Machine learning-based anomaly detection and alert system for oil pumps

1 . A method, performed by a processing device, comprising:

receiving input parameters comprising at least a pump speed command to a pump of a vehicle at a time t, and an electric current of the pump prior to the time t;

applying, to the input parameters, a machine learning model, configured to determine one or more output parameters comprising at least a predicted electric current of the pump at the time t, in response to the pump speed command, wherein the machine learning model is trained to emulate a load response of a healthy pump based at least on the electric current of the pump prior to the time t and in response to the pump speed command; and

determining, based at least on the predicted electric current of the pump and a sensed electric current of the pump at the time t, whether an anomaly is associated with the pump of the vehicle; and

transmitting, by the processing device, a signal to cause the vehicle to present a visual indication of the anomaly or transmitting to a second processing device, an indication of the anomaly.

2 . The method of claim 1 , wherein the input parameters further comprises a fluid temperature associated with the pump, sensed prior to the time t.

3 . The method of claim 1 , wherein the input parameters comprise a first vehicle state that the vehicle is operating the pump to transfer thermal energy to a battery of the vehicle.

4 . The method of claim 3 , wherein the input parameters comprise a second vehicle state that the vehicle is not operating the pump to transfer thermal energy to the battery of the vehicle.

5 . The method of claim 4 , wherein determining whether the anomaly is associated with the pump of the vehicle comprises:

determining a difference between i) a time-series of the sensed electric current and ii) a time-series of the predicted electric current; and

determining that the anomaly is associated with the pump in response to when the difference satisfies a threshold.

6 . The method of claim 5 , wherein determining whether the anomaly is associated with the pump of the vehicle further comprises:

grouping the time-series of the sensed electric current and the time-series of the predicted electric current into first sessions and second sessions, based on the first vehicle state and the second vehicle state; and

determining that the anomaly is associated with the pump in response to when a combined difference of the first sessions satisfies a first threshold.

7 . The method of claim 6 , wherein determining whether the anomaly is associated with the pump of the vehicle further comprises determining that the anomaly is associated with the pump in response to when a threshold number of second sessions is satisfied, and when a second combined difference of a subset of the second sessions satisfies a second threshold.

8 . The method of claim 1 , wherein the pump is configured to pump oil through an electric motor of the vehicle.

9 . The method of claim 1 , wherein the pump is configured to pump coolant through the vehicle.

10 . The method of claim 1 , wherein the anomaly comprises at least one of: an oil leakage, an oil fill amount, pump blockage, or pump debris.

11 . The method of claim 1 , wherein the anomaly comprises a severity of the anomaly.

12 . A server computer system comprising:

a memory; and

one or more processors coupled with the memory configured to perform operations, comprising:

receiving input parameters comprising at least a pump speed command to a pump of a vehicle at a time t, and an electric current of the pump prior to the time t;

applying, to the input parameters, a machine learning model, configured to determine one or more output parameters comprising at least a predicted electric current of the pump at the time t, in response to the pump speed command, wherein the machine learning model is trained to emulate a load response of a healthy pump based at least on the electric current of the pump prior to the time t and in response to the pump speed command; and

determining, based at least on the predicted electric current of the pump and a sensed electric current of the pump at the time t, whether an anomaly is associated with the pump of the vehicle; and

transmitting a signal to cause the vehicle to present a visual indication of the anomaly or transmitting to a second server computer, an indication of the anomaly.

13 . The server computer system of claim 12 , wherein the input parameters further comprises a fluid temperature associated with the pump, sensed prior to the time t.

14 . The server computer system of claim 12 , wherein the input parameters comprise a first vehicle state that the vehicle is operating the pump to transfer thermal energy to a battery of the vehicle.

15 . The server computer system of claim 14 , wherein the input parameters comprise a second vehicle state that the vehicle is not operating the pump to transfer thermal energy to the battery of the vehicle.

16 . The server computer system of claim 15 , wherein determining whether the anomaly is associated with the pump of the vehicle comprises:

determine a difference between i) a time-series of the sensed electric current and ii) a time-series of the predicted electric current; and

determine that the anomaly is associated with the pump in response to when the difference satisfies a threshold.

17 . A non-transitory computer-readable memory storing instructions that, when executed by one or more processors, cause the one or more processors to perform operations, comprising:

receiving input parameters comprising at least a pump speed command to a pump of a vehicle at a time t, and an electric current of the pump prior to the time t;

applying, to the input parameters, a machine learning model, configured to determine one or more output parameters comprising at least a predicted electric current of the pump at the time t, in response to the pump speed command, wherein the machine learning model is trained to emulate a load response of a healthy pump based at least on the electric current of the pump prior to the time t and in response to the pump speed command;

determining, based at least on the predicted electric current of the pump and a sensed electric current of the pump at the time t, whether an anomaly is associated with the pump of the vehicle; and

transmitting a signal to cause the vehicle to present a visual indication of the anomaly or transmitting to a second server computer, an indication of the anomaly.

18 . The computer-readable memory of claim 17 , wherein the input parameters further comprises a fluid temperature associated with the pump, sensed prior to the time t.

19 . The computer-readable memory of claim 17 , wherein the input parameters comprise a first vehicle state that the vehicle is operating the pump to transfer thermal energy to a battery of the vehicle.

20 . The computer-readable memory of claim 19 , wherein the input parameters comprise a second vehicle state that the vehicle is not operating the pump to transfer thermal energy to the battery of the vehicle.

21 . A method, performed by a processing device, comprising:

receiving input parameters comprising at least a pump speed command to a pump of a vehicle at a time t, and an electric current of the pump prior to the time t;

applying, to the input parameters, a machine learning model, configured to determine one or more output parameters comprising at least a predicted electric current of the pump at the time t, in response to the pump speed command; and

determining, based at least on the predicted electric current of the pump and a sensed electric current of the pump at the time t, whether an anomaly is associated with the pump of the vehicle wherein determining whether the anomaly is associated with the pump of the vehicle comprises grouping a time-series of the sensed electric current and a time-series of the predicted electric current into first sessions and second sessions, based on a first vehicle state that the vehicle is operating the pump to transfer thermal energy to a battery of the vehicle, and a second vehicle state that the vehicle is not operating the pump to transfer thermal energy to a battery of the vehicle, and determining that the anomaly is associated with the pump in response to when a combined difference of the first sessions satisfies a first threshold; and

transmitting, by the processing device, a signal to cause the vehicle to present a visual indication of the anomaly or transmitting to a second processing device, an indication of the anomaly.