Systems and methods for monitoring equipment
Systems and methods for monitoring apparatus(es) and equipment using a sensor cluster are described. The systems and methods can be used to automatically repair or otherwise address actual and predicted failure modes of the apparatus(es).
1 . A system for monitoring an apparatus comprising a motor, the system comprising:
a vibration sensor comprising a multi-axis accelerometer;
a housing surrounding the vibration sensor;
a mounting interface coupled to the housing and configured to couple the system to the apparatus away from the motor;
a signal conditioning and communications subsystem coupled to the vibration sensor within the housing and configured to receive a vibration signal stream from the vibration sensor; and
a processing subsystem coupled to the signal conditioning and communications subsystem and comprising a neural processing unit (NPU), the processing subsystem contained within the housing and comprising non-transitory media storing instructions that, when executed, perform operations for:
receiving vibration data derived from the vibration signal stream;
performing a set of transformation operations upon said vibration data;
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 comprises signatures associated with a set of harmonic faults of the apparatus, a set of synchronous faults of the apparatus, a set of sub-harmonic and sub-synchronous faults of the apparatus, and a set of non-synchronous faults of the apparatus; and
returning an analysis comprising a recommended action for improving or maintaining proper performance of the apparatus, based upon the set of unique signatures.
2 . The system of claim 1 , wherein the NPU is an NPU with 1 trillions of operations per second (TOPS) capability with energy use performance of less than 1 picojoule per operation.
3 . The system of claim 1 , wherein NPU comprises self-attention time-series transformer architecture comprising an encoder block comprising multi-head attention subarchitecture.
4 . The system of claim 3 , wherein the self-attention time-series transformer architecture of the NPU omits a decoder block.
5 . The system of claim 1 , wherein the apparatus comprises an unmanned aerial vehicle.
6 . The system of claim 5 , wherein the set of subcomponents comprises subcomponents of an engine of the unmanned aerial vehicle and flight control surfaces of the unmanned aerial vehicle.
7 . The system of claim 1 , wherein the apparatus comprises an electric utility apparatus.
8 . The system of claim 1 , wherein the set of subcomponents comprises a bearing, a shaft, a belt, and a gear of the apparatus.
9 . A system for monitoring an apparatus comprising a motor, the system comprising:
a vibration sensor comprising a multi-axis accelerometer;
a housing surrounding the vibration sensor;
a mounting interface coupled to the housing and configured to couple the system to the apparatus away from the motor;
a signal conditioning and communications subsystem coupled to the vibration sensor within the housing and configured to receive a vibration signal stream from the vibration sensor; and
a processing subsystem coupled to the signal conditioning and communications subsystem and comprising a neural processing unit (NPU), the processing subsystem contained within the housing and comprising on-chip self-attention time-series transformer architecture for processing a vibration signal stream, wherein the processing subsystem comprises non-transitory media storing instructions that, when executed, perform operations for:
receiving vibration data derived from the vibration signal stream;
performing a set of transformation operations upon said vibration data;
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 comprises signatures associated with a set of overhang faults of the apparatus, underhang faults of the apparatus, misalignment faults of the apparatus, imbalance faults of the apparatus, and multi-modal faults of the apparatus; and
returning an analysis comprising a recommended action for improving or maintaining proper performance of the apparatus, based upon the set of unique signatures.
10 . A method for monitoring an apparatus, the method comprising:
providing a mounting interface between a vibration sensor coupled to a signal processing subsystem, and the apparatus, without direct contact between the vibration sensor and a motor of the apparatus, wherein the vibration sensor comprises a multi-axis accelerometer;
sampling a vibration signal stream generated from the vibration sensor during operation of the apparatus;
performing a set of transformation operations upon the vibration signal stream, wherein the set of transformation operations comprises operations applied by self-attention time-series transformer architecture;
identifying a set of unique signatures corresponding to faults of a set of subcomponents of the apparatus from the set of transformation operations, wherein the set of unique signatures comprises signatures associated with a set of harmonic faults of the apparatus, a set of synchronous faults of the apparatus, a set of sub-harmonic and sub-synchronous faults of the apparatus, and a set of non-synchronous faults of the apparatus; and
returning an analysis comprising a recommended action for improving or maintaining proper performance of the apparatus, based upon the set of unique signatures.
11 . The method of claim 10 , wherein said self-attention time-series transformer architecture comprises an encoder block comprising multi-head attention subarchitecture.
12 . The method of claim 11 , wherein said self-attention time-series transformer architecture omits a decoder block.
13 . The method of claim 10 , wherein the apparatus comprises an unmanned aerial vehicle and wherein the set of subcomponents comprises subcomponents of an engine and flight control surfaces of the unmanned aerial vehicle.
14 . The method of claim 13 , further comprising executing the recommended action, wherein the recommended action comprises controlling flight operation of the unmanned aerial vehicle in response to a fault of at least one of the set of subcomponents.
15 . The method of claim 10 , wherein the apparatus comprises an electric utility apparatus.
16 . The method of claim 15 , further comprising executing the recommended action, wherein the recommended action comprises shutting down the electric utility apparatus.