Variable load degaussing power supply test method and system

Abstract

The invention discloses a variable load degaussing power supply test method and system, and relates to the technical field of degaussing, and the variable load degaussing power supply test method mainly comprises the steps: simulating an actual operation load through a programmable load, carrying out the simulation load test of a degaussing power supply, and collecting test data to obtain historical data; the method comprises the following steps: preprocessing historical data, and dividing the processed data into a training set and a test set; according to the training set and the test set, training an artificial neural network model by using a loss function and an optimizer to obtain a trained artificial neural network model; and obtaining a prediction result by using the trained artificial neural network model according to real-time data of the degaussing power supply. By implementing the variable load degaussing power supply testing method and system provided by the invention, the performance of the degaussing power supply can be conveniently measured and predicted in real time.

Description

Technical Field

[0001] The present invention relates to the technical field of demagnetization, and more particularly to a variable load demagnetization power supply testing method and system. Background Art

[0002] With the continuous advancement of magnetic field detection technology, magnetic field signals retained on ships will become targets for aerial and underwater magnetic field detection, making them vulnerable to magnetic fuze weapons in wartime, posing a significant threat to ship navigation and operational safety. Regular degaussing is becoming an increasingly important task in ship protection. The degaussing power supply plays a crucial role in the degaussing system. A high-performance degaussing power supply can produce excellent degaussing results, so it is necessary to test the performance of the degaussing power supply to ensure the stable and efficient operation of the degaussing system.

[0003] When the current degaussing power supply test system is tested, different test items such as the stability, protection mechanism, and dynamic response time of the degaussing power supply need to be achieved by manually adjusting the load configuration. For example, when the test system switches from an overload test to a short-circuit test, it is first necessary to disconnect the output of the degaussing power supply to ensure that there is no voltage at the load end. Then, the fixed load used for the overload test is removed from the output end of the degaussing power supply. Then, the copper busbar or short-circuit line used for the short-circuit test is connected to the output end of the degaussing power supply. The required measuring equipment must also be connected to the test circuit before the test can begin. For short-circuit testing, the test time should not be too long and should be as short as possible to avoid long-term high current damage to the power supply and test equipment. When testing the load characteristics of the degaussing power supply, frequent switching is also involved for resistors or inductors of different values.

[0004] In the test mode switching process described above, each switch requires rebuilding the test system load and reconnecting the measurement equipment to the test circuit before testing can begin. This switching process is not only cumbersome but also requires special attention to safety.

[0005] The current degaussing power supply test system already has the function of detecting real-time output errors, but lacks the function of further processing the output error data.

[0006] The above content is only used to assist in understanding the technical solution of the present invention and does not constitute an admission that the above content is prior art. Summary of the Invention

[0007] The object of the present invention is to provide a variable load degaussing power supply testing method and system, which can conveniently measure and predict the performance of the degaussing power supply in real time.

[0008] The present invention provides a variable load demagnetization power supply testing method, comprising the following steps: S1: using a programmable load to simulate an actual operating load, performing a simulated load test on the demagnetization power supply, and collecting test data to obtain historical data; S2: preprocessing the historical data, and then dividing the preprocessed historical data into a training set and a test set; S3: using a loss function and an optimizer to train an artificial neural network model based on the training set and the test set, to obtain a trained artificial neural network model; S4: using the trained artificial neural network model to obtain a prediction result based on real-time data of the demagnetization power supply.

[0009] Furthermore, the simulated load test on the degaussing power supply includes an overload test, and the overload test includes: setting the programmable load to a constant current mode; setting the load current of the programmable load, gradually increasing the load current, starting the degaussing power supply and the programmable load, detecting the output voltage, output current, and output power of the degaussing power supply, and recording the test data; and judging whether the protection mechanism of the degaussing power supply is effective based on the test data.

[0010] Furthermore, the overload test includes an overload recovery test, which includes: after the overload test, reducing the load current of the programmable load to below the rated value of the demagnetization power supply, and obtaining a conclusion on whether the demagnetization power supply is working normally based on the output voltage, output current, and output power.

[0011] Furthermore, the simulated load test on the demagnetization power supply also includes a stability test, and the stability test includes: setting the programmable load as a dynamic load, setting the load parameters to change gradually over time, starting the demagnetization power supply and the programmable load, detecting output current data, and updating test data; the output current data includes ripple factor, output current rise time and fall time, and current overshoot, and then judging whether the output of the demagnetization power supply meets the demagnetization requirements based on the test data.

[0012] Furthermore, the historical data includes degaussing power supply input voltage fluctuation, input current, efficiency, degaussing power supply working time, temperature, humidity, aging degree, automatically adjusted feedback current value and corresponding static output error.

[0013] Furthermore, the static output error includes amplitude error, attenuation characteristic error, time error, and overall waveform error.

[0014] Furthermore, step S2 specifically includes: S21: using S22: normalizing the data after the anomaly removal to obtain normalized data; S23: dividing the normalized data in chronological order, and dividing the normalized data into a training set and a test set.

[0015] Furthermore, the artificial neural network model is an LSTM model, the loss function is the mean square error, and the optimizer is an Adam optimizer.

[0016] Furthermore, step S4 specifically includes: inputting the real-time data of the degaussing power supply into the trained artificial neural network model to obtain a prediction error value sequence; traversing the prediction error value sequence to find the time point when the error exceeds the preset error threshold for the first time to obtain a prediction result.

[0017] The present invention also provides a variable load demagnetization power supply test system, which includes the following modules: a test data module, configured to use a programmable load to simulate an actual operating load, perform a simulated load test on the demagnetization power supply, and collect test data to obtain historical data; a data preprocessing module, configured to preprocess the historical data, and then divide the preprocessed historical data into a training set and a test set; a model training module, configured to use a loss function and an optimizer to train an artificial neural network model based on the training set and the test set to obtain a trained artificial neural network model; a prediction module, configured to use the trained artificial neural network model to obtain a prediction result based on the real-time data of the demagnetization power supply.

[0018] The variable load degaussing power supply testing method and system provided by the present invention have the following beneficial effects: The present invention uses a programmable load to perform a simulated load test on a demagnetizing power supply to obtain test data; pre-processes the test data and historical data to obtain a training set and a test set; uses a loss function and an optimizer to train an artificial neural network model based on the training set and the test set to obtain a trained artificial neural network model; and uses the trained artificial neural network model to obtain a prediction result based on the real-time data of the demagnetizing power supply. The present invention proposes introducing a programmable load into a demagnetizing power supply test system and controlling the load configuration of the programmable load through an industrial computer, thereby replacing the manual operation of changing the load according to different test items in the demagnetizing power supply test system, thereby simplifying the test process of the test system; and at the same time, introduces an AI prediction model to predict the output error of the demagnetizing power supply. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] The present invention will be further described below with reference to the accompanying drawings and embodiments, in which: Figure 1This is a flow chart of the variable load degaussing power supply testing method and system provided by the present invention; Figure 2 This is a control interface diagram provided by the present invention; Figure 3 This is a flow chart of the overload test provided by the present invention; Figure 4 This is a stability test flow chart provided by the present invention; Figure 5 This is a flow chart of the LSTM model training provided by the present invention; Figure 6 This is the output error prediction flow chart provided by the present invention. DETAILED DESCRIPTION

[0020] In order to have a clearer understanding of the technical features, purposes and effects of the present invention, specific embodiments of the present invention are now described in detail with reference to the accompanying drawings.

[0021] Figure 1 FIG1 is a schematic diagram showing a variable load degaussing power supply testing method according to an embodiment of the present invention. In this embodiment, the variable load degaussing power supply testing method includes the following steps: S1: Use a programmable load to simulate the actual operating load, perform a simulated load test on the demagnetization power supply, and collect test data to obtain historical data; In an exemplary embodiment, the simulated load test on the degaussing power supply includes an overload test, and the overload test includes: setting the programmable load to a constant current mode; setting the load current of the programmable load, gradually increasing the load current, starting the degaussing power supply and the programmable load, detecting the output voltage, output current, and output power of the degaussing power supply, and recording test data; confirming that the load current exceeds the overcurrent protection threshold of the degaussing power supply, and determining whether the protection mechanism of the degaussing power supply is effective based on the test data; In an exemplary embodiment, the overload test includes an overload recovery test, which includes: after the overload test, reducing the load current of the programmable load to below the rated value of the degaussing power supply, and obtaining a conclusion on whether the degaussing power supply is operating normally based on the output voltage, output current, and output power; In an exemplary embodiment, the simulated load test on the degaussing power supply further includes a stability test, wherein the stability test includes: setting the programmable load as a dynamic load, setting the load parameters to gradually change over time, starting the degaussing power supply and the programmable load, detecting output current data, and updating test data; the output current data includes ripple factor, output current rise time and fall time, and current overshoot, and then determining whether the output of the degaussing power supply meets the degaussing requirements based on the test data; S2: preprocessing the historical data, and then dividing the preprocessed historical data into a training set and a test set; In an exemplary embodiment, the historical data includes the degaussing power supply input voltage fluctuation, input current, efficiency, degaussing power supply operating time, temperature, humidity, aging degree, automatically adjusted feedback current value, and corresponding static output error; In an exemplary embodiment, the static output error includes an amplitude error, an attenuation characteristic error, a time error, and an overall waveform error; In an exemplary embodiment, step S2 specifically includes: S21: using criterion, removing abnormal data from the historical data to obtain data after the abnormalities are removed; S22: normalizing the data after the abnormalities are removed to obtain normalized data; S23: dividing the normalized data obtained in chronological order, and dividing the normalized data into a training set and a test set; S3: training the artificial neural network model using a loss function and an optimizer according to the training set and the test set to obtain a trained artificial neural network model; In an exemplary embodiment, the artificial neural network model is an LSTM model, the loss function is a mean square error, and the optimizer is an Adam optimizer; As an exemplary embodiment, step S3 specifically includes: using the AI ​​toolkit of LabVIEW, loading the TensorFlow framework, defining the artificial neural network model as an LSTM model; defining the input layer, selecting the number of units, and the output layer; selecting the mean square error (MSE) as the loss function and the Adam optimizer as the optimizer, compiling the model, and training the model using the training set to obtain a trained artificial neural network model; S4: Obtaining a prediction result using the trained artificial neural network model according to the real-time data of the demagnetization power supply; In an exemplary embodiment, step S4 specifically includes: inputting the real-time data of the degaussing power supply into a trained artificial neural network model to obtain a prediction error value sequence; traversing the prediction error value sequence to find the time point at which the error first exceeds a preset error threshold, and obtaining a prediction result; As an exemplary embodiment, the preset error threshold is 0.01; As an exemplary embodiment, in step S4, real-time data from the degaussing power supply is fed into a trained LSTM model for prediction. The model then predicts a sequence of error values ​​for a period of time in the future. The predicted error values ​​are then iterated over to find the first operating time that exceeds the set error threshold, and the predicted time point is output.

[0022] This embodiment provides a variable load demagnetization power supply test system, which includes the following modules: a test data module, configured to use a programmable load to simulate an actual operating load, perform a simulated load test on the demagnetization power supply, collect test data, and then divide the historical data into a training set and a test set; a data preprocessing module, configured to preprocess the historical data to obtain a training set and a test set; a model training module, configured to use a loss function and an optimizer to train an artificial neural network model based on the training set and the test set to obtain a trained artificial neural network model; and a prediction module, configured to use the trained artificial neural network model to obtain a prediction result based on the real-time data of the demagnetization power supply.

[0023] In some embodiments, the variable load degaussing power supply test method can also be implemented in the following manner. In this embodiment, the variable load degaussing power supply test method includes: This embodiment introduces a programmable load in the demagnetization power supply test system, which simplifies the test process of the demagnetization power supply: a programmable load is introduced in the current demagnetization power supply test system. The programmable load supports multiple modes such as static load, dynamic load, constant current, constant voltage, etc. to meet different test requirements. After the programmable load is introduced, the test system can quickly switch different load conditions remotely under the control of the software, and can flexibly adjust the load parameters such as current, voltage, power, etc., without the need to simplify the test process like the previous test system. And for different test items, different load configurations are performed, and you only need to select different test items for automatic testing. At the same time, an AI prediction model is introduced to train the AI ​​model with the real-time collected output error data of the actual operation history and the corresponding parameters such as operation time, humidity, temperature, and demagnetization power supply input voltage fluctuation. After that, the real-time data is combined with the AI ​​model to predict the time of the output error upper limit and provide early warning.

[0024] Take the overload test of a single degaussing power supply as an example to design a test project. Connect the programmable load to the degaussing power supply test system and the programmable load to the industrial computer. The industrial computer issues corresponding control instructions and configures the programmable load according to different test items. The control interface based on LabVIEW is as follows: Figure 2 The output test data of the relevant degaussing power supply is sent to the industrial computer through the programmable load. The industrial computer records and saves the data during the test and displays the test data at the same time.

[0025] When the test system needs to verify the output capacity and protection mechanism of the degaussing power supply under overload conditions, first set the programmable load in constant current mode on the industrial computer's control terminal. Gradually increase the load current until it exceeds the degaussing power supply's rated value (such as 120%, 150%, 200% of the rated current). Start the degaussing power supply and programmable load, and begin the overload test. Use the industrial computer to observe the output voltage and current waveforms of the degaussing power supply to verify whether the overcurrent protection is triggered. During this process, the main observation is whether the degaussing power supply's output voltage is stable or decreasing.

[0026] During overload testing, the output voltage, output current, and output power are primarily measured. When the load current does not exceed the degaussing power supply's overcurrent protection threshold, the increased load current causes a voltage drop across the power supply's internal impedance, resulting in a slight decrease in output voltage. The output current and output power increase with increasing load current. When the current measured by the programmable load exceeds the degaussing power supply's threshold, the output data will vary depending on the power supply's protection mechanism. For example, if the degaussing power supply utilizes overcurrent limiting protection, the output current will be limited to the degaussing power supply's set upper current limit, and the output voltage will drop to a lower value to prevent further current increases. The output power is typically limited to a level slightly above the rated power. When the degaussing power supply's protection mechanism is triggered, the time interval between the load current exceeding the set threshold and the actual activation of the protection mechanism must be monitored to determine whether it exceeds the degaussing power supply's design requirements. After the degaussing power supply protection mechanism is triggered, an overload recovery test is performed. First, reduce the programmable load's load current to below the degaussing power supply's rated value. Then, observe the output voltage, current, and power waveforms again to verify that the degaussing power supply is functioning properly and exceeds its design requirements. The relevant test data is recorded. The overload test process is as follows Figure 3 shown.

[0027] It is very necessary to detect the output current stability of the demagnetization power supply according to different loads in the demagnetization power supply test system. The following is a design for testing different loads for the demagnetization power supply test system that introduces a programmable load. After the design of the demagnetization power supply is completed, its output characteristics (such as initial current, decay time constant, output waveform, etc.) are usually fixed according to the design specifications. However, due to the load changes of the demagnetization power supply, testing and verification are required. The test process is as follows: First, set the programmable load to dynamic load mode (the programmable load supports complex load modes, and the combination of R and L can be directly set to perform tests under specific load modes), and set the load parameters to change gradually over time, for example, the load resistance gradually increases from 1Ω to 10Ω, and the duration is 5 seconds. Start the demagnetization power supply and the programmable load for testing. By observing the output current waveform, verify whether the output current of the demagnetization power supply can work normally, and whether its ripple factor, output current rise time and fall time, current overshoot, etc. meet the design requirements of the demagnetization power supply. Such as Figure 4 Shown is a flow chart for stability testing.

[0028] Output current errors in the degaussing power supply can cause the degaussing power supply or the load to operate abnormally for extended periods. This long-term operation can accelerate the aging or wear of the degaussing power supply components, further directly impacting the degaussing system's effectiveness. Therefore, predicting the output current error of the degaussing power supply is crucial to the entire degaussing system. Predicting the output error of the degaussing power supply allows for preventative maintenance, reducing equipment failure rates, downtime, and maintenance costs.

[0029] The following is a design for the error prediction of the demagnetization power supply output current.

[0030] First, it is necessary to collect relevant data, including historical data of the degaussing power supply during normal operation. These historical data include the degaussing power supply input voltage fluctuation, input current, efficiency, degaussing power supply working time, temperature, humidity, aging degree, automatic adjustment feedback current value and corresponding output current error value; After that, the abnormal data in the test is removed and all the data is normalized. The data is then divided into training set and test set, and the data is divided in chronological order.

[0031] Static output errors include: amplitude error, attenuation characteristic error, time error, and overall waveform error; Amplitude error: The deviation between the maximum value of each rectangular wave and the set value: , Attenuation characteristic error: The deviation between the rate at which the current gradually decays and the set value: , Time error: The deviation between the duration of the rectangular wave and the set value: , Overall waveform error: the degree of deviation between the actual waveform and the theoretical waveform: , The following is the design. Use LabVIEW's AI toolkit to load the TensorFlow framework to define the LSTM (Long Short-Term Memory) model. Match the number of input layers with the test variables of the demagnetization power supply test system, and then select the appropriate number of LSTM units (50). The output layer will output the predicted error value. After defining the LSTM network structure, select the appropriate loss function (mean square error MSE) and optimizer (Adam) to compile the model, and then train the model using the collected original data set. The LSTM model training process is as follows: Figure 5 shown.

[0032] The output error prediction of a degaussing power supply exhibits time-dependent and nonlinear characteristics. The LSTM model can effectively capture the long-term dependencies in time series data, enabling accurate output error prediction. The following describes the application of the LSTM model in output error prediction.

[0033] Calculate the mean of each input feature variable using the above historical running data: First, preprocess the data according to the The following are the specific implementation steps to eliminate abnormal data: First, calculate according to the following formula: , Where N is the number of feature input variables, Input variable value for the feature of the i-th data point; According to the calculated mean , calculate the variance: , The standard deviation is ;according to The outliers are removed according to the criterion; the input voltage, input current, temperature and other data of the degaussing power supply may have outliers due to sensor failure or environmental interference; for each input characteristic variable, check whether it meets the , eliminate the data points that do not meet the requirements to ensure the reliability of the data.

[0034] After the data is filtered, the data needs to be further normalized. The dimensions and value ranges of the input feature variables (such as voltage, current, and temperature) are different. Normalization can eliminate the influence of the feature variables. For each input feature variable, find its minimum value. and maximum value , such as the formula: , Map the normalized data to the [0,1] interval.

[0035] Then the LSTM model is trained. The LSTM unit consists of a forget gate, an input gate, an output gate, and a memory unit: Forget Gate: , Input Gate: , , Memory unit update: , Output Gate: , , in, is the input vector of the current time step; is the hidden state of the previous time step; is the memory unit state at the previous time step; 、 、 、 are the weight matrices of the forget gate, input gate, candidate value, and output gate respectively; 、 、 、 They are the bias vectors of the forget gate, input gate, candidate value, and output gate respectively; For the Sigmoid activation function: ; is the hyperbolic tangent activation function; We then measure the difference between the model's predictions and the true values, choosing the mean square error (MSE) as the loss function. We calculate the squared error between each prediction and the true value, and average the error across all samples: , in, is the true value; is the predicted value; When training LSTM models, you may encounter the following issues: long sequences of data can cause vanishing or exploding gradients; traditional gradient descent methods may not be able to effectively escape local minima; and using a fixed learning rate can lead to slow training (when the learning rate is too small) or instability (when the learning rate is too large). To address these issues, a method for efficiently updating model parameters is needed.

[0036] Select the Adam optimizer, which dynamically adjusts the learning rate for each parameter; , , , The Adam optimizer increases the learning rate in regions with small gradients and decreases it in regions with large gradients, thereby accelerating convergence. Adam combines momentum with an adaptive learning rate to accelerate convergence in the gradient direction. Furthermore, Adam's momentum mechanism can help the model escape local minima and saddle points, increasing the likelihood of finding the global optimal solution. Its formula is as follows: , Repeatedly train the LSTM output error prediction model until the loss function converges.

[0037] The final LSTM model output can be achieved through a fully connected layer: , The real-time data of the demagnetization power supply is input into the trained LSTM model for prediction. The model predicts the error value sequence in the future. Then, by traversing the predicted error values, the working time that exceeds the set error threshold for the first time is found, and the predicted time point is output. The output error prediction flow chart is as follows: Figure 6 shown.

[0038] The embodiments of the present invention are described above in conjunction with the accompanying drawings, but the present invention 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 invention, ordinary technicians in this field can also make many forms without departing from the scope of protection of the present invention and the claims, all of which are protected by the present invention.

Claims

1. A variable load degaussing power supply testing method, characterized in that: The following steps are involved: S1: Use a programmable load to simulate the actual operating load, perform a simulated load test on the demagnetization power supply, and collect test data to obtain historical data; S2: preprocessing the historical data, and then dividing the preprocessed historical data into a training set and a test set; S3: training the artificial neural network model using a loss function and an optimizer according to the training set and the test set to obtain a trained artificial neural network model; S4: According to the real-time data of the demagnetization power supply, the trained artificial neural network model is used to obtain a prediction result.

2. The variable load degaussing power supply testing method according to claim 1, characterized in that: The simulated load test on the degaussing power supply includes an overload test, and the overload test includes: setting the programmable load to a constant current mode; setting the load current of the programmable load, gradually increasing the load current, starting the degaussing power supply and the programmable load, detecting the output voltage, output current, and output power of the degaussing power supply, and recording test data; and judging whether the protection mechanism of the degaussing power supply is effective based on the test data.

3. The variable load degaussing power supply testing method according to claim 2, characterized in that: The overload test includes an overload recovery test, which includes: after the overload test, reducing the load current of the programmable load to below the rated value of the degaussing power supply, and obtaining a conclusion on whether the degaussing power supply is working normally based on the output voltage, output current, and output power.

4. The variable load degaussing power supply testing method according to claim 1, characterized in that: The simulated load test on the demagnetization power supply also includes a stability test, which includes: setting the programmable load as a dynamic load, setting the load parameters to gradually change over time, starting the demagnetization power supply and the programmable load, detecting output current data, and updating test data; the output current data includes ripple factor, output current rise time and fall time, and current overshoot, and then judging whether the output of the demagnetization power supply meets the demagnetization requirements based on the test data.

5. The variable load degaussing power supply testing method according to claim 1, characterized in that: The historical data includes the degaussing power supply input voltage fluctuation, input current, efficiency, degaussing power supply working time, temperature, humidity, aging degree, automatically adjusted feedback current value and corresponding static output error.

6. The variable load degaussing power supply testing method according to claim 5, characterized in that: The static output error includes amplitude error, attenuation characteristic error, time error, and overall waveform error.

7. The variable load degaussing power supply testing method according to claim 1, characterized in that: Step S2 specifically includes: S21: Exploitation Criteria, eliminating abnormal data in the historical data to obtain data after eliminating abnormalities; S22: normalizing the data after removing the abnormalities to obtain normalized data; S23: Divide the normalized data into a training set and a test set in chronological order.

8. The variable load degaussing power supply testing method according to claim 1, characterized in that: The artificial neural network model is an LSTM model, the loss function is the mean square error, and the optimizer is an Adam optimizer.

9. The variable load degaussing power supply testing method according to claim 1, characterized in that: Step S4 specifically includes: inputting the real-time data of the degaussing power supply into the trained artificial neural network model to obtain a prediction error value sequence; traversing the prediction error value sequence to find the time point when the error exceeds the preset error threshold for the first time to obtain a prediction result.

10. A variable load degaussing power supply test system, characterized in that: The system includes the following modules: The test data module is configured to: use a programmable load to simulate an actual operating load, perform a simulated load test on the demagnetization power supply, and collect test data to obtain historical data; A data preprocessing module is configured to: preprocess the historical data, and then divide the preprocessed historical data into a training set and a test set; The model training module is configured to: train the artificial neural network model using a loss function and an optimizer according to the training set and the test set to obtain a trained artificial neural network model; The prediction module is configured to obtain a prediction result based on the real-time data of the demagnetization power supply and the trained artificial neural network model.

Images