Systems and methods to diagnose vehicles based on the voltage of automotive batteries
View Patent ↗Systems and methods for battery-based vehicle diagnostics are provided. Various embodiments include a battery-based diagnostics system that guards vehicles against anomalies with a cyber-physical approach. The diagnostics system can be implemented as an add-on module of commodity vehicles attached to automotive batteries, thus providing vehicles an additional layer of protection. The automotive battery can operate in strong dependency with many physical components of the vehicle, which is observable as correlations between battery voltage and the vehicle's corresponding operational parameters, e.g., a faster revolutions-per-minute (RPM) of the engine, in general, leads to a higher battery voltage. These embodiments exploit such physically induced correlations to diagnose vehicles by cross-validating the vehicle information with battery voltage and may be based on a set of data-driven norm models constructed online.
1. A method to facilitate vehicle diagnostics, the method comprising:
constructing a predictive norm model configured to determine correlations between a battery of a vehicle and one or more components of the vehicle, wherein the correlations comprise a first correlation between voltages of the battery and output values of the one or more other components over a time period and a second correlation between maximum values of the voltages and maximum values of the output values over the time period;
measuring real-time voltages of the battery and real-time output values of the one or more components;
supplying the real-time voltages to the predictive norm models;
determining, via the predictive norm model, predicted output values associated with the one or more components of the vehicle based on applying the correlations to the real-time voltages;
determining whether a state of the one or more components of the vehicle is anomalous by cross-validating the real-time output values with the predicted output values; and
generating an alert if the state of the one or more components is anomalous.
2. The method of claim 1 , wherein constructing the predictive norm model comprises training the predictive norm model with at least a first feature vector of the voltages and the maximum values of the voltages and a second feature vector of the output values and the maximum values of the output values.
3. The method of claim 2 , wherein determining the correlations comprises producing a correlation feature vector based on values of the first feature vector and corresponding values of the second feature vector over the time period.
4. The method of claim 3 , wherein applying the correlations comprises applying the correlation feature vector to the real-time voltages.
5. The method of claim 1 , wherein cross-validating further comprises determining that the state of the one or more components of the vehicle is anomalous when the real-time output values differ from the predicted output values of the predictive norm model.
6. The method of claim 5 , further comprising verifying that the state of the one or more components of the vehicle is anomalous by comparing the real-time output values with one or more other known operating parameters.
7. The method of claim 1 , wherein cross-validating further comprises determining that the state of the one or more components of the vehicle is normal when the real-time output values align with the predicted output values of the predictive norm model.
8. A system to facilitate vehicle diagnostics, the system comprising:
a processor;
a battery monitor to collect voltage of a battery in real-time and to generate a corresponding voltage signal;
a converter to transform the voltage signal into digital signal;
a communications module to collect vehicle information from an in-vehicle network of a vehicle; and
a machine learning engine, under the control of the processor, to:
ingest the digital signal as an input to a predictive norm model of the machine learning engine, wherein the predictive norm model is configured to determine correlations between the battery and other components of the vehicle, and wherein the correlations comprise a first correlation between the voltage of the battery and output values of the other components over a time period and a second correlation between maximum values of the voltage and maximum values of the output values over the time period;
generate, via the predictive norm model, predicted vehicle information based on applying the correlations to the digital signal; and
determine an anomalous state within the other components of the vehicle based on a comparison between the predicted vehicle information and the vehicle information collected by the communications module.
9. The system of claim 8 , wherein the predictive norm model is trained with at least a first feature vector of the voltage and the maximum values of the voltage and a second feature vector of the output values and the maximum values of the output values.
10. The system of claim 9 , wherein determining the correlations comprises producing a correlation feature vector based on values of the first feature vector and corresponding values of the second feature vector over the time period.
11. The system of claim 10 , wherein applying the correlations comprises applying the correlation feature vector to the digital signal.
12. The system of claim 8 , wherein the machine learning engine determines that the state of the other components of the vehicle are anomalous when the predicted vehicle information differs from the vehicle information.
13. The system of claim 12 , wherein, upon determining the state of the other components of the vehicle are anomalous, the machine learning engine validates that the state of the other components of the vehicle are anomalous by comparing the vehicle information to other known correlations.
14. The system of claim 8 , wherein the predicted vehicle information is indicative of the output values of the other components based on the digital signal.
15. The system of claim 8 , wherein the communications module, upon verification that an anomaly has occurred, notifies a vehicle operator that an anomaly has occurred.
16. A system to facilitate vehicle diagnostics, the system comprising:
a memory having stored thereon predictive correlations between a battery and one or more components within a machine, wherein the predictive correlations comprise a first correlation between voltage of the battery and output values of the components over a time period and a second correlation between maximum values of the voltage and maximum values of the output values over the time period;
a diagnostic module configured to:
ingest real-time voltage signals of the machine and generate one or more predicted output values of the components based on applying the predictive correlations to the real-time voltage signals; and
determine an operating condition of the machine by cross-validating the one or more predicted output values with real-time output values of the components, wherein the operating condition includes an anomalous state and a normal state,
wherein the normal state is indicative of a convergence between the one or more predicted output values and the operating condition of the machine, and
wherein the anomalous state is indicative of a divergence between the one or more predicted output values and the operating condition of the machine; and
a communication system to generate an alert if the state of the operating condition of the machine is anomalous.
17. The system of claim 16 , wherein to generate one or more predicted output values, the diagnostic module utilizes a predictive norm model.
18. The system of claim 16 , wherein the diagnostic module determines that the anomalous state occurred in the operating condition of the machine when at least one of the predicted output values differs from a corresponding real-time output value of the components.
19. The system of claim 18 , wherein the diagnostic module, prior to the generation of the alert, validates that the anomalous state has occurred in by comparing the operating condition of the machine to other known correlations.
20. The system of claim 16 , wherein the diagnostic module determines that the normal state occurred in the operating condition of the machine when at least one of the predicted output values aligns with a corresponding real-time output value of the components.