The advent of Industry 4.0 is bringing more innovation and progress but also puts more pressure to the production process. The cost of unscheduled downtime in the manufacturing industry can be high. Ensuring that machines and components remain operational requires the right maintenance strategy. We ensure a shift at clients from corrective maintenance to preventive and predictive maintenance.
The corrective maintenance strategy still used by most clients ensures that assets are used to their maximum capacity and preventive maintenance costs are avoided. If the machine suddenly starts smoking, it is clear that the maximum service life has been reached. However, the total cost of repair often turns out to be higher than that of a scheduled maintenance service. Unscheduled downtime results in production staff being sent home, installation engineers having to work overtime and clients being disappointed.
Predictive maintenance offers a solution
Predictive maintenance is usually based on the actual condition of machines, also called Condition-Based Maintenance (CBM). The advantage over a Time-Based Maintenance (TBM) strategy is that no over-maintaining is necessary. If the lifespan of machines varies greatly, many machines will be serviced too early. A client whose company has over 250 robotic arms in use provides an example. ‘Each robotic arm makes a unique movement with a unique gripper, which results in a large variance in lifespan’ says Wiljan Vos, Data Consultant for Beenen (part of Batenburg Industrial Automation). ‘Replacing robots after 5 years when they can still produce another 5 years is obviously a waste of money, which is why we switched to the CBM strategy.’
As simple as predictive maintenance may sound, and however tempting the benefits may be, properly implementing the various tools and methodologies is far from an easy task. For predictive maintenance to provide a solution, you need to have suitable data. Sensor data from machines is combined with Event Logs, SCADA data and maintenance history to determine the condition of a machine. By working closely with software and maintenance engineers, analysts know how to interpret the data. Vos: ‘Data scientists or maintenance engineers alone won’t get you there; cooperation is necessary.’
Intensive cooperation with clients and extensive analyses of machines and data result in the acquisition of a great deal of new knowledge. Knowledge that is then used to improve the current production process. This ultimately proves to be a very useful by-product of a CBM project. Vos: ‘We have helped to improve the repair process and have also gained a better understanding of how robotic arms work and are now able to implement optimisations.’
Machine Learning
Detecting and identifying errors in robotic arms is primarily done by analysts. Machine Learning then ensures that deviations can be automatically detected. This may involve the exceeding of a simple threshold but also the results of complex or statistical calculations. In time, the system will be able to detect more deviations and be even more valuable. ‘Don’t think that every error can be identified well in advance after implementation. Improving the process of predictive maintenance requires a lot of data and time,’ says Vos. Installation engineers are gradually becoming familiar with the new system, which is causing a cultural shift. The transition from corrective to preventive work requires a different mindset from installation engineers, maintenance engineers and production staff. On the other hand, machine availability will increase at lower maintenance costs.
[This article was also published in Link Magazine number 2 April 2020]