Predictive maintenance helps to predict machine failures and malfunctions based on:

Data

Today, sensor data such as measurements of vibrations, pressure, flow, temperature, etc. can be transferred to an internet platform for analysis by means of the Internet Of Things (IoT). It is also possible to bring together different data sources, such as the  history of the machine's operations, maintenance history, production data (production quality, product failures, production downtime, etc.) and environmental data such as temperature, humidity, air quality, etc. Suddenly, many more parameters can be involved in the comparison with the failure data, revealing new insights and connections.

Predictive Maintenance Model

Based on machine learning algorithms, data scientists arrive at a predictive maintenance model that actually predicts when asset failures occur, based on trends of one or more parameters. As more data is kept, the model can also be further adjusted.

Collaboration with Maintenance Engineers

An important role is reserved for maintenance engineers and reliability engineers who, based on their many years of experience, help determine the asset selection, provide information about available data, and are involved in the creation of the predictive maintenance model from the start of a predictive maintenance project.

Our approach for a pilot project, where Realdolmen possesses the necessary expertise in each of these areas:

• Asset selection: Which assets are critical to your production process or guarantee the proper functioning of your installation or machine? Improvements in the maintenance of these assets will provide you with the most value.
   
• Data identification: Qualitative data is very important. We investigate the availability (historical and current), relevance, reliability and quality of your existing data.
   
• Data exploration: Our data scientists use the data to work with various analysis tools and machine learning to convert the data into a Predictive Maintenance Model. The model uses algorithms based on the trends and history of one or more parameters to indicate when asset failures are imminent.

• Data monitoring: The interaction with the maintenance staff is crucial in testing the model against reality. Dashboards and reporting are important in communication with engineers. As more data is kept, the model can also be further adjusted.
   
• Predict & prescribe: The model helps to improve maintenance. The output of the model can be integrated in the CMMS to control predictive actions. In the final phase, additional artificial intelligence can be employed to take the step to prescriptive maintenance. This involves the system using the CMMS to make proposals with regard to desired actions.