The invention relates to a system and to a method for determining a state of a device by means of a trained support-vector machine. According to the invention, an operating parameter space is divided into classification volumes, at least one of which indicates a normal state and at least one other of which indicates a fault state of the device. A current state of the device can therefore be determined by determining where a current operating parameter point is to be arranged in the operating parameter space. The invention further relates to methods and to variants of the system in order to facilitate a cause evaluation and to determine particularly relevant operating parameters for the fault determination.

A learning diagnosis apparatus performs learning from failure data to create a diagnostic model, and stores a model, a failure cause part, and sensor data of the equipment in a rare case data table when the number of cases of the failure cause part of the equipment is less than a predetermined number. Then, based on the diagnostic model created by a learning unit, an estimated probability of causing a failure is calculated for each part of the equipment in which a failure has occurred. Based on the rare case data table, a sensor data match rate between sensor data of the equipment in which the failure has occurred and past sensor data of the model of the equipment is calculated. Then, the calculated sensor data match rate for each part of the equipment in which the failure has occurred is displayed.

A method is provided for repair of a mechanical or electromechanical system of a machine. An inference engine receives indication of a failure mode of the system, and measurements of operating conditions of the machine, and the inference engine defines a current problem including (a) the failure mode of the system, and (b) a pattern in the measurements. The inference engine searches a knowledge base with historical problems including (a) failure modes of systems of the machine, and (b) patterns in measurements of the operating conditions, and respective solutions with (c) repair actions performed to address the respective ones of the failure modes, for a respective solution to a historical problem most similar to the current problem. This is inferred as a solution to the current problem, and the inference engine generates an output display indicating the repair action of the solution to address the failure mode of the system.

The present disclosure provides a method for anomaly classification for an industrial control system (ICS) communication network based on statistical learning and deep learning. This method designs LSTM deep learning structure parameters and performs modeling analysis based on a large amount of traffic data during normal operation of the ICS communication network; based on real-time communication traffic data thresholds generated by a SARIMA online statistical learning model, designs correlated algorithms to analyze a numerical relationship between background traffic and real-time traffic; and classifies ICS communication network anomalies according to an ICS network anomaly classification algorithm. In the present disclosure, an ICS test network range involving virtual and physical devices and a test platform in Zhejiang Province are used for experimental analysis, a physical simulation platform is built in a laboratory environment for validation, and detailed examples are provided to verify the reliability and accuracy of the algorithm.

System and methods for detecting anomalies and identifying faults of a gas turbine engine may include a recorder in communication with a processor. The recorder may be configured to capture archival data of the gas turbine engine. A flight normalizer module may be configured to produce normalized results based on the archival data. A flight parameter features module may be configured to generate flight parameter features based on the normalized results. A data warehouse module may be configured to determine suspected fault classes by comparing the flight parameter features against training parameter features stored in the data warehouse module based on queries from the flight parameter features module. A majority vote module may be configured to determine a diagnosed fault class based on the suspected fault classes.

A system for providing real time aircraft engine sensor analysis includes a computer system configured to receive an engine operation data set in real time. The computer system includes a machine learning based analysis tool and a user interface configured to display a real time analysis of the engine operation data set. The user interface includes at least one portion configured to identify a plurality of anomalies in the engine operation data set.

A system for determining the cause of an abnormality in a semiconductor manufacturing process includes an abnormality mode determination module, a selection module, and a root cause analysis module. The abnormality mode determination module is used to determine the similarity between wafer bin maps containing the abnormal data. When the similarity among the wafer maps is higher than a reference value, the selection module executes the steps of: determining a bad lot based on the wafer maps where the similarity is higher than the reference value; determining a time span within which the bad lot is generated; selecting other bad lots occurring in the time span and satisfying a failure model; selecting a good lot based on a fixed lot interval. The root cause analysis module is used to execute the steps of calculating the correlation among data to obtain confidence indexes.

To perform a prognostic health analysis for an asset, a plurality of independent stochastic simulations are performed using transition probabilities of a discrete Markov Chain model. A prognostic asset health state evolution is computed over a time horizon from the plurality of independent stochastic simulations. An output is generated based on the computed prognostic asset health state evolution.

Aspects of the present disclosure provide a method and apparatus for maintenance recommendations. Embodiments include receiving data related to a plurality of fault codes. Embodiments include determining, for each respective fault code of the plurality of fault codes, a unique subset of parameters for the respective fault code based on frequencies of the parameters within the plurality of fault codes in the data. Embodiments include receiving a plurality of fault records and a plurality of maintenance records. Embodiments include, for each respective maintenance record of the plurality of maintenance record, identifying one or more candidate fault records of the plurality of fault records and correlating a subset of the one or more candidate fault records with the respective maintenance record based on the unique subset of the parameters for the respective fault code corresponding to each respective candidate fault record of the one or more candidate fault records.

Systems, methods, and other embodiments associated with providing long-term predictions for specific maintenance are described. In general, the system or method will: Retrieve a current amount of usage for a component part of a device, a failure probability curve for the component part, and a planned maintenance schedule for the device. Update the failure probability curve based on trends of data obtained from sensors associated with the component part, and causal factors associated with the device. Determine that the likelihood of failure for the component part after a first upcoming planned maintenance and before a second upcoming planned maintenance exceeds a threshold. Generate a work order for specific maintenance on the component part to be performed during the first upcoming planned maintenance. Transmit the work order to cause resources to be timely obtained and labor allocated to perform the specific maintenance during the first upcoming planned maintenance.