenarios identified?” and “What features & functionalities are needed from predictive maintenance to foresee these failure scenarios?”
Is the organisation data ready?
Data in manufacturing setup can be complex. Data velocity may vary from 10 data points / millisecond to 1 data point/day. Data may be structured coming from machines or may be unstructured text coming from logbooks, equipment manuals or applications. One of the major challenges is to relate the application data, which has latency, to the machine data which is real-time. Other data changes include, the failure data and its validity to be used for predictive maintenance.
The question that arises for an organisation is: “Do we have a single source of integrated real-time machine data needed for prediction?”
Has the organisation done the cost-benefit analysis?
Predictive solutions are a costly affair. This may involve device play, platform play and advance analytics. Thus it is critical for organisation to assign the KPIs for each failure scenarios identified which it plans to address through the predictive maintenance. This should also consider the initial low accuracy of prediction and contingency arising from the wrong predictions.
The questions here are: “Have we done cost benefits analysis for predictive maintenance solutions?” and “What is the estimated threshold limit & payback period for these predictive solutions?”
Is the organisation operational & maintenance team trained to read predictive insights?
The road to predictive goes from descriptive. Leap-frogging to predictive can lead to the knowledge gap. To start with, it is important to train the operational personnel on descriptive analytics coming from current data. Setup a clear demarcation of the descriptive, predictive and prescriptive. Educate them on value of descriptive and predictive scenarios. Finally, not all prediction leads to failure. It is critical to train the operational personnel on the predictive solution. So the team has clear expectation from the predictive solution and can take a call on the false alarms.
The question here is: “Are organisation personnel trained for the predictive solution and use these predictive solutions for increasing operational efficiency?”
Industry 4.0 is the next evolution in the manufacturing industry and every manufacturer will have to enable connected or smart manufacturing sooner and later. The process though is not as easy as it seems and the team responsible for digitalising the process needs to be cognizant of the ideal framework or model to be deployed, including precise research and meticulous planning to have a successful predictive maintenance implemented.
The author is a domain specialist at Happiest Minds Technologies. Across his career, he has explored multiple innovative & new technology and its implication on manufacturing systems.