Big Data, Machine Learning and the Future of Condition Monitoring

Julkaistu 05.07.2018

Big Data processing and machine learning have lately been popular topics in various publications and events. I see great potential benefits for the industry in these subjects, on various levels. Especially interesting is the use of Big Data processing an machine learning in condition monitoring. There are several different applications, but the goal is to optimize condition monitoring and move toward a more proactive or predictive maintenance, which also improves availability.

There are a often many instruments that measure industrial process data in real time. This measurement data is transferred to the control room, where the operator oversees the process state by interpreting the data. The measurement data is also often stored in order to make it easier to find the root causes for equipment failure, among other things. As a result, there is a large amount of process data which both reflects the process state at a set point in time, but also indicates the operability and condition of equipment that are part of the process.

By analyzing this process Big Data and comparing it to e.g. relevant equipment failure statistics, it is possible with the help of machine learning to produce applications that monitor process data in real time, and use it to identify process equipment ageing during operation. Big Data that is filtered with machine learning can be used to make projections that predict equipment failure probability in the future. These projections can be used to facilitate preventive maintenance and optimize the timing of maintenance procedures.

Preventive condition monitoring and maintenance increase equipment availability, when maintenance procedures are completed before equipment failure. This also improves process safety, when the availability of safety-related equipment increases. By moving from periodic inspection and maintenance to predictive actions, the costs of said procedures is also optimized, as they can be timed more accurately with the help of the information gained from the processed Big Data. When planning process maintenance outages, the data and projections can be consulted to see which devices need to be maintained or inspected during the upcoming outage, and which ones will most likely stay operable until the next outage. This reduces the amount of unnecessary procedures, such as ones where a process device is opened for inspection, with the conclusion that the device is still in fine condition. Each inspection where a device is opened includes a risk that a mistake is made during re-assembly that leads to equipment failure. Therefore, minimizing unnecessary inspections will reduce the overall risk of failure for the device.

Big Data processing with the help of machine learning are tools for future condition monitoring, which will make industrial processes safer, more efficient, and more reliable. I recommend every reader to follow research and development related to these topics, of which there are already many examples out there!


Profile picture of Joel Maunula

Head of Operability

Joel Maunula

Laitosturvallisuus ja -optimointi

+358 50 575 1372