# US 11,113,905 B2 · Fault detection system and method for vehicle system prognosis > Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt - Number: US 11,113,905 B2 - Document ID: US11113905B2 - Status: Granted - Kind: B2 - Application number: 16/776,867 - Filed: 2020-01-30 - Published: 2020-05-28 - Granted: 2021-09-07 - Earliest claimed: 2017-07-07 - Assignees: The Boeing Company - Applicants: The Boeing Company - Inventors: Dmitriy Korchev; Charles E. Martin; Tsai-Ching Lu; Steve Slaughter; Alice A. Murphy; Christopher R. Wezdenko - Web page: https://patentgrep.com/patent/US11113905B2 ## CPC - G07C 5/0808: {Diagnosing performance data (testing of vehicles G01M17/00; testing of electrical installation on vehicles G01R31/005)} - G07C 5/0816: {Indicating performance data, e.g. occurrence of a malfunction} - G07C 5/085: {using electronic data carriers} - H04L 67/12: specially adapted for proprietary or special-purpose networking environments, e.g. medical networks, sensor networks, networks in vehicles or remote metering networks - H04L 69/04: Protocols for data compression, e.g. ROHC - H04L 69/40: for recovering from a failure of a protocol instance or entity, e.g. service redundancy protocols, protocol state redundancy or protocol service redirection (management of faults, events, alarms or notifications in data switching networks H04L41/06) - G07C: TIME OR ATTENDANCE REGISTERS; REGISTERING OR INDICATING THE WORKING OF MACHINES; GENERATING RANDOM NUMBERS; VOTING OR LOTTERY APPARATUS; ARRANGEMENTS, SYSTEMS OR APPARATUS FOR CHECKING NOT PROVIDED FOR ELSEWHERE - H04L: TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION (arrangements common to telegraphic and telephonic communication H04M) ## Abstract A fault detection system including one or more sensors onboard a vehicle to detect a characteristic of the vehicle and generate sensor signals corresponding to the characteristic, a processor onboard the vehicle to receive the sensor signals, generate one or more fast Fourier transform vectors based on the sensor signals so that the one or more fast Fourier transform vectors are representative of the characteristic, generate an analysis model from a time history of the fast Fourier transform vectors, and determine, using the analysis model, a degree to which the one or more fast Fourier transform vectors could have been generated by the analysis model, and an indicator to communicate an operational status of the vehicle to an operator or crew member of the vehicle based on the degree to which the one or more fast Fourier transform vectors could have been generated by the analysis model. ## Claims ### Claim 1 A fault detection system comprising: one or more sensors onboard a vehicle, the one or more sensors being configured to detect a predetermined characteristic of the vehicle and generate a plurality of raw unlabeled sensor signals corresponding to the predetermined characteristic; a processor onboard the vehicle, the processor being configured to receive the plurality of sensor signals from the one or more sensors, generate one or more fast Fourier transform vectors based on the plurality of sensor signals so that the one or more fast Fourier transform vectors are representative of the predetermined characteristic, generate an analysis model for the predetermined characteristic from a time history of the fast Fourier transform vectors, where the analysis model is trained by the processor with a training data set of fast Fourier transform vectors that are generated from the plurality of raw unlabeled sensor signals obtained only under normal health data operating conditions of the predetermined characteristic, and determine, using the analysis model, a degree to which the one or more fast Fourier transform vectors could have been generated by the analysis model; and an indicator in communication with the processor, the indicator being configured to communicate an operational status of the vehicle to an operator or crew member of the vehicle based on the degree to which the one or more fast Fourier transform vectors could have been generated by the analysis model. ### Claim 2 (depends on 1) The fault detection system of claim 1, wherein the plurality of sensor signals obtained only under the normal health data operating conditions of the predetermined characteristic are obtained over one or more predetermined time periods. ### Claim 3 (depends on 2) The fault detection system of claim 2, wherein the processor is configured to generate a time series of fast Fourier transform vectors corresponding to the one or more predetermined time periods, the time series of fast Fourier transform vectors forming the training data set of fast Fourier transform vectors. ### Claim 4 (depends on 1) The fault detection system of claim 1, wherein the training data set of fast Fourier transform vectors is generated from the plurality of sensor signals, the plurality of sensor signals being obtained from the vehicle and at least another vehicle. ### Claim 5 (depends on 1) The fault detection system of claim 1, wherein the processor is configured to filter fast Fourier transform vectors from the training data set of fast Fourier transform vectors based on the mean and standard deviation of at least one spectral bin of the training data set of fast Fourier transform vectors. ### Claim 6 (depends on 1) The fault detection system of claim 1, wherein the processor is configured to compress an original number of spectral bins of the training data set of fast Fourier transform vectors to a compressed number of spectral bins so that at least one spectral bin in the compressed number of spectral bins is an average amplitude of a corresponding number of spectral bins in the original number of spectral bins. ### Claim 7 (depends on 1) The fault detection system of claim 1, wherein the processor is further configured to filter spurious fast Fourier transform vectors from the one or more fast Fourier transform vectors. ### Claim 8 A method comprising: detecting, with one or more sensors onboard a vehicle, a predetermined characteristic of the vehicle and generating, with the one or more sensors, a plurality of sensor signals corresponding to the predetermined characteristic; receiving, with a processor onboard the vehicle, the plurality of sensor signals from the one or more sensors, generating, with the processor onboard the vehicle, one or more fast Fourier transform vectors based on the plurality of sensor signals so that the one or more fast Fourier transform vectors are representative of the predetermined characteristic, generating, with the processor onboard the vehicle, an analysis model for the predetermined characteristic from a time history of the fast Fourier transform vectors, where the analysis model is trained by the processor with a training data set of fast Fourier transform vectors that are generated from the plurality of raw unlabeled sensor signals obtained only under normal health data operating conditions of the predetermined characteristic, and determining, with the processor onboard the vehicle using the analysis model, a degree to which the one or more fast Fourier transform vectors could have been generated by the analysis model; and communicating an operational status of the vehicle to an operator or crew member of the vehicle, with an indictor in communication with the processor, based on the degree to which the one or more fast Fourier transform vectors could have been generated by the analysis model. ### Claim 9 (depends on 1) The fault detection system of claim 1, wherein the processor is further configured to generate the analysis model by: applying dictionary learning to the training data set of fast Fourier transform vectors to create a learned dictionary, decomposing the one or more fast Fourier transform vectors in the training data set of fast Fourier transform vectors where decomposition of the one or more fast Fourier transform vectors in the training data set of fast Fourier transform vectors represents a low rank portion of the analysis model and a residual of the decomposition represents a sparse portion of the analysis model, reconstructing the one or more fast Fourier transform vectors in the training data set of fast Fourier transform vectors to generate a set of weights for each or at least one of the one or more fast Fourier transform vectors in the training data set of fast Fourier transform vectors, calculating a residual vector for each or at least one fast Fourier transform vector in the training data set of fast Fourier transform vectors, the residual vector representing a difference between the respective fast Fourier transform vector in the training data set of fast Fourier transform vectors and the reconstructed fast Fourier transform vector corresponding to the respective fast Fourier transform vector, and generating a low rank Gaussian model for the low rank portion and a sparse Gaussian model for the sparse portion based on the residual vector for each or at least one fast Fourier transform vector in the training data set of fast Fourier transform vectors. ### Claim 10 (depends on 1) The fault detection system of claim 1, wherein the processor is further configured to determine, using the analysis model, the degree to which the one or more fast Fourier transform vectors could have been generated by the analysis model by one or more of: determining a low rank portion of the one or more fast Fourier transform vectors, by decomposing the one or more fast Fourier transform vectors with a learned dictionary of the analysis model, and employing a low rank Gaussian model to determine the degree to which the one or more fast Fourier transform vectors could have been generated by the analysis model; and determining a sparse portion of the one or more fast Fourier transform vectors, by calculating a residual vector that represents a difference between a respective one of the one or more fast Fourier transform vectors and a reconstructed fast Fourier transform corresponding to respective one of the one or more fast Fourier transform vectors, and employing a sparse Gaussian model to determine the degree to which the one or more fast Fourier transform vectors could have been generated by the analysis model. ### Claim 11 (depends on 1) The fault detection system of claim 1, wherein the vehicle is an aerospace vehicle. ### Claim 12 (depends on 1) The fault detection system of claim 1, wherein the vehicle is an airplane. ### Claim 13 (depends on 8) The fault detection system of claim 8, wherein the processor is further configured to compress an original number of spectral bins of the one or more fast Fourier transform vectors to a compressed number of spectral bins so that at least one spectral bin in the compressed number of spectral bins is an average amplitude of a corresponding number of spectral bins in the original number of spectral bins. ### Claim 14 (depends on 13) The method of claim 13, wherein the plurality of sensor signals obtained only under the normal health data operating conditions of the predetermined characteristic are obtained over one or more predetermined time periods. ### Claim 15 (depends on 14) The method of claim 14, wherein the processor is configured to generate a time series of fast Fourier transform vectors corresponding to the one or more predetermined time periods, the time series of fast Fourier transform vectors forming the training data set of fast Fourier transform vectors. ### Claim 16 (depends on 13) The method of claim 13, wherein the training data set of fast Fourier transform vectors is generated from the plurality of sensor signals, the plurality of sensor signals being obtained from the vehicle and at least another vehicle. ### Claim 17 (depends on 13) The method of claim 13, wherein the processor is configured to filter fast Fourier transform vectors from the training data set of fast Fourier transform vectors based on the mean and standard deviation of at least one spectral bin of the training data set of fast Fourier transform vectors. ### Claim 18 (depends on 13) The method of claim 13, wherein the processor is configured to compress an original number of spectral bins of the training data set of fast Fourier transform vectors to a compressed number of spectral bins so that at least one spectral bin in the compressed number of spectral bins is an average amplitude of a corresponding number of spectral bins in the original number of spectral bins. ### Claim 19 (depends on 1) The fault detection system of claim 1, wherein the processor is further configured to filter spurious fast Fourier transform vectors from the training data set of fast Fourier transform vectors. ### Claim 20 (depends on 19) The fault detection system of claim 19, wherein the processor is configured to filter the spurious fast Fourier transform vectors from the training data set of Fast Fourier transform vectors based on a pre-defined statistical norm threshold.