Predictive maintenance and measurement quality control: the hidden information within data collected by monitoring systems

Monitoring systems developed by CAE manage metadata related to measurements, associating each data point with a parameter that defines its estimated quality. This assessment is verified and, if necessary, updated at every stage of the chain: from the sensor to the datalogger and up to the central software systems.

It is precisely at the central level, with the aim of offering high-value services to customers, that dedicated software solutions operate, such as Detective. This software is capable of analyzing in just a few seconds the vast amounts of data coming from sensors across all field-installed stations. Thanks to its ability to quickly process incoming data, Detective can report potential anomalies in real-time and assigns a specific data quality marker to each one. This evaluation is then made available to the operator for further assessment or, if the data is used as input for any forecasting model, to determine whether it should be included based on its reliability.

With the goal of assessing data reliability, among the numerous checks implemented by Detective are those defined by ISPRA in the “Guidelines for the validation of hydro-meteorological data.” As a result, dozens of algorithms perform checks on rain gauges, air thermometers, hydrometers, flow rates, and other variables.

Following the philosophy of these checks, similar verifications have been implemented in Detective for pressure, radiation, relative humidity, wind direction, and wind speed. In its current version, Detective includes 110 different checks on meteorological parameters, diagnostic data, and the operation of both the central system and the field network. The program’s structure allows for the easy addition of new checks or the extension of existing ones to other types of parameters in the future.

Here, thanks to Machine Learning techniques, the identification of potential anomalous data is both fast and effective, enabling operators to detect drift phenomena and other malfunctions that would otherwise be difficult to identify.

In addition to measurement validation, CAE has developed specific expertise in analyzing diagnostic data from maintained systems. Based on its experience, CAE has introduced control over numerous additional parameters to obtain a comprehensive view of field equipment performance. Issues such as energy imbalance, component wear, vandalism, and many others can be identified.

This is where predictive maintenance takes shape. Identifying precursor behaviors of a malfunction before it turns into a failure, with the consequent loss of measurements, is an advantage made possible by modern techniques, enabling improved efficiency, typically associated with well-planned maintenance activities.