Hybrid risk model for maintenance optimization and system for executing such method

A computer-implemented method for maintenance optimization of a fleet or group of plants, wherein the plants comprise one or more turbomachinery assets, wherein each turbomachinery asset comprises a turbomachine and related auxiliary systems, and wherein each of the turbomachinery assets comprises one or more installed sensors, each installed sensor capable of generating operating signals of the turbomachinery assets,

• the method comprising the steps of:
• calculating model configuration parameters by comparing healthy conditions and unhealthy conditions of the turbomachinery assets found in historical fleet data;
• acquiring data constituted by operating signals from the sensors of the turbomachinery assets within a defined timeframe;
• filtering and decorrelating the operating signals to remove correlation between operating signals from different systems, resulting in a decorrelated signals;
• extracting features from the decorrelated signals;
• using a multivariate problem approach, comparing the features to the model configuration parameters to identify one or more anomalies in the features;
• using a classifier,
• establishing anomaly classes for the anomalies and associating the anomaly classes to the turbomachinery assets of the fleet they refer to;
• and
• classifying the anomalies according to the class as system anomalies, which are system malfunctions or degradations related to the turbomachinery assets of the fleet, or into sensor anomalies, which are sensor malfunctions,
• wherein,
• if system anomalies, then executing a risk assessment for estimating the risk of any event that requires a maintenance task and a proper time for the maintenance task to occur,
• else, if sensor anomalies, then executing a severity assessment for assigning a severity to the anomalies identified as sensor malfunctions;
• using the severity, assigning maintenance dispositions for the maintenance tasks;
• generating a file that includes the maintenance dispositions;
• and
• sending the file to an end user,
• wherein the anomaly classes are selected from historical data or by expert knowledge, and
• wherein the anomaly classes are selected from among the following:
• signal freezing, step-change, symmetric noise, asymmetric noise, spikes, abnormal or anomalous range, or drift.