Abnormality detection system for automobile window

Abstract

The invention relates to the technical field of data processing, and discloses an automobile window anomaly detection system, which is characterized in that the tire inner volume of a target vehicle is acquired, and current data and running states of N window motors, tire pressure data and temperature data of M tires and a damping coefficient of a suspension system are acquired according to a fixed time interval u; based on the tire inner volume and temperature data, performing uniform dimension processing on the tire pressure data of each tire at the corresponding moment to obtain standard tire pressure data of each tire; based on the running states of the car window motors, screening the current data of the N car window motors to obtain first screening data and second screening data; constructing a current prediction model based on the first screening data and the second screening data; and in a second preset time period, performing anomaly detection on the current data at each moment based on predicted current data output by the current prediction model to obtain an anomaly detection result.

Description

Technical Field

[0001] The present invention relates to the technical field of data processing, and more particularly to an abnormality detection system for automobile windows. Background Art

[0002] The existing management of the lifting and lowering of automobile windows is based on motor drive, specifically: a current parameter threshold is set. When the real-time current parameter of the motor is greater than the current parameter threshold, it usually means that the movement of the window is actively hindered (for example, the window is hindered by a person or an object, which may cause a pinching problem). At this time, the window is stopped. The existing technology has the following problems: when the vehicle is bumpy, the window will also bump, and the real-time parameters of the motor will also change accordingly, resulting in abnormal stop or lowering of the window. Summary of the invention

[0003] The present invention provides an abnormality detection system for automobile windows, which solves the technical problems raised in the background technology.

[0004] The present invention provides an abnormality detection system for a vehicle window, comprising:

[0005] The data acquisition module is used to collect Get the target vehicle's tire volume within a certain time interval. Collect: current data and operating status of N window motors, tire pressure data and temperature data of M tires, and damping coefficient of suspension system;

[0006] A dimension unification module is used to unify the tire pressure data of each tire at a corresponding moment based on the tire internal volume and temperature data to obtain the standard tire pressure data of each tire;

[0007] A data screening module, used for screening the current data of N window motors based on the running status of the window motors to obtain first screening data and second screening data;

[0008] A model building module, for building a current prediction model based on the first screening data and the second screening data; the output of the current prediction model represents the predicted current data of the window motor in the working state within the second preset time period; wherein the working state includes a rising state and a falling state;

[0009] The abnormality detection module is used to perform abnormality detection on the current data at each moment within a second preset time period based on the predicted current data output by the current prediction model to obtain abnormality detection results; wherein the abnormality detection results include: normal current state and abnormal current state.

[0010] Furthermore, data is collected at fixed time intervals, including:

[0011] In the first preset time period, it is determined that at least one of the N window motors is in working state at the i-th moment, and then the current data of the corresponding window motor at the i-th moment, the tire pressure data and temperature data of the M tires, and the damping coefficient of the suspension system are collected. , Is a positive integer.

[0012] Further, unified dimensional processing includes:

[0013] Set dimension temperature ;

[0014] Based on the tire volume and the temperature data of the mth tire at the i-th moment, the tire pressure data of the mth tire at the i-th moment is converted into dimensioned temperature The standard tire pressure data is calculated as follows: ;

[0015] in, represents the standard tire pressure data of the mth tire at the i-th moment, represents the tire pressure data of the mth tire at the i-th moment, represents dimensionless temperature, represents the temperature data of the mth tire at the i-th moment, represents the adjustment weight, Indicates the tire volume. , represents the natural base, , Is a positive integer.

[0016] Further, obtaining the first screening data and the second screening data includes:

[0017] The operating states include an ascending state, a descending state and a standby state; according to the operating states of the N window motors at the i-th moment, the current data of the window motors with the same operating state are divided into the same type to obtain the current data of the ascending window, the current data of the descending window and the current data of the standby window at the i-th moment; the current data of the ascending window is used as the first screening data, and the current data of the descending window is used as the second screening data.

[0018] Further, the current prediction model includes a rising prediction unit and a falling prediction unit;

[0019] Get the Euclidean distance between each window motor and each tire in the target vehicle, and construct training data. The training data includes: sample data and sample labels, as follows:

[0020] Initial sample data is constructed. Each element of the initial sample data is constructed based on the standard tire pressure data of M tires at the i-th moment and the i+1-th moment. For the m-th element m(n) in the initial sample data, the calculation and determination process is as follows:

[0021] Calculate the absolute value of the difference between the standard tire pressure data of the mth tire at the ith moment and the i+1th moment; multiply the absolute value by the Euclidean distance between the nth window motor and the mth tire to obtain the mth element in the initial sample data;

[0022] Based on the damping coefficient of the suspension system at the i-th moment, the initial sample data is processed: ;

[0023] in, represents the mth element in the standard sample data, represents the damping coefficient of the suspension system at the i-th moment, represents the damping coefficient weight;

[0024] The standard sample data is normalized and used as sample data of the current prediction model, and the current data of the nth window motor at the i-th moment is normalized and used as sample labels.

[0025] Furthermore, the rising prediction unit includes: if the current data of the nth window motor at the i-th moment is the first screening data, then based on the corresponding sample data and sample labels, the rising prediction unit is trained to obtain the rising prediction unit; wherein the rising prediction unit is constructed based on a one-dimensional convolutional neural network.

[0026] Furthermore, the descent prediction unit includes: if the current data of the nth window motor at the i-th moment is the second screening data, then based on the corresponding sample data and sample labels, the descent prediction unit is trained to obtain the descent prediction unit; wherein the descent prediction unit is constructed based on a one-dimensional convolutional neural network.

[0027] Furthermore, the loss functions of the ascending prediction unit and the descending prediction unit are both mean square error loss functions.

[0028] Furthermore, an abnormality detection result is obtained, including: obtaining the current data and operating status of the nth window motor at the kth moment of the target vehicle within the second preset time period, the tire pressure data and temperature data of M tires, and the damping coefficient of the suspension system; if the operating status of the nth window motor is a rising state or a falling state, constructing an input sample at the kth moment; wherein the input sample is constructed in the same manner as the sample data of the current prediction model; based on the operating status of the nth window motor, the input sample is input into the rising prediction unit or the falling prediction unit to obtain the predicted current data output by the current prediction model at the kth moment; if the absolute difference between the predicted current data and the current data of the nth window motor at the kth moment is ≤ the preset threshold, the current state is normal, otherwise the current state is abnormal.

[0029] The beneficial effect of the present invention is that by comprehensively collecting multi-dimensional data such as window motor current, operating status, tire pressure and temperature, and suspension system damping coefficient, and using unified dimension processing and a current prediction model based on a one-dimensional convolutional neural network, the abnormal state of the window can be accurately predicted and detected. Compared with the traditional single threshold detection method, the present invention effectively overcomes the problem of misjudgment of current fluctuations caused by vehicle bumps, greatly improves the safety of window operation and detection accuracy, and thus avoids the risk of accidental stop or injury caused by misjudgment. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] Figure 1 It is a module diagram of an abnormality detection system for a car window of the present invention. DETAILED DESCRIPTION

[0031] The subject matter described herein will now be discussed with reference to example embodiments. It should be understood that the discussion of these embodiments is only to enable those skilled in the art to better understand and implement the subject matter described herein, and the functions and arrangements of the elements discussed may be changed without departing from the scope of protection of the contents of this specification. Each example may omit, replace or add various processes or components as needed. In addition, the features described relative to some examples may also be combined in other examples.

[0032] like Figure 1 As shown, a vehicle window abnormality detection system includes:

[0033] The data acquisition module is used to collect Get the target vehicle's tire volume within a certain time interval. Collect: current data and operating status of N window motors, tire pressure data and temperature data of M tires, and damping coefficient of suspension system;

[0034] A dimension unification module is used to unify the tire pressure data of each tire at a corresponding moment based on the tire internal volume and temperature data to obtain the standard tire pressure data of each tire;

[0035] A data screening module, used for screening the current data of N window motors based on the running status of the window motors to obtain first screening data and second screening data;

[0036] A model building module, for building a current prediction model based on the first screening data and the second screening data; the output of the current prediction model represents the predicted current data of the window motor in the working state within the second preset time period; wherein the working state includes a rising state and a falling state;

[0037] The abnormality detection module is used to perform abnormality detection on the current data at each moment within a second preset time period based on the predicted current data output by the current prediction model to obtain abnormality detection results; wherein the abnormality detection results include: normal current state and abnormal current state.

[0038] It should be noted that the tire pressure data, temperature data and the damping coefficient of the suspension system are all obtained based on the existing sensors of the vehicle. When a car encounters a bump, it is first reflected in the tire pressure data of the car. Secondly, it is filtered by the active suspension system of the car and finally reflected to the window motor. Therefore, the tire pressure data of the car, the temperature data in the car tire, and the damping coefficient of the suspension system can be transmitted to the window motor with a nonlinear effect, thereby interfering with the current data of the window motor. The interference of the current data is fitted by a neural network, so that the predicted current value at each moment can be directly obtained based on the data of the car sensor. Based on the predicted current value as a standard, the current parameters at each moment are compared, and the abnormal state is judged based on the threshold.

[0039] Comparative data between traditional threshold detection and the present invention:

[0040] Specifically, it can be seen from the comparison data that the present invention does not cause abnormal stopping or lowering of the car window during bumpy road sections, and the reaction time to pressure is significantly better than the traditional threshold detection method.

[0041] In one embodiment of the present invention, collecting data at fixed time intervals includes:

[0042] In the first preset time period, it is determined that at least one of the N window motors is in working state at the i-th moment, and then the current data of the corresponding window motor at the i-th moment, the tire pressure data and temperature data of the M tires, and the damping coefficient of the suspension system are collected. , Is a positive integer.

[0043] Specifically, collecting data at fixed time intervals ensures the continuity and consistency of the data, providing a reliable time series data basis for subsequent data processing and model building; on the other hand, by collecting data only when the window motor is in working condition, it can ensure that the collected information is closely related to the actual operating status, avoiding the collection of idle data or noise data.

[0044] In one embodiment of the present invention, the unified dimension processing includes:

[0045] Set dimension temperature ;

[0046] Based on the tire volume and the temperature data of the mth tire at the i-th moment, the tire pressure data of the mth tire at the i-th moment is converted into dimensioned temperature The standard tire pressure data is calculated as follows: ;

[0047] in, represents the standard tire pressure data of the mth tire at the i-th moment, represents the tire pressure data of the mth tire at the i-th moment, represents the dimensionless temperature, represents the temperature data of the mth tire at the i-th moment, represents the adjustment weight, Indicates the tire volume. , represents the natural base, , Is a positive integer.

[0048] Specifically, the tire pressure data is processed in a unified dimension, that is, by setting a standard dimension temperature and combining the tire volume with the temperature data corresponding to the current moment, the original tire pressure data is converted to obtain standard tire pressure data under a unified temperature condition. The reason is that tire pressure is greatly affected by temperature, and tire pressure data under different temperature environments fluctuates greatly. Directly using the original data may cause deviations in subsequent model training and anomaly detection; and through unified dimension processing, the influence of temperature differences on tire pressure can be eliminated, thereby ensuring the consistency and comparability of the data, providing more accurate and stable sample data for the current prediction model, and further improving the accuracy of the system's detection of abnormal window conditions.

[0049] In one embodiment of the present invention, obtaining the first screening data and the second screening data includes:

[0050] The operating states include an ascending state, a descending state and a standby state; according to the operating states of the N window motors at the i-th moment, the current data of the window motors with the same operating state are divided into the same type to obtain the current data of the ascending window, the current data of the descending window and the current data of the standby window at the i-th moment; the current data of the ascending window is used as the first screening data, and the current data of the descending window is used as the second screening data.

[0051] Specifically, the operating states of the window motors are divided into rising state, falling state and standby state, and then the current data in the same state are classified into one category according to the state of each motor at the i-th moment, and finally the current data in the rising state is used as the first screening data, and the current data in the falling state is used as the second screening data. The reason for this is that the current characteristics of the window motors are significantly different under different operating states, and extracting the data in the rising and falling states separately helps to build a more targeted current prediction model and improve the accuracy of anomaly detection; at the same time, eliminating the data in the standby state can reduce the interference of noise and irrelevant information on model training.

[0052] In one embodiment of the present invention, the current prediction model includes a rising prediction unit and a falling prediction unit;

[0053] Get the Euclidean distance between each window motor and each tire in the target vehicle, and construct training data. The training data includes: sample data and sample labels, as follows:

[0054] The initial sample data is constructed based on the standard tire pressure data of M tires at the i-th moment and the i+1-th moment: ;

[0055] in, represents the mth element in the initial sample data, Indicates The standard tire pressure data of the mth tire at a certain moment, represents the Euclidean distance between the nth window motor and the mth tire, , is a positive integer; the initial sample data is constructed in a vector structure, and the dimension of the initial sample data is M;

[0056] Based on the damping coefficient of the suspension system at the i-th moment, the initial sample data is processed: ;

[0057] in, represents the mth element in the standard sample data, represents the damping coefficient of the suspension system at the i-th moment, represents the damping coefficient weight;

[0058] The standard sample data is normalized and used as sample data of the current prediction model, and the current data of the nth window motor at the i-th moment is normalized and used as sample labels.

[0059] Specifically, the Euclidean distance between each window motor and each tire is obtained, and the initial sample data is constructed according to the standard tire pressure data of M tires at the i-th moment and the i+1-th moment. The initial sample data is then adjusted in combination with the damping coefficient and its weight of the suspension system at the i-th moment to obtain the standard sample data, which is used as the input of the current prediction model. At the same time, the actual current data of the window motor at the corresponding moment is used as the sample label to train the prediction units corresponding to the rising and falling states in the model.

[0060] In one embodiment of the present invention, a rise prediction unit includes: if the current data of the nth window motor at the i-th moment is the first screening data, then based on the corresponding sample data and sample labels, a rise prediction unit is trained to obtain the rise prediction unit; wherein the rise prediction unit is constructed based on a one-dimensional convolutional neural network.

[0061] Specifically, when the current data of the nth window motor at the i-th moment belongs to the first screening data (i.e., data corresponding to the rising state), the sample data and sample labels corresponding to the state are used to train a one-dimensional convolutional neural network to construct a prediction unit for predicting the current change in the rising state. The one-dimensional convolutional neural network has good local feature extraction capabilities in processing time series data and can capture the dynamic changes in the current of the window during the rising process.

[0062] In one embodiment of the present invention, a descent prediction unit includes: if the current data of the nth window motor at the i-th moment is the second screening data, then based on the corresponding sample data and sample labels, a descent prediction unit is trained to obtain the descent prediction unit; wherein the descent prediction unit is constructed based on a one-dimensional convolutional neural network.

[0063] Specifically, when the window motor is in the lowering state, that is, its current data belongs to the second screening data, the system uses the corresponding sample data and sample labels to build a lowering prediction unit through one-dimensional convolutional neural network training. The current change characteristics of the window during the lowering process may be different from those in the rising state. By establishing a prediction unit for the lowering state alone, the accuracy of abnormality detection can be improved.

[0064] In one embodiment of the present invention, the loss functions of the ascending prediction unit and the descending prediction unit are both mean square error loss functions.

[0065] Specifically, both the rising prediction unit and the falling prediction unit use mean square error (MSE) as the loss function during the training process. The mean square error can quantify the difference between the predicted current value and the actual collected current data, so that the model can more accurately correct the wrong prediction during the training process. In addition, the continuity and differentiability of MSE provide good support for the gradient-based optimization method, which helps the stable convergence of the rising prediction unit and the falling prediction unit, and ultimately improves the overall anomaly detection accuracy.

[0066] In one embodiment of the present invention, an abnormality detection result is obtained, including: obtaining the current data and operating status of the nth window motor of the target vehicle at the kth moment within the second preset time period, the tire pressure data and temperature data of M tires, and the damping coefficient of the suspension system; if the operating status of the nth window motor is a rising state or a falling state, constructing an input sample at the kth moment; wherein the input sample is constructed in the same manner as the sample data of the current prediction model; based on the operating status of the nth window motor, the input sample is input into the rising prediction unit or the falling prediction unit to obtain the predicted current data output by the current prediction model at the kth moment; if the absolute difference between the predicted current data and the current data of the nth window motor at the kth moment is ≤ the preset threshold, the current state is normal, otherwise the current state is abnormal.

[0067] Specifically, the trained current prediction model is used to predict the window motor at the kth moment, and compared with the actual current data collected to determine whether the current state is normal. Specifically, when the nth window motor is detected to be in a rising or falling state, the input sample is constructed in the same way as the model training, and then the input sample is input into the corresponding rising or falling prediction unit to obtain the predicted current data. The predicted current data is used to reflect the predicted current data of the window motor when it rises or falls based on the bumpiness of the road surface at the kth moment. If the difference between the predicted current data and the actual current data exceeds the preset threshold, it means that the window can be subjected to additional force, resulting in abnormal current of the window motor, so there is a risk of pinching people or objects.

[0068] The above describes an embodiment of the present embodiment, but the present embodiment is not limited to the above-mentioned specific implementation methods. The above-mentioned specific implementation methods are merely illustrative and not restrictive. Under the guidance of the present embodiment, ordinary technicians in this field can also make many forms, all of which are within the protection of the present embodiment.

Claims

1. A vehicle window abnormality detection system, characterized in that: include: The data acquisition module is used to collect Get the target vehicle's tire volume within a certain time interval. Collect: current data and operating status of N window motors, tire pressure data and temperature data of M tires, and damping coefficient of suspension system; A dimension unification module is used to unify the tire pressure data of each tire at a corresponding moment based on the tire internal volume and temperature data to obtain the standard tire pressure data of each tire; A data screening module, used for screening the current data of N window motors based on the running status of the window motors to obtain first screening data and second screening data; A model building module, for building a current prediction model based on the first screening data and the second screening data; the output of the current prediction model represents the predicted current data of the window motor in the working state within the second preset time period; wherein the working state includes a rising state and a falling state; The abnormality detection module is used to perform abnormality detection on the current data at each moment within a second preset time period based on the predicted current data output by the current prediction model to obtain abnormality detection results; wherein the abnormality detection results include: normal current state and abnormal current state.

2. The abnormality detection system for automobile windows according to claim 1, characterized in that: Collect data at regular intervals, including: In the first preset time period, it is determined that at least one of the N window motors is in working state at the i-th moment, and then the current data of the corresponding window motor at the i-th moment, the tire pressure data and temperature data of the M tires, and the damping coefficient of the suspension system are collected. , Is a positive integer.

3. The abnormality detection system for automobile windows according to claim 2, characterized in that: Unified dimensional processing, including: Set dimension temperature ; Based on the tire volume and the temperature data of the mth tire at the i-th moment, the tire pressure data of the mth tire at the i-th moment is converted into dimensioned temperature The standard tire pressure data below.

4. The abnormality detection system for automobile windows according to claim 3, characterized in that: The first screening data and the second screening data are obtained, including: The operating states include an ascending state, a descending state and a standby state; according to the operating states of the N window motors at the i-th moment, the current data of the window motors with the same operating state are divided into the same type to obtain the current data of the ascending window, the current data of the descending window and the current data of the standby window at the i-th moment; the current data of the ascending window is used as the first screening data, and the current data of the descending window is used as the second screening data.

5. The abnormality detection system for automobile windows according to claim 4, characterized in that: A current prediction model, including a rising prediction unit and a falling prediction unit; Get the Euclidean distance between each window motor and each tire in the target vehicle, and construct training data. The training data includes: sample data and sample labels, as follows: Initial sample data is constructed. Each element of the initial sample data is constructed based on the standard tire pressure data of M tires at the i-th moment and the i+1-th moment. For the m-th element m(n) in the initial sample data, the calculation and determination process is as follows: Calculate the absolute value of the difference between the standard tire pressure data of the mth tire at the ith moment and the i+1th moment; multiply the absolute value by the Euclidean distance between the nth window motor and the mth tire to obtain the mth element in the initial sample data; Based on the damping coefficient of the suspension system at the i-th moment, the initial sample data is processed; The standard sample data is normalized and used as sample data of the current prediction model, and the current data of the nth window motor at the i-th moment is normalized and used as sample labels.

6. The abnormality detection system for automobile windows according to claim 5, characterized in that: The rising prediction unit includes: if the current data of the nth window motor at the i-th moment is the first screening data, then the rising prediction unit is trained based on the corresponding sample data and sample labels; wherein the rising prediction unit is constructed based on a one-dimensional convolutional neural network.

7. The abnormality detection system for automobile windows according to claim 6, characterized in that: The descent prediction unit includes: if the current data of the nth window motor at the i-th moment is the second screening data, then the descent prediction unit is trained based on the corresponding sample data and sample labels; wherein the descent prediction unit is constructed based on a one-dimensional convolutional neural network.

8. The abnormality detection system for automobile windows according to claim 7, characterized in that: The loss functions of the upward prediction unit and the downward prediction unit are both mean square error loss functions.

9. The abnormality detection system for automobile windows according to claim 8, characterized in that: Obtaining an abnormality inspection result, including: obtaining the current data and operating status of the nth window motor of the target vehicle at the kth moment within the second preset time period, the tire pressure data and temperature data of M tires, and the damping coefficient of the suspension system; if the operating status of the nth window motor is a rising state or a falling state, constructing an input sample at the kth moment; wherein the input sample is constructed in the same manner as the sample data of the current prediction model; based on the operating status of the nth window motor, inputting the input sample into the rising prediction unit or the falling prediction unit to obtain the predicted current data output by the current prediction model at the kth moment; if the absolute difference between the predicted current data and the current data of the nth window motor at the kth moment is ≤ a preset threshold, the current state is normal, otherwise the current state is abnormal.

Images