Predictive maintenance is one of the most obvious applications of big data and analytics in an industrial environment. The reason seems obvious. Manufacturing companies capture a lot of data from their machines. Until now they have found little use for them. Using process data to predict machine failure and intervene in advance seems a perfect way to improve machine availability. But does that mean that maintenance will now become the domain of data scientists instead of engineers? I think the answer is no. On the contrary, predictive maintenance could well be a way to bring maintenance back to the shop floor.
Total productive maintenance is a known concept for a long time already. It aims to improve production efficiency, employee morale and job satisfaction by actively involving operators in the maintenance of their equipment. According to different sources, TPM allows manufacturers to reach operational equipment efficiencies up to 85%. TPM gives a number of interesting clues and how predictive maintenance can be applied in a real life environment.
The most important pillar in TPM is autonomous maintenance. When you believe that operators are the people who know their machines best, it is logical to have them involved in the maintenance of their equipment. By giving them access to data and smart analytics on the relationship between process measures and the health of their machine we can make them key participants in the maintenance process. Next to training this requires of course good preparation.
The first step of total predictive maintenance would be 5s. 5s allows to make problems visible. When we use big data and analytics applications, 5s would apply to the physical environment of the machine but also to the data received from the machines sensors. We notice that companies store plenty of process data knowing that the data are of very low quality. In the 5s step a data scientist would work with the operators to identify which signals can be considered reliable. Then the reliable data series can be made visible on the shop floor.
In a continuous way the data scientist can work with the operators and the engineering staff to build alternative maintenance plans based on data and the collective knowledge of the entire team. This approach will not only lead to better maintenance plan but also to a much higher involvement of the operators in the day-to-day management of their equipment. Once the right alerts have been set up the operators can become more autonomous in maintaining the equipment. Through visual alerts created by an analytics tool they can see what is happening with their equipment at any moment and decide about the best timing to perform maintenance activities.
Does this require more highly skilled operators? I do not think so. When the alerts are set up with the team the operators will own the rules that are applied. They will know why certain alerts are created and they will understand what measures underpin each alert. This will increase their ownership and allow them to manage their own environment.
New technologies often seem very complex to use at first. But when they become more mature the complexity of the technology disappears and only the power of its functionalities remain visible. With simple and intuitive user interfaces it must be possible to reach that level of maturity very quickly in predictive maintenance. Once you reach that point, advanced technology and employee involvement can go hand in hand to reach the next level of total productive maintenance. And total predictive maintenance will have become a fact.