AI (artificial intelligence) gives rise to new possibilities in solving problems in enterprise asset management. It is one of the new technologies for Industry 4.0. Preventing unexpected breakdown is one of the key objectives of maintenance. AI may shed light on new approaches to it.
Asset monitoring is done through meter reading analysis. For instance, meter triggers can be defined in Calem to watch for reading spikes that are out of bounds (see this blog). This blog presents an AI approach to meter reading analysis. For instance, readings of an asset are used to train a model to recognize the reading patterns (the first chart below). When readings are not following patterns (highlighted in red in the second chart), a work order is raised. The approach makes it possible to catch potential issues before asset breakdowns.
Asset readings are timeseries with datetimes and readings. A data set of readings are prepared for the past three years as shown in the chart below. The data set is used to train a model to discover reading patterns of a meter.
A validation set of readings are provided to check for anomaly (chart below). The validation set is about a third of the time span of the training set.
A model is trained to discover the reading patterns based on the training set above. Early stopping is used to stop the model training to avoid overfitting.
The chart below shows a segment of readings of training and prediction. The model looks good.
The model is applied to the validation data set. It discovered an anomaly highlighted in red. The anomaly will raise a work order in Calem so the anomaly can be investigated. Preventive or corrective actions might be necessary.
Additional Resources
By accepting you will be accessing a service provided by a third-party external to https://eam.calemeam.com/