Controller performance test method and system

Abstract

The invention relates to the technical field of controller performance testing, and discloses a controller performance testing method and system. According to the method, operation data of a controller under normal working, fault states and different load conditions are collected, and data cleaning, denoising and normalization preprocessing are carried out. Then, adopting a deep learning algorithm based on LSTM (Long Short Term Memory), and training a performance test model by using the preprocessed data; the model can receive different control signals, predict performance of the controller under corresponding conditions, and perform quantitative evaluation on power consumption, rotating speed, temperature and response time of the controller. In addition, the invention further provides a controller performance test system which comprises a data acquisition module, a data preprocessing module, a model training module, a performance prediction module and a quantitative evaluation module, and automation and intellectualization of the test process are achieved. The application of the invention can comprehensively and accurately reflect the performance of the controller and improve the test efficiency and accuracy.

Description

Technical Field

[0001] The present invention relates to the technical field of controller performance testing, and in particular to a controller performance testing method and a system thereof. Background Art

[0002] Electric tricycles are a common means of transportation, and their controllers are core components. Their performance is directly related to key indicators such as vehicle maneuverability, safety, and energy consumption. However, traditional controller performance testing methods often rely on manual testing and simple data analysis, which not only has low test efficiency, but also makes it difficult to fully and accurately reflect the performance of the controller under different working conditions. In addition, with the continuous development of electric tricycle technology, the complexity and intelligence of controllers are constantly improving, and the requirements for performance testing are also getting higher and higher. Therefore, how to design an efficient and accurate electric tricycle controller performance testing method has become a problem that needs to be solved urgently. Summary of the invention

[0003] In view of the above-mentioned technical deficiencies, an object of the present invention is to provide a controller performance testing method and system thereof to solve the problem of low controller testing efficiency in the prior art.

[0004] In order to solve the above technical problems, the present invention adopts the following technical solutions: In a first aspect, the present invention provides a controller performance testing method, the method comprising: Step S100: collecting controller operation data under normal working state, fault state and different load conditions; Step S200: performing data preprocessing on the collected controller operation data; Step S300: Using a LSTM-based deep learning algorithm, and using the pre-processed controller running data to train a performance test model; Step S400: input different control signals and use the performance test model to predict the performance of the controller; Step S500: Quantitatively evaluate various performance indicators of the controller according to the prediction results of the performance test model.

[0005] Preferably, in a possible implementation manner of the first aspect, the controller operation data includes controller voltage data, controller current data, controller speed data, controller temperature data, controller register value, controller flag status and controller response time data.

[0006] Preferably, in a possible implementation manner of the first aspect, the data preprocessing includes data cleaning, data denoising and data normalization.

[0007] Preferably, in a possible implementation manner of the first aspect, the performance test model training process includes: Divide the preprocessed controller operation data into a training set and a validation set; Use the training set data to train the LSTM-based deep learning algorithm; Use validation set data to verify and adjust the performance test model during training.

[0008] Preferably, in a possible implementation manner of the first aspect, the performance test model structure includes: The input layer is used to receive the pre-processed controller operation data; LSTM layer, which contains multiple long short-term memory units and is used to extract time series features from the data; The fully connected layer receives the output of the LSTM layer and converts it into the predicted value of the controller performance indicator; Output layer, outputs the prediction results of the performance test model.

[0009] Preferably, in a possible implementation manner of the first aspect, the prediction result includes controller power consumption, controller rotation speed, controller temperature and controller response time.

[0010] Preferably, in a possible implementation manner of the first aspect, step S400 specifically includes: Input different control signals into the trained performance test model, the control signals including the controller working mode, the controller fault state and the controller load mode; The performance test model predicts the performance of the controller under corresponding conditions based on the input control signal.

[0011] Preferably, in a possible implementation manner of the first aspect, step S500 specifically includes: According to the prediction results of the performance test model, the controller power consumption, controller speed, controller temperature and controller response time are quantitatively rated; Compare the quantitative evaluation results with the preset performance index threshold to determine whether the controller meets the performance requirements; Generate a performance evaluation report containing the quantitative evaluation results of various performance indicators.

[0012] In a second aspect, the present invention provides a controller performance testing system, the system comprising: A data acquisition module is used to collect the controller operation data under normal working state, fault state and different load conditions; The data preprocessing module is used to perform data cleaning, data denoising and data normalization on the collected controller operation data; The model training module uses a LSTM-based deep learning algorithm and uses the preprocessed controller to run data to train the performance test model; The performance prediction module is used to input different control signals and use the trained performance test model to predict the performance of the controller; The quantitative evaluation module performs quantitative evaluation based on the prediction results of the performance test model and generates a performance evaluation report containing the quantitative evaluation results of various performance indicators.

[0013] The beneficial effects of the present invention are: by collecting the operating data of the controller under normal operation, fault state and different load conditions, and performing data preprocessing, then using the LSTM-based deep learning algorithm to train the performance test model, and finally using the model to predict and quantitatively evaluate the performance of the controller. This method can comprehensively and accurately reflect the performance of the controller under different working conditions, and improve the efficiency and accuracy of the test. At the same time, the present invention also provides a performance testing system, which realizes the automation and intelligence of the test process, and further improves the test efficiency and accuracy. The application of this method and system is of great significance for ensuring the controllability, safety and reducing energy consumption of electric tricycles, and provides strong technical support for the research and development and production of electric tricycle controllers. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0015] Figure 1 A flow chart of a controller performance testing method is provided for this application.

[0016] Figure 2 A controller performance test system structure diagram is provided for this application.

[0017] Explanation of the accompanying drawings: 1-data acquisition module, 2-data preprocessing module, 3-model training module, 4-performance prediction module, 5-quantitative evaluation module. DETAILED DESCRIPTION

[0018] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0019] Embodiment 1: Figure 1 As shown, the present invention provides a controller performance testing method, the method comprising: Step S100: Collecting controller operation data under normal working state, fault state and different load conditions.

[0020] Specifically, the controller operation data includes controller voltage data, controller current data, controller speed data, controller temperature data, controller register value, controller flag bit status and controller response time data.

[0021] The controller voltage data is obtained through the voltage sensor, the controller current data is obtained through the current sensor, the controller speed data is obtained through the speed sensor, the controller temperature data is obtained through the temperature sensor, the controller register value is obtained by directly reading the controller internal register, the controller flag bit status is obtained by accessing the controller status register or 3 status bits, and the controller response time data is obtained by recording the time difference from the controller receiving the instruction to the completion of execution.

[0022] In this embodiment, the voltage sensor adopts the LF210-S type Hall voltage sensor, which is connected to the voltage output terminal of the controller. By measuring the voltage and converting it into an electrical signal, and then processing it through analog-to-digital conversion (ADC), the voltage data of the controller is obtained. The current sensor adopts the LF1005-S / SP16 type Hall current sensor, which is connected in series in the current path of the controller, directly measures the current and converts it into an electrical signal, and then monitors the current data of the control system after ADC processing. The speed sensor adopts the HRS740 type Hall speed sensor, which is installed on the drive shaft of the controller. When the drive shaft rotates, the sensor triggers the Hall effect to generate an electrical signal, and the speed data is calculated by measuring the frequency of the electrical signal. The temperature sensor adopts the B58703M1103A type NTC temperature sensor, which is in contact with the heat dissipation component of the controller, and the temperature is calculated by measuring the change in the resistance value. After ADC conversion, it can be used for temperature monitoring.

[0023] The controller register value is obtained by directly reading the controller's internal register. Use tools such as the devmem command to directly access the physical memory address and read the value of a specific register to obtain information such as the controller's internal state and configuration parameters. The controller flag bit status is obtained by accessing the controller's status register or a specific status bit. Finally, the controller response time data is obtained by recording the time difference from the controller receiving the instruction to the completion of the execution. Set timestamps at the controller's input and output ends respectively, and record the time when the instruction is sent and received. The controller response time is obtained by calculating the time difference.

[0024] The controller operation data of the controller under normal operation, fault state and different load conditions are collected. First, under normal working conditions, the collection of controller operation data is realized through a series of sensors and direct reading of internal registers. The voltage sensor LF210-S, the current sensor LF1005-S / SP16, the speed sensor HRS740 and the temperature sensor B58703M1103A are used to collect the voltage data, current data, speed data and temperature data of the controller respectively. In addition, the controller register value is obtained by directly reading the internal register of the controller through tools such as the devmem command, and the controller flag bit status is obtained by accessing the status register of the controller or a specific status bit. The controller response time data is obtained by recording the time difference from the controller receiving the instruction to the completion of execution. Specifically, the timestamp is set at the input and output ends of the controller, and the time of sending and receiving the instruction is recorded, and the time difference is calculated to obtain it.

[0025] In the fault state, the acquisition process is similar to the normal working state, but the focus is on identifying and recording data under various fault conditions. Fault conditions include controller overheating, voltage abnormality, current overload, unstable speed, register value error, abnormal flag state, and response time delay. For example, when the controller is overheated, the temperature sensor will capture data of abnormal temperature rise; when the voltage is abnormal, the voltage sensor will record a voltage value outside the normal range; when the current is overloaded, the current sensor will detect an excessive current value. These fault data are crucial for subsequent analysis and diagnosis of controller faults.

[0026] Under different load conditions, the acquisition process is also based on the above sensors and the method of directly reading internal registers. Load conditions are divided into different load amounts, including light load conditions, medium load conditions, heavy load conditions and overload conditions. Light load conditions refer to loads less than or equal to 30% of the design load, medium load conditions refer to loads greater than 30% of the design load and less than or equal to 70% of the design load, heavy load conditions refer to loads greater than 70% of the design load and less than or equal to 100% of the design load, and overload refers to loads greater than 100% of the design load. Under different load amounts, the operating data of the controller will be different. For example, parameters such as voltage, current, speed and temperature may increase with the increase of load. By collecting data under different load conditions, the performance and stability of the controller under different working conditions can be analyzed.

[0027] Step S200: preprocessing the collected controller operation data.

[0028] Specifically, data preprocessing includes data cleaning, data denoising, and data normalization.

[0029] In this embodiment, the controller operation data is saved in a DataFrame. First, the fillna method in the pandas library is used to fill the missing values ​​in the data by forward filling (method='ffill'), and the result is directly saved in the original DataFrame (inplace=True). Taking the controller voltage data as an example, a voltage threshold range (voltage_threshold) is set, with a lower limit of 10 and an upper limit of 30. Then, all rows with voltage values ​​within the threshold range are filtered out from the DataFrame by Boolean indexing to remove data that does not meet the requirements. The drop_duplicates method in the pandas library is used, and the result is also saved in the original DataFrame.

[0030] In terms of data denoising, a function named moving_average_filter is defined, which uses a moving average filtering algorithm to smooth the input data to reduce the interference of noise. The input parameters of the function include the data to be processed and the window size (5 in this embodiment). The function first calculates the cumulative sum of the data, then calculates the moving average by sliding the window, and returns the calculation result. Finally, in the data normalization stage, a MinMaxScaler object is created, and the feature columns that need to be normalized ('voltage', 'current', 'speed', and 'temperature') are specified. Then the values ​​of these feature columns are extracted and normalized using the fit_transform method of the MinMaxScaler object.

[0031] Step S300: Using the LSTM-based deep learning algorithm, the preprocessed controller runs data to train the performance test model.

[0032] Specifically, the performance test model training process includes dividing the preprocessed controller operation data into a training set and a validation set; using the training set data to train the LSTM-based deep learning algorithm; using the validation set data to verify and adjust the performance test model during the training process. The preprocessed controller operation data is obtained from the control center.

[0033] The performance test model structure includes: an input layer for receiving preprocessed controller operation data; an LSTM layer, which contains multiple long short-term memory units and is used to extract time series features in the data; a fully connected layer, which receives the output of the LSTM layer and converts it into a predicted value of the controller performance indicator; and an output layer, which outputs the predicted results of the performance test model.

[0034] In this embodiment, the preprocessed controller operation data is divided into a training set and a validation set in an 8:2 ratio, and the training set data is used to train the LSTM-based deep learning algorithm; the validation set data is used to verify and adjust the performance test model during the training process.

[0035] For the process of training the performance test model of the LSTM-based deep learning algorithm, a Sequential model is first created as the basic framework of the neural network. Then an input layer is added to the model. The shape of the input layer is specified by the function parameter input_shape. It is responsible for receiving the input data and passing it to the subsequent layers for processing. Then a long short-term memory network layer (LSTM layer) is added to the model. The LSTM layer contains 100 LSTM units and sets return_sequences=True to return the sequence output. The output of this layer will retain the time dimension of the input sequence so that the subsequent layers can further process it. Then a fully connected layer (Dense layer) is added to the model. The fully connected layer contains 64 neurons and uses the ReLU function as the activation function. This layer is responsible for nonlinearly transforming the output of the LSTM layer and extracting useful feature information. Finally, the output layer is added. The output layer is also a fully connected layer, but it contains 4 neurons, corresponding to the 4 predicted performance indicators: controller power consumption, controller speed, controller temperature, and controller response time. The output of this layer will be used as the final prediction result of the model.

[0036] After building the model, the model is compiled using the Adam optimizer and the mean square error (MSE) loss function, and the LSTM-based deep learning algorithm is trained using the training set data. The Adam optimizer is an optimization algorithm based on gradient descent, which has the characteristics of adaptive learning rate adjustment. The MSE loss function is used to calculate the average of the sum of squared errors between the model's predicted value and the actual value to evaluate the performance of the model. After each round of training, the validation set data is used to verify and adjust the performance test model during the training process.

[0037] Step S400: input different control signals and use the performance test model to predict the performance of the controller.

[0038] Specifically, different control signals are input into the trained performance test model, and the control signals include the controller working mode, the controller fault status and the controller load mode; the performance test model predicts the performance of the controller under corresponding conditions based on the input control signals.

[0039] In this embodiment, the control signal includes normal working state, fault state, and different load conditions. The fault state includes controller overheating, voltage abnormality, current overload, speed instability, register value error, flag state abnormality, and response time delay. Different load conditions include light load condition, medium load condition, heavy load condition, and overload condition. The control signal is input into the performance test model through the programming interface. After receiving the input signal, the model predicts the performance of the controller under the corresponding conditions, including the key performance indicators such as controller power consumption, controller speed, controller temperature, and controller response time.

[0040] In order to ensure the accuracy and reliability of the prediction results, the input control signal is checked and preprocessed. On the one hand, by verifying the legitimacy and validity of the control signal, the interference of invalid or erroneous signals on the model prediction results is avoided. On the other hand, the control signal is properly encoded and formatted to ensure that it can be seamlessly connected with the input interface of the performance test model.

[0041] Step S500: Quantitatively evaluate various performance indicators of the controller according to the prediction results of the performance test model.

[0042] Specifically, based on the prediction results of the performance test model, the controller power consumption, controller speed, controller temperature and controller response time are quantitatively rated; the quantitative evaluation results are compared with the preset performance indicator thresholds to determine whether the controller meets the performance requirements; and a performance evaluation report containing the quantitative evaluation results of each performance indicator is generated.

[0043] In this embodiment, after obtaining the prediction results of the performance test model, a detailed quantitative evaluation is performed on the four key performance indicators of the controller, namely, power consumption, rotation speed, temperature, and response time. First, for the controller power consumption, the power consumption threshold is set to not exceed 120% of the rated power, that is, if the controller power consumption prediction value exceeds 120% of the rated power, it is considered that its power consumption is too high and needs further optimization. At the same time, according to the size of the power consumption prediction value, the power consumption performance is divided into four levels: excellent (≤90% of the rated power), good (90% to 105% of the rated power), general (105% to 120% of the rated power), and unqualified (>120% of the rated power).

[0044] For controller speed, the speed threshold is set to not less than 90% of the design maximum speed and not more than 110% of the design maximum speed. Similarly, based on the speed prediction value, the speed performance is divided into four levels: excellent (100% to 110% of the design maximum speed), good (95% to 100% of the design maximum speed), average (90% to 95% of the design maximum speed) and unqualified (<90% of the design maximum speed).

[0045] In terms of controller temperature evaluation, the temperature threshold is set to no more than 80 degrees Celsius. At the same time, according to the temperature prediction value, the temperature performance is divided into four levels: normal (≤60℃), slightly high (60℃ to 70℃), high (70℃ to 80℃) and too high (>80℃).

[0046] Finally, for the controller response time, the response time threshold is set to no more than 50 milliseconds. Based on the response time prediction value, the response time performance is divided into four levels: excellent (≤20 milliseconds), good (20 milliseconds to 30 milliseconds), average (30 milliseconds to 40 milliseconds), and unqualified (>40 milliseconds, but no more than 50 milliseconds; if it exceeds 50 milliseconds, it is considered severely unqualified).

[0047] After completing the quantitative rating of each performance indicator, these rating results are compared with the preset performance indicator thresholds to generate a performance evaluation report containing the quantitative evaluation results of each performance indicator.

[0048] Embodiment 2: Figure 2 As shown, the present invention provides a controller performance testing system, the system comprising: The data acquisition module 1 is used to collect the controller operation data under normal working state, fault state and different load conditions; The data preprocessing module 2 is used to perform data cleaning, data denoising and data normalization on the collected controller operation data; Model training module 3 uses a LSTM-based deep learning algorithm and uses the preprocessed controller to run data to train the performance test model; The performance prediction module 4 is used to input different control signals and use the trained performance test model to predict the performance of the controller; The quantitative evaluation module 5 performs quantitative evaluation according to the prediction results of the performance test model and generates a performance evaluation report including the quantitative evaluation results of various performance indicators.

[0049] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalents, the present invention is also intended to include these modifications and variations.

Claims

1. A controller performance testing method, characterized in that: The method comprises: Step S100: collecting controller operation data under normal working state, fault state and different load conditions; Step S200: performing data preprocessing on the collected controller operation data; Step S300: Using a LSTM-based deep learning algorithm, and using the pre-processed controller running data to train a performance test model; Step S400: input different control signals and use the performance test model to predict the performance of the controller; Step S500: Quantitatively evaluate various performance indicators of the controller according to the prediction results of the performance test model.

2. The controller performance testing method according to claim 1, characterized in that: The controller operation data includes controller voltage data, controller current data, controller speed data, controller temperature data, controller register value, controller flag bit status and controller response time data.

3. The controller performance testing method according to claim 1, characterized in that: The data preprocessing includes data cleaning, data denoising and data normalization.

4. The controller performance testing method according to claim 1, characterized in that: The performance test model training process includes: Divide the preprocessed controller operation data into a training set and a validation set; Use the training set data to train the LSTM-based deep learning algorithm; Use validation set data to verify and adjust the performance test model during training.

5. The controller performance testing method according to claim 4, characterized in that: The performance test model structure includes: The input layer is used to receive the pre-processed controller operation data; LSTM layer, which contains multiple long short-term memory units and is used to extract time series features from the data; The fully connected layer receives the output of the LSTM layer and converts it into the controller performance indicator prediction value; Output layer, outputs the prediction results of the performance test model.

6. The controller performance testing method according to claim 5, characterized in that: The prediction results include controller power consumption, controller speed, controller temperature and controller response time.

7. The controller performance testing method according to claim 1, characterized in that: The step S400 specifically includes: Input different control signals into the trained performance test model, the control signals including the controller working mode, the controller fault state and the controller load mode; The performance test model predicts the performance of the controller under corresponding conditions based on the input control signal.

8. The controller performance testing method according to claim 1, characterized in that: The step S500 specifically includes: According to the prediction results of the performance test model, the controller power consumption, controller speed, controller temperature and controller response time are quantitatively rated; Compare the quantitative evaluation results with the preset performance index threshold to determine whether the controller meets the performance requirements; Generate a performance evaluation report containing the quantitative evaluation results of various performance indicators.

9. A controller performance testing system, characterized in that: The system comprises: A data acquisition module is used to collect the controller operation data under normal working state, fault state and different load conditions; The data preprocessing module is used to perform data cleaning, data denoising and data normalization on the collected controller operation data; The model training module uses a LSTM-based deep learning algorithm and uses the preprocessed controller to run data to train the performance test model; The performance prediction module is used to input different control signals and use the trained performance test model to predict the performance of the controller; The quantitative evaluation module performs quantitative evaluation based on the prediction results of the performance test model and generates a performance evaluation report containing the quantitative evaluation results of various performance indicators.

Images