System and method for evaluating system events and executing responses
AssigneeFluid Power AI, LLC
InventorsJose Maria Guadalupe Gomez, Francis Mitchell Toglia, Mario Gomez Narvaez
Life of this patent
Estimating the term…
Abstract
A system includes sensors for monitoring signals, and a processing system executes one or more methods for identification of system events, from the signals, corresponding to state changes and performance of the system and/or its subcomponents. Event identification is performed with classification and/or other machine learning algorithms, with generation of novel training data sets. The sensor(s) can also be used to determine power consumption information about the system and/or its subcomponents. The system processes event-associated outputs for execution of actions for improving system performance, along with other downstream applications.
Claims20
A system for monitoring an apparatus, the system comprising:
- a sensor cluster comprising a vibration sensor;
- an interface comprising a housing coupling the sensor cluster to the apparatus, wherein the vibration sensor is positioned to monitor vibration induced by a motor of the apparatus;
- a monitor coupled to the sensor cluster and comprising a controller for sampling data derived from the sensor cluster and transmitting data from the monitor;
- and
- a processing subsystem coupled to the interface and comprising a non-transitory computer-readable medium comprising instructions stored thereon, that when executed by the processing subsystem perform one or more steps of:
- receiving a dataset derived from outputs of the sensor cluster;
- performing a set of transformation operations upon the dataset, using a neural network model comprising an encoder;
- identifying a set of unique signatures corresponding to states of a set of subcomponents of the apparatus, from the set of transformation operations, wherein the set of unique signatures are extracted from cyclic oscillation features, changes in multiple harmonics, and magnitudes of harmonics of signals from the sensor cluster, and wherein states of the set of subcomponents of the apparatus comprise failure modes of the set of subcomponents attributed to the set of unique signatures;
- and
- returning an analysis comprising a recommended action for improving or maintaining proper performance of the apparatus, based upon the set of unique signatures.
A method for monitoring an apparatus, the method comprising:
- establishing an interface between a sensor cluster and the apparatus, the interface comprising a housing coupling the sensor cluster to the apparatus, wherein the sensor cluster comprises at least one of:
- a vibration sensor positioned to monitor vibration induced by the apparatus, a temperature sensor positioned to monitor temperature, a current sensor positioned to monitor current, and a voltage sensor positioned to monitor voltage, through the sensor interface;
- sampling a set of data streams, derived from outputs of the sensor cluster;
- performing a set of transformation operations upon the set of data streams, wherein the set of operations comprises processing the set of data streams with a neural network model comprising one direction and an encoder configured to process time series data of the dataset;
- identifying a set of unique signatures corresponding to states and events of a set of subcomponents of the apparatus, from the set of transformation operations, wherein the set of unique signatures are extracted from harmonic values of signals from the sensor cluster and states of the set of subcomponents of the apparatus comprise failure modes of the set of subcomponents attributed to the set of unique signatures, wherein at least one of the set of unique signatures is associated with a cavitation event involving fluid of the apparatus;
- and
- returning an analysis of the set of subcomponents of the apparatus, based upon the set of unique signatures.
Dependencies are inferred from the claim text.