July/August 2020

16 | Plant & Works Engineering www.pwemag.co.uk July/August 2020 Maintenance Matters Focus on: Asset Management/Maintenance Software I t may not make financial sense to install sensors everywhere and suitable technology may not always be available. In these scenarios, performing preventive maintenance on some assets at regular intervals (for example with a thermal imaging camera) may be the best approach. But predictive maintenance is undoubtedly the future. Alerts are configured into analytical programs to notify users when assets are out of tolerance. When this happens, maintenance staff will utilise all the data to best implement plans to remove assets or schedule downtime. Facilities can be shut down assets at a planned time. Predictive maintenance techniques powered by artificial intelligence (AI) are helping enterprises across industries to find patterns that can avoid machine failures. To reliably detect potential faults, a predictive maintenance system needs to have prior knowledge of all possible fault situations. By “teaching” the maintenance system about all the various possible fault situations, AI-based machine learning software enables it to detect specific faults in real-time and initiate remedial action. “Unlike AI, traditional business intelligence systems are not designed to handle huge volumes of industrial Internet of things (IIoT) data,” says Venkata Naveen, Disruptive Tech Analyst at data analytics company GlobalData. “Predictive maintenance is a key cost-saving digital strategy for any enterprise. AI-powered predictive maintenance can help enterprises to save money and time on maintenance, machine downtime while extending the life of their heavy equipment.” If machines can talk and exchange data, then potential problems can be detected by running background AI-based software that can identify anomalies and other normally unforeseeable faults. Once the predictive maintenance system knows the exact point in the future at which a machine will be shut down for maintenance, it can automatically initiate the associated logistics processes. This ensures that all the relevant work and parts ordering processes are properly coordinated. Mitsubishi Electric has developed AI-based diagnostic technology that harnesses machine learning algorithms to analyse sensor data from a machine and generate a model of its transition between different operational states. The model is then used to set optimal conditions for detecting abnormalities during each operational state, enabling operators to gauge signs of machinery failure before actual breakdowns. Also, by harnessing data from a sensor and using the drive to process the information into actionable insight, a Mitsubishi Electric 800 Series VSD will look after itself as well as the general health of a complete drive train. “One of the critical challenges for predictive maintenance is streamlining the flow of data from machines to a central system with a low level of latency and high security, which, given the advancements in 5G connectivity and cybersecurity, can be overcome,” Naveen concludes. “Despite the stumbling blocks, Every moment that a production line or facility is shut down owing to technical problems is a major cost burden. Happily, most facilities today do have a coherent maintenance programme. However, many facility managers question whether preventive maintenance or predictive maintenance is best. In fact, best-in-class maintenance programmes incorporate both. Andy Pye reports. Exposing the software underbelly There’s a clear pathway to improving the reliability of your injection moulding fleet, electric machines have the edge in that they have fewer moving parts and fewer wear parts.