Creating insights in machinery performance & enabling failure analysis

A client of Settels Savenije foresaw issues with a module of its machinery, risking an unpredictable and shorter lifespan than expected. Experience learns that the modules often fail within a time well below the expected lifetimes.

This unpredictable lifespan leads to frequent, high-cost (un)planned downtime and expensive replacement of components. Moreover, the complexity level of future generations of these machines will keep increasing, causing even faster degradation under the current circumstances.

To increase the lifespan of these modules, Settels Savenije aims to learn more about the operating behavior of these machines, their failure modes, and the causes of failure. This information can then be used to make a comparison between different suppliers and to improve the equipment.

At the start, Settels Savenije had built multiple test setups to simulate operational field situations and accelerate occurrences of failure modes. The behavior of the equipment is continuously measured by sensors and this data is stored in InfluxDB, a specific database for large volumes of time series data.

In turn, Bright Cape’s responsibility was to process the high volumes of raw sensor data (approx. 216 million datapoints, 7-8 GB, per day) and convert these into useful insights. An understanding of the performance and standard behavior of the equipment could be achieved via an extensive set of performance metrics, which are properties or characteristics of the observed measurements.

These metrics are calculated using advanced analytics and displayed in several dashboards to enable daily monitoring and failure analysis. A so-called task analysis was used to align the user interface of these dashboards with the tasks performed by the end-users. Here, the goal, scope, (sub)tasks, and type of visualizations were defined in close collaboration with Settels Savenije’s engineers.

The existing hardware and data infrastructure of the setups were extended by Bright Cape with an end-to-end data analytics solution, including an additional data storage layer, ETL pipeline, and a fully automated data synchronization and processing solution. On top of this, a total of four interconnected dashboards were built in Grafana, an interactive time-series visualization platform.

These dashboards enable engineers to monitor machinery performance on a daily basis, perform a comparison inter- & intra-configurations, and investigate deviations from standard behavior using a root cause analysis. Once these simulated field situations are put into practice, this could lead to potential savings of tens of thousands of dollars per reduced hour of maintenance downtime. In addition, the time it takes to perform an inspection and create an analytical report has been significantly reduced from 2-3 weeks to 1.5 hours.