Nuclear power plant solenoid valve life prediction method and terminal device

Abstract

The invention relates to a nuclear power plant solenoid valve life prediction method and terminal equipment. The method comprises the following steps: based on a set sequence model and an attention mechanism, establishing a multi-layer perceptron model which can perform sequence data processing and perform training by pertinently utilizing set feature data; obtaining a training set and a verification set, training the multi-layer perceptron model by using the training set, determining hyper-parameters of the multi-layer perceptron model by using the verification set, and obtaining the trained multi-layer perceptron model; performing model distillation processing on the trained multi-layer sensor model to obtain a prediction model capable of predicting the service life of the electromagnetic valve; acquiring characteristic data of the tested electromagnetic valve; and inputting the feature data into the prediction model to obtain a prediction result output by the prediction model. The service life of the electromagnetic valve is predicted, and the prediction model has the advantages of being simple in structure and high in prediction accuracy.

Description

Technical Field

[0001] The invention relates to the technical field of nuclear power plant equipment maintenance, and in particular to a life prediction method and terminal equipment for a solenoid valve in a nuclear power plant. Background Art

[0002] Nuclear power plants are equipped with a large number of solenoid valves that participate in the operation and control of nuclear power plants. Therefore, the normal function of the solenoid valves is closely related to the operation of nuclear power plants. At present, nuclear power plants will only replace solenoid valves when they find that the solenoid valves have malfunctions or the solenoid valves have been in service for longer than the set time (i.e. regularly). However, there is a lot of uncertainty in replacing solenoid valves when the solenoid valves have malfunctions. In particular, some solenoid valves used in important control and protection equipment may suddenly fail when an accident occurs in the nuclear power plant, which may lead to the deterioration of the accident due to the inability to perform the corresponding protection actions. Therefore, nuclear power plants are in urgent need of a technical solution that can predict the life of solenoid valves in order to achieve the purpose of replacing solenoid valves in advance. Summary of the invention

[0003] The technical problem to be solved by the present invention is to provide a method for predicting the life of a solenoid valve in a nuclear power plant and a terminal device.

[0004] The technical solution adopted by the present invention to solve the technical problem is: constructing a method for predicting the life of a solenoid valve in a nuclear power plant, comprising:

[0005] Based on the set sequence model and attention mechanism, a multi-layer perceptron model is established that can process sequence data and use the set feature data for targeted training;

[0006] Obtaining a training set and a validation set, training the multilayer perceptron model using the training set, and then determining hyperparameters of the multilayer perceptron model using the validation set to obtain the trained multilayer perceptron model;

[0007] Performing model distillation processing on the trained multi-layer perceptron model to obtain a prediction model capable of predicting the life of the solenoid valve;

[0008] Acquire characteristic data of the solenoid valve under test;

[0009] The feature data is input into the prediction model to obtain the prediction result output by the prediction model.

[0010] Preferably, before using the training set to train the multilayer perceptron model, the method further comprises:

[0011] Performing imbalanced data processing on the training set; wherein the imbalanced data processing includes oversampling and undersampling.

[0012] Preferably, before using the training set to train the multilayer perceptron model, the method further comprises:

[0013] A deep feature learning structure is established in the hidden layers of the multilayer perceptron model.

[0014] Preferably, the step of training the multilayer perceptron model using the training set comprises:

[0015] Based on an ensemble learning algorithm, the multilayer perceptron model is trained using the training set.

[0016] Preferably, the step of determining the hyperparameters of the multilayer perceptron model using the validation set comprises:

[0017] The performance of the trained multilayer perceptron model is evaluated by a cross-validation algorithm and the validation set, so as to adjust the hyperparameters of the multilayer perceptron model by using a hyperparameter tuning algorithm according to the evaluation results.

[0018] Preferably, before performing model distillation processing on the trained multilayer perceptron model, the method further includes:

[0019] S21, obtaining a test set; wherein the test set includes a plurality of sample data, and each sample data included in the test set is different from all sample data included in the training set and the validation set;

[0020] S22, inputting the test set into the trained multilayer perceptron model to obtain a test result output by the trained multilayer perceptron model;

[0021] S23, judging whether the test result meets the set requirements, if yes, allowing the trained multilayer perceptron model to be subjected to model distillation processing, if no, adjusting the hyperparameters of the multilayer perceptron model, and executing S24;

[0022] S24, obtaining a test set to be updated that is different from the training set, the validation set, and the used test set, defining the test set to be updated as the new test set, and returning to S22.

[0023] Preferably, the set characteristic data includes the coil current, coil voltage, coil resistance, coil inductance, coil pull-in time and coil release time of the solenoid valve, as well as the opening time and closing time of the solenoid valve.

[0024] The present invention also constructs a terminal device, comprising:

[0025] A control unit is used to run a computer program and implement the steps of the method for predicting the life of a solenoid valve in a nuclear power plant provided by an embodiment of the present invention when executing the computer program.

[0026] Preferably, the terminal device further includes:

[0027] An adjustable power supply, connected to the control unit, and configured to output a test power supply according to a power supply control signal output by the control unit;

[0028] Multifunctional acquisition card, used to measure the coil resistance and coil inductance of the solenoid valve under test;

[0029] The switching matrix is ​​connected to the adjustable power supply, the control unit, the multifunctional acquisition card and the solenoid valve under test, and is used to cooperate with the control unit, the adjustable power supply and the multifunctional acquisition card to test the coil current, coil voltage, coil resistance and coil inductance of the solenoid valve under test.

[0030] Preferably, the adjustable power supply includes a digital potential circuit and a voltage stabilization control circuit;

[0031] The digital potential circuit is connected to the control unit to output a regulation signal according to the power control signal, and the voltage stabilization control circuit is connected to the digital potential circuit to output an output voltage of the test power supply according to the regulation signal;

[0032] The switching matrix includes a first relay K2, a second relay K19, a resistor R20, a resistor R21, a resistor R14, a resistor R38 and an optocoupler relay U10;

[0033] The first common contact of the first relay K2 is connected to the first end of the photoreceptor of the optical coupling relay U10, the first normally closed contact is connected to the positive pole of the test power supply of the voltage stabilization control circuit, the second common contact is connected to the second normally closed contact of the second relay K19 via the resistor R38, the second normally closed contact is grounded, the positive end of the excitation coil is connected to the first DC voltage, and the negative end of the excitation coil is connected to the control unit. The second end of the photoreceptor of the optical coupling relay U10 is connected to the second common contact of the first relay K2 via the resistor R20 and the resistor R21. The connection node of the resistor R21 is connected to the multifunctional acquisition card, the anode of the light-emitting body of the optical coupling relay U10 is connected to the first DC voltage, and the cathode of the light-emitting body is connected to the control unit via the resistor R14, the first common contact of the second relay K19 is connected to the first end of the coil of the electromagnetic valve under test, the first normally closed contact is connected to the second end of the light-receiving body of the optical coupling relay U10, and the second common contact is connected to the second end of the coil of the electromagnetic valve under test, and the first common contact, the first normally open contact, the second common contact and the second normally open contact of the second relay K19 are also connected to the multifunctional acquisition card respectively;

[0034] When measuring the coil current and the coil voltage, the control unit controls the optocoupler relay U10 to be turned on, the first relay K2 to be demagnetized, and the second relay K19 to be demagnetized, so as to measure the second normally closed contact voltage of the second relay K19 and the connection node voltage of the resistor R20 and the resistor R21 through the multi-functional acquisition card, and enables the control unit to calculate the coil current and the coil voltage according to the connection node voltage of the resistor R20 and the resistor R21 and the second normally closed contact voltage of the second relay K19;

[0035] When measuring the coil resistance and the coil inductance, the control unit controls the optocoupler relay U10 to turn off, the first relay K2 to demagnetize, and the second relay K19 to be energized, so that the first common contact, the first normally open contact, the second common contact, and the second normally open contact of the second relay K19 are also respectively connected to the multifunctional acquisition card to measure the coil resistance and the coil inductance through the multifunctional acquisition card.

[0036] The implementation of the present invention has the following beneficial effects: a method for predicting the life of a solenoid valve in a nuclear power plant is provided; first, based on a set sequence model and an attention mechanism, a multilayer perceptron model capable of processing sequence data and targetedly training with set feature data is established; then, the multilayer perceptron model is trained with a training set, and the hyperparameters of the multilayer perceptron model are determined with a validation set to obtain a trained multilayer perceptron model; then, the trained multilayer perceptron model is subjected to model distillation processing to obtain a prediction model capable of predicting the life of the solenoid valve; finally, the feature data of the solenoid valve under test is obtained, and the feature data is input into the prediction model to obtain a prediction result capable of characterizing the remaining life of the solenoid valve under test, thereby realizing the life prediction of the solenoid valve, and the prediction model has the advantages of simple structure and high prediction accuracy, which plays a positive role in improving the operational stability of the nuclear power plant. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] The present invention will be further described below with reference to the accompanying drawings and embodiments, in which:

[0038] Figure 1 is a program flow chart of a method for predicting the life of a solenoid valve in a nuclear power plant in some embodiments of the present invention;

[0039] Figure 2 is a model structure diagram of a multilayer perceptron model in some embodiments of the present invention;

[0040] Figure 3 is a flowchart of a program for training a multilayer perceptron model with a test set in some embodiments of the present invention;

[0041] Figure 4is a circuit structure block diagram of a terminal device in some embodiments of the present invention;

[0042] Figure 5 is a circuit schematic diagram of a digital potentiometer circuit in some embodiments of the present invention;

[0043] Figure 6 is a circuit schematic diagram of a voltage stabilizing control circuit in some embodiments of the present invention;

[0044] Figure 7 is a circuit schematic diagram of a multifunctional acquisition card in some embodiments of the present invention;

[0045] Figure 8 is a circuit schematic diagram of a switching matrix in some embodiments of the present invention;

[0046] Fig. 9 is a circuit schematic diagram of a main control circuit in some embodiments of the present invention;

[0047] Fig.10 is a circuit schematic diagram of a Darlington array in some embodiments of the present invention. DETAILED DESCRIPTION

[0048] 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.

[0049] It should be noted that the flowcharts shown in the accompanying drawings are only exemplary and do not necessarily include all the contents and operations / steps, nor do they have to be executed in the order described. For example, some operations / steps can be decomposed, while some operations / steps can be combined or partially combined, so the actual execution order may change according to actual conditions.

[0050] The block diagrams shown in the accompanying drawings are merely functional entities and do not necessarily correspond to physically independent entities. That is, these functional entities may be implemented in software form, or in one or more hardware modules or integrated circuits, or in different networks and / or processor devices and / or microcontroller devices.

[0051] The present invention provides a method for predicting the life of a solenoid valve in a nuclear power plant. Figure 1 The method may include step S10, step S20, step S30, step S40 and step S50.

[0052] Step S10 includes: establishing a multilayer perceptron model (MLP model for short) that can process sequence data and use set feature data for training in a targeted manner based on a set sequence model and an attention mechanism. In this step, an initial multilayer perceptron model is first constructed, and then a set sequence model is introduced into the initial multilayer perceptron model, so that the multilayer perceptron model can process sequence data, such as sequence data with time series characteristics such as the opening time and closing time of the solenoid valve; then, an attention mechanism is introduced into the initial multilayer perceptron model, so that the multilayer perceptron model can pay more attention to the set feature data during the training process, and the set feature data are all data that are closely related to the life of the solenoid valve. The set feature data can specifically include the coil current, coil voltage, coil resistance, coil inductance, coil pull-in time and coil release time of the solenoid valve, as well as the opening time and closing time of the solenoid valve. It can be understood that the introduction of the attention mechanism helps to improve the prediction accuracy of the model.

[0053] It should be noted that the multilayer perceptron model can be constructed using the MLP algorithm, where the structure of the multilayer perceptron model can be referenced Figure 2 , the activation function of the multilayer perceptron model can be: The multilayer perceptron model has the following advantages:

[0054] 1. It is highly flexible and can be applied to various types of problems. It can perform classification, regression and clustering tasks on data such as coil current, coil voltage, coil resistance, coil inductance, coil pull-in time and coil release time of the solenoid valve;

[0055] 2. Excellent nonlinear modeling capabilities, able to learn and represent complex nonlinear relationships, and therefore able to train and analyze nonlinear data in set feature data or feature data;

[0056] 3. Strong adaptability. It can adapt to different data and problems by adjusting the size of the hidden layer, the number of layers, the activation function and other hyperparameters, which helps to improve the life prediction performance of the model;

[0057] 4. Good parallel computing capabilities, which can perform efficient training and reasoning in a parallel computing environment, not only fast training speed, but also help reduce the amount of calculation of the control unit (such as GPU);

[0058] 5. It has strong feature learning ability and can automatically learn and extract important features from input data (i.e. set feature data), without the need to manually design features, thus reducing the burden of feature analysis and learning process;

[0059] 6. Strong generalization ability. Under appropriate regularization and optimization strategies, it has strong generalization ability and can accurately predict new data (i.e. the characteristic data of the tested solenoid valve).

[0060] Understandably, compared to deep learning models such as deep convolutional neural networks, the multilayer perceptron model has the advantages of simple structure, low hardware configuration requirements, and the output results of the model are relatively easy to understand and explain.

[0061] In order to improve the accuracy of model life prediction and improve the efficiency of the multi-layer perceptron model in capturing dependencies and sequence patterns on time series, in some embodiments, the sequence model can be set to a recurrent neural network or a long short-term memory network.

[0062] Step S20 includes: obtaining a training set and a validation set, training the multilayer perceptron model using the training set, and then using the validation set to determine the hyperparameters of the multilayer perceptron model to obtain a trained multilayer perceptron model.

[0063] In this step, taking the training set as an example, the training set may include sample data measured at multiple different time sequences during the service of multiple solenoid valves, and each sample data may include the following measured at the corresponding time sequence of the corresponding solenoid valve: the coil current, coil voltage, coil resistance, coil inductance, coil pull-in time and coil release time of the solenoid valve, the temperature and humidity of the environment in which the solenoid valve is located, the opening time and closing time of the solenoid valve, and the actual remaining life of the solenoid valve. It should be noted that the sample data can be obtained by testing multiple solenoid valves using existing testing equipment, such as using an oscilloscope to test the change curve of the coil current and the change curve of the coil voltage, and a resistance tester to test the coil resistance, etc.; the coil pull-in time and the coil release time can be determined by analyzing the change curve of the coil current and the change curve of the coil voltage; the opening time and closing time of the solenoid valve can be determined by testing the medium flow or air pressure change on the pipeline where the solenoid valve is located, for example, the moment when the fluid flow drops to near zero can be determined as the closing time of the solenoid valve, and the moment when the fluid flow rises from a value near zero to greater than the set opening threshold can be determined as the opening time of the solenoid valve.

[0064] In order to prevent the problem of imbalanced data distribution in the training set from affecting the prediction performance of the trained multilayer perceptron model, in some embodiments, before executing the step of training the multilayer perceptron model using the training set, the following step may also be included: performing imbalanced data processing on the training set; wherein the imbalanced data processing includes oversampling and undersampling.

[0065] In some embodiments, before executing the step of training the multilayer perceptron model using the training set, it also includes: establishing a deep feature learning structure in the hidden layer of the multilayer perceptron model. Specifically, a deep feature learning structure such as a convolutional layer (i.e., a convolutional neural network) can be entered into the first N layers of the multiple hidden layers of the multilayer perceptron model, so that the multilayer perceptron model can automatically learn useful or even more representative features from the training set, thereby improving the prediction performance. Wherein, N is a natural number, which can be customized according to actual needs.

[0066] In some embodiments, the step of training the multilayer perceptron model using the training set may include: training the multilayer perceptron model using the training set based on an ensemble learning algorithm. It is understandable that training the multilayer perceptron model using the training set based on the ensemble learning algorithm can combine the prediction results of the learners of multiple multilayer perceptron models, thereby improving the prediction accuracy and robustness of the multilayer perceptron model. The ensemble learning algorithm may include at least one of a Bagging algorithm, a Boosting algorithm, and a Stacking algorithm.

[0067] In some embodiments, the step of determining the hyperparameters of the multilayer perceptron model using the validation set may include: evaluating the performance of the trained multilayer perceptron model through a cross-validation algorithm and a validation set, and adjusting the hyperparameters of the multilayer perceptron model using a hyperparameter tuning algorithm according to the evaluation results. The cross-validation algorithm may be any one of a leave-one-out validation algorithm, a K-fold cross-validation algorithm, and a stratified K-fold cross-validation algorithm; and the hyperparameter tuning algorithm may be a grid search algorithm or a random search algorithm. It is understandable that the prediction performance of the multilayer perceptron model can be accurately evaluated through the cross-validation algorithm and the validation set, and then the hyperparameters of the multilayer perceptron model can be adjusted according to the evaluation results in combination with the hyperparameter tuning algorithm, so as to achieve the tuning of the hyperparameters, thereby improving the prediction capability of the multilayer perceptron model. After determining the hyperparameters, the trained multilayer perceptron model may be obtained.

[0068] In some embodiments, Figure 3 As shown, before performing model distillation processing on the trained multilayer perceptron model, steps S21, S22, S23 and S24 may be performed to further improve the prediction accuracy of the multilayer perceptron model.

[0069] Step S21 includes: obtaining a test set; wherein the test set includes a plurality of sample data, and each sample data included in the test set is different from all sample data included in the training set and the validation set.

[0070] Step S22 includes: inputting the test set into the trained multilayer perceptron model to obtain the test result output by the trained multilayer perceptron model.

[0071] It should be noted that the component structures of the test set and the validation set are the same as those of the training set, so the methods for obtaining the sample data included in the three are the same. The main difference is that the sample data included in the test set are different from each other, and each sample data in the test set is different from any sample data in the validation set and the training set. It can be understood that since each sample data in the test set corresponds to the sample data measured at a certain time sequence during the service period of a certain solenoid valve, and the retirement time of the solenoid valve is known, the actual remaining life of the solenoid valve can be obtained by subtracting the service time sequence corresponding to the sample data from the retirement time. In some embodiments, assuming that the total service time of a solenoid valve is 12 months, if the test personnel collect sample data of the solenoid valve once a month, 11 sample data corresponding to the solenoid valve can be obtained in the end, and the 11 sample data constitute a test set A. When executing step S22, the sample data in the test set A will be input into the trained multilayer perceptron model in sequence, and then 11 corresponding remaining life estimation results will be obtained. Then, taking the sample data of the 5th month as an example, if the sample of the 5th month is input into the trained multilayer perceptron model, the remaining life estimation result obtained is X months, and the error percentage between X and 7 (equal to the difference between the total service time of 12 months and the current time series of 5 months) is calculated, and the error percentage is used to locate the remaining life estimation result corresponding to the sample data. After obtaining the remaining life estimation results of all sample data, the average value of the remaining life estimation results of all sample data is calculated. It can be understood that the average value can represent the prediction accuracy of the multilayer perceptron model, so the average value can be output as the test result.

[0072] Step S23 includes: judging whether the test result meets the set requirements, if so, allowing the trained multilayer perceptron model to be distilled, otherwise adjusting the hyperparameters of the multilayer perceptron model, and executing step S24. Specifically, the step of judging whether the test result meets the set requirements may include: when the average value is greater than the set accuracy, judging that the test result meets the set requirements, and when the average value is not greater than the set accuracy, judging that the test result does not meet the set requirements. The setting range of the set accuracy can be 80% to 95%.

[0073] Step S24 includes: obtaining a test set to be updated that is different from the training set, the validation set, and the used test set, defining the test set to be updated as a new test set, and returning to step S22. It should be noted that the method for obtaining sample data of the test set to be updated is the same as that of the test set, but each sample data included in the test set to be updated is different from any sample data included in the validation set, the training set, and all the test sets used in the previous execution of step S22.

[0074] Step S30 includes: performing model distillation processing on the trained multi-layer perceptron model to obtain a prediction model capable of predicting the life of the solenoid valve. This step implements model distillation processing through an existing model distillation method, thereby obtaining a model with a more streamlined structure, stronger generalization ability and higher computational efficiency, that is, obtaining the prediction model.

[0075] Step S40 includes: obtaining characteristic data of the solenoid valve under test. In this step, the characteristic data of the solenoid valve under test may include the coil current, coil voltage, coil resistance, coil inductance, coil pull-in time and coil release time of the solenoid valve under test, the temperature and humidity of the environment in which the solenoid valve under test is located, and the opening time and closing time of the solenoid valve under test. The method for obtaining the characteristic data of the solenoid valve under test is similar to the method for obtaining the sample data, except that the solenoid valve object obtained is different, so it will not be repeated here.

[0076] Step S50 includes: inputting the characteristic data into the prediction model to obtain the prediction result output by the prediction model. In this step, the output prediction result may be the remaining predicted life of the solenoid valve under test, which may be displayed by a human-computer interaction unit (such as a display screen).

[0077] The technical solution of the present invention first establishes a multilayer perceptron model capable of processing sequence data and training with set feature data based on a set sequence model and an attention mechanism; then, the multilayer perceptron model is trained with a training set, and the hyperparameters of the multilayer perceptron model are determined with a validation set to obtain a trained multilayer perceptron model; then, the trained multilayer perceptron model is subjected to model distillation processing to obtain a prediction model capable of predicting the life of the solenoid valve; finally, the feature data of the solenoid valve under test is obtained, and the feature data is input into the prediction model to obtain a prediction result capable of characterizing the remaining life of the solenoid valve under test, thereby realizing the life prediction of the solenoid valve, and the prediction model has the advantages of simple structure and high prediction accuracy, which plays a positive role in improving the operating stability of the nuclear power plant.

[0078] like Figure 4As shown, the present invention further provides a terminal device, which includes a control unit. The control unit is used to run a computer program, and when executing the computer program, the steps of the method for predicting the life of a solenoid valve in a nuclear power plant provided by an embodiment of the present invention are implemented.

[0079] In some embodiments, Figure 4 As shown, the terminal device may also include an adjustable power supply, a multi-function acquisition card and a switching matrix.

[0080] See also Figure 4 The adjustable power supply is connected to the control unit, and the adjustable power supply is used to output the test power supply according to the power control signal output by the control unit.

[0081] In some embodiments, Figure 5 and Figure 6 As shown, the adjustable power supply may include a digital potential circuit and a voltage stabilization control circuit. The digital potential circuit is connected to the control unit to output an adjustment signal according to the power control signal, and the voltage stabilization control circuit is connected to the digital potential circuit to output the output voltage of the test power supply according to the adjustment signal.

[0082] In some embodiments, Figure 5 As shown, the digital potentiometer circuit may include a digital potentiometer U8, a resistor R43, an adjustable resistor R49, a resistor R44, a resistor R45, a resistor R42, a resistor R46 and a capacitor C36. The digital potentiometer U8 may be a digital potentiometer chip of model AD5270. Specifically, the SYNC pin, SCLK pin, DIN pin, and SDO pin of the digital potentiometer U8 are respectively connected to the control unit to receive the voltage regulation signal output by the control unit for controlling the output voltage of the test power supply. The W pin of the digital potentiometer U8 is connected to the first end of the adjustable resistor R49 and is connected to the ground via the capacitor C36. The second end of the adjustable resistor R49 is connected to the adjustable end and is connected to the ground via the resistor R44, the resistor R45, and the resistor R46. The connection node of the resistor R44 and the resistor R45 is connected to the A pin of the digital potentiometer U8 via the resistor R42. The SDO pin of the digital potentiometer U8 is also connected to the third DC voltage (the corresponding network label is VCC_3.3V) via the resistor R43. The connection node of the resistor R45 and the resistor R46 is connected to the voltage stabilization control circuit to input the regulation signal (the corresponding network label is ADJ). It can be understood that the digital potentiometer U8, resistor R43, adjustable resistor R49, resistor R44, resistor R45, resistor R42 and resistor R46 constitute a voltage divider circuit. The control unit can control the resistance value of the digital potentiometer U8 by outputting a voltage regulation signal, and then control the voltage of the regulation signal to achieve control of the output voltage of the test power supply.

[0083] In some embodiments, Figure 6As shown, the voltage stabilization control circuit may include a voltage stabilization control chip U11, a resistor RSE21, a resistor Rsense1, a resistor Rramp1, a capacitor Cramp1, a capacitor Css1, a capacitor Ccomp1, an inductor L5, an NMOS tube M1, an NMOS tube M2, a resistor Rfbt1, a diode D115, and a diode D114. Among them, the voltage stabilization control chip U11 may be a voltage stabilization control chip of model LM5118MH. Specifically, the EN pin of the voltage stabilization control chip U11 is connected to the second DC voltage (corresponding network number is +24V_I N), VI via the resistor RSE21. The N pin is connected to the second DC voltage, the FB pin is connected to the VOUT pin via the resistor Rfbt1 and connected to the digital potentiometer circuit to access the adjustment signal, the VOUT pin is also connected to the cathode of the diode D114 and outputs the test power supply, the RAMP pin is connected to its VCC pin via the resistor Rramp1 and connected to the ground via the capacitor Cramp1, the SS pin is connected to the ground via the capacitor Css1, the COMP pin is connected to its FB pin via the capacitor Ccomp1, and the anode of the diode D114 is connected to the voltage stabilization control chip U1 via the inductor L5. The HS pin of the voltage stabilizing control chip U11 is connected to the HS pin of the voltage stabilizing control chip U11, and the other is connected to the drain of the NMOS tube M2. The gate of the NMOS tube M2 is connected to the LO pin of the voltage stabilizing control chip U11, and the source is grounded. The HS pin of the voltage stabilizing control chip U11 is also connected to the source of the NMOS tube M1 and the cathode of the diode D115. The anode of the diode D115 is connected to the CS pin of the voltage stabilizing control chip U11 and is connected to the CSG pin of the voltage stabilizing control chip U11 and the ground through the resistor Rsense1. The gate of the NMOS tube M1 is connected to the HO pin of the voltage stabilizing control chip U11, and the drain is connected to the second DC voltage. It can be understood that the voltage regulator control chip U11, the resistor RSE21, the resistor Rsense1, the inductor L5, the NMOS tube M1, the NMOS tube M2, the resistor Rfbt1, the diode D115 and the diode D114 constitute a boost / buck switching power supply. The voltage regulator control chip U11 can control the conduction and shutdown of the NMOS tube M1 and the NMOS tube M2 according to the voltage of the ADJ signal, thereby controlling the output voltage of the test power supply (the corresponding network label is HVDC+), and can achieve 24V to 50V DC voltage control, so as to drive solenoid valves with different working voltages for life prediction.

[0084] Furthermore, if Figure 5 As shown, the voltage stabilization control circuit may also include peripheral electronic components such as capacitor Cout1, capacitor Cin1, resistor Ruv1, resistor Ruv2, etc. For specific electronic components and connection structures, please refer to Figure 5 , I will not go into details here.

[0085] See also Figure 4The multifunctional acquisition card is connected to the control unit, and the multifunctional acquisition card is used to measure the voltage, and the coil resistance and coil inductance of the electromagnetic valve under test. The multifunctional acquisition card can be an existing LCR tester, as long as it can measure resistance, inductance and voltage, and is not limited here. Further, the circuit diagram of the multifunctional acquisition card can be referred to Figure 7 .

[0086] See also Figure 4 The switching matrix is ​​connected with the adjustable power supply, the control unit, the multi-function acquisition card and the solenoid valve under test. The switching matrix is ​​used to cooperate with the control unit, the adjustable power supply and the multi-function acquisition card to test the coil current, coil voltage, coil resistance and coil inductance of the solenoid valve under test.

[0087] In some embodiments, Figure 8As shown, the switching matrix may include a first relay K2, a second relay K19, a resistor R20, a resistor R21, a resistor R14, a resistor R38 and an optocoupler relay U10. The first relay K2 and the second relay K19 may be relays of model G6S-2-10DC12. Specifically, the first common contact of the first relay K2 (i.e., the 4th pin of the first relay K2) is connected to the first end of the photoreceptor of the optocoupler relay U10, the first normally closed contact of the first relay K2 (i.e., the 4th pin of the first relay K2) is connected to the positive pole of the test power supply of the voltage stabilization control circuit, the second common contact of the first relay K2 (i.e., the 9th pin of the first relay K2) is connected to the second normally closed contact of the second relay K19 (i.e., the 10th pin of the second relay K19) and the multi-function acquisition card (not shown in the figure, the multi-function acquisition card also has) via a resistor R38, the second normally closed contact of the first relay K2 (i.e., the 10th pin of the first relay K2) is grounded, the positive end of the excitation coil of the first relay K2 (i.e., the 1st pin of the first relay K2) is connected to the first DC voltage (the corresponding network label is 12V), the negative end of the excitation coil of the first relay K2 (i.e., the 12th pin of the first relay K2) is connected to the control unit, and the second end of the photoreceptor of the optocoupler relay U10 It is connected to the second common contact of the first relay K2 via resistors R20 and R21, the connection node of resistors R20 and R21 is connected to the multi-function acquisition card, the anode of the light-emitting body of the optocoupler relay U10 is connected to the first DC voltage, the cathode of the light-emitting body of the optocoupler relay U10 is connected to the control unit via resistor R14, the first common contact of the second relay K19 (i.e., the 4th pin of the second relay K19) is connected to the first end of the coil of the solenoid valve under test, the first normally closed contact of the second relay K19 (i.e., the 3rd pin of the second relay K19) is connected to the second end of the photoreceptor of the optocoupler relay U10, the second common contact of the second relay K19 (i.e., the 9th pin of the second relay K19) is connected to the second end of the coil of the solenoid valve under test, and the first common contact, the first normally open contact (i.e., the 5th pin of the second relay K19), the second common contact and the second normally open contact (i.e., the 8th pin of the second relay K19) of the second relay K19 are also connected to the multi-function acquisition card respectively.

[0088] See also Figure 8When measuring the coil current and coil voltage, the control unit controls the optocoupler relay U10 to be turned on, the first relay K2 to be demagnetized, and the second relay K19 to be demagnetized, so that the positive end of the measuring power supply, the first normally closed contact, the first common contact of the first relay K2, the optocoupler relay U10, the first normally closed contact of the second relay K19, the first common contact of the second relay K19, the coil of the solenoid valve to be measured (i.e., the first and second common contacts of the second relay K19), the second normally closed contact of the second relay K19, the resistor R38, the second common contact of the first relay K2, and the second normally closed contact of the first relay K2 form a loop, and then the second normally closed contact voltage of the second relay K19 and the connection node voltage of the resistor R20 and the resistor R21 are measured through the multi-functional acquisition card, and the control unit can calculate the coil current and coil voltage according to the connection node voltage of the resistor R20 and the resistor R21 and the voltage across the resistor R38. In this embodiment, the multifunctional acquisition card is communicatively connected with the control unit, so after measuring the second normally closed contact voltage of the second relay K19 and the connection node voltage of the resistor R20 and the resistor R21, the relevant measured voltage can be sent to the control unit. It can be understood that after the control unit obtains the connection node voltage of the resistor R20 and the resistor R21, it can first restore the total voltage after the coil series connection, the resistor R38, the first and the second relays of the measured solenoid valve according to the pre-stored resistance ratio of the resistor R20 and the resistor R21 and the connection node voltage of the resistor R20 and the resistor R21 (that is, the second normally closed contact voltage of the second relay K19). Since the conduction impedance of the first and second relays is very small and can be ignored, the coil voltage can be obtained by subtracting the voltage across the resistor R38 (that is, the voltage at point ADDC-1) from the restored total voltage. Since the resistance value of the resistor R38 is known in advance, the current flowing through the resistor R38 can be calculated based on Ohm's law to obtain the coil current.

[0089] See also Figure 8 When measuring the coil resistance and coil inductance, the control unit controls the optocoupler relay U10 to turn off, the first relay K2 to demagnetize, and the second relay K19 to energize, so that the first common contact, the first normally open contact, the second common contact, and the second normally open contact of the second relay K19 are also connected to the multi-function acquisition card respectively, which is equivalent to connecting the two ends of the coil of the solenoid valve under test to the multi-function acquisition card in parallel. In this way, the multi-function acquisition card can measure the coil resistance through the four-wire method and measure the coil inductance at the same time.

[0090] In some embodiments, the control unit may include Fig. 9 The main control circuit shown and Fig.10 The Darlington array U6 is shown.

[0091] like Fig. 9As shown, the main control circuit may include a microprocessor U1 and related peripheral electronic devices (including capacitors C1, C2, etc.). The connection relationship between the microprocessor U1 and the related peripheral electronic devices can be referred to Fig. 9 In addition, the microprocessor U1 may be a microprocessor of model STM32L151 RCT6.

[0092] like Figure 8 , Fig. 9 , Fig.10 As shown, the Darlington array U6 connects the microprocessor U1, the first relay K2 and the second relay K19. Since the control signal voltage output by the microprocessor U1 is not greater than 5V, the first relay K2 and the second relay K19 cannot be directly controlled. The function of the Darlington array U6 is to amplify the relay control signal (including CPUEN_001, CPUEN_002, ..., CPUEN_007) output by the microprocessor U1 so that the microprocessor U1 can control the excitation or demagnetization of the first relay K2 and the second relay K19. In addition, the Darlington array U6 can be a Darlington tube with a model of ULN2803A.

[0093] In this specification, each embodiment is described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the embodiments can be referred to each other. For the device disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the relevant parts can be referred to the method part.

[0094] Professionals may further appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of the two. In order to clearly illustrate the interchangeability of hardware and software, the composition and steps of each example have been generally described in the above description according to function. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professionals and technicians may use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of the present invention.

[0095] The steps of the method or algorithm described in conjunction with the embodiments disclosed herein may be implemented directly using hardware, a software module executed by a processor, or a combination of the two. The software module may be placed in a random access memory (RAM), a memory, a read-only memory (ROM), an electrically programmable ROM, an electrically erasable programmable ROM, a register, a hard disk, a removable disk, a CD-ROM, or any other form of storage medium known in the art.

[0096] It can be understood that the above embodiments only express the preferred implementation modes of the present invention, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the patent scope of the present invention. It should be pointed out that, for ordinary technicians in this field, the above technical features can be freely combined without departing from the concept of the present invention, and several deformations and improvements can be made, which all belong to the protection scope of the present invention. Therefore, all equivalent changes and modifications made to the scope of the claims of the present invention should belong to the coverage of the claims of the present invention.

Claims

1. A method for predicting the life of a solenoid valve in a nuclear power plant, characterized in that: include: Based on the set sequence model and attention mechanism, a multi-layer perceptron model is established that can process sequence data and use the set feature data for targeted training; Obtaining a training set and a validation set, training the multilayer perceptron model using the training set, and then determining hyperparameters of the multilayer perceptron model using the validation set to obtain the trained multilayer perceptron model; Performing model distillation processing on the trained multi-layer perceptron model to obtain a prediction model capable of predicting the life of the solenoid valve; Acquire characteristic data of the solenoid valve under test; The feature data is input into the prediction model to obtain the prediction result output by the prediction model.

2. The method for predicting the life of a solenoid valve in a nuclear power plant according to claim 1, characterized in that: Before using the training set to train the multilayer perceptron model, the method further includes: Performing imbalanced data processing on the training set; wherein the imbalanced data processing includes oversampling and undersampling.

3. The method for predicting the life of a solenoid valve in a nuclear power plant according to claim 2, characterized in that: Before using the training set to train the multilayer perceptron model, the method further includes: A deep feature learning structure is established in the hidden layers of the multilayer perceptron model.

4. The method for predicting the life of a solenoid valve in a nuclear power plant according to claim 3, characterized in that: The step of training the multilayer perceptron model using the training set comprises: Based on an integrated learning algorithm, the multilayer perceptron model is trained using the training set.

5. The method for predicting the life of a solenoid valve in a nuclear power plant according to claim 1, characterized in that: The step of determining the hyperparameters of the multilayer perceptron model using the validation set comprises: The performance of the trained multilayer perceptron model is evaluated by a cross-validation algorithm and the validation set, so as to adjust the hyperparameters of the multilayer perceptron model by using a hyperparameter tuning algorithm according to the evaluation results.

6. The method for predicting the life of a solenoid valve in a nuclear power plant according to claim 5, characterized in that: Before performing model distillation processing on the trained multi-layer perceptron model, the method further includes: S21, obtaining a test set; wherein the test set includes a plurality of sample data, and each sample data included in the test set is different from all sample data included in the training set and the validation set; S22, inputting the test set into the trained multilayer perceptron model to obtain a test result output by the trained multilayer perceptron model; S23, judging whether the test result meets the set requirements, if yes, allowing the trained multilayer perceptron model to be subjected to model distillation processing, if no, adjusting the hyperparameters of the multilayer perceptron model, and executing S24; S24, obtaining a test set to be updated that is different from the training set, the validation set, and the used test set, defining the test set to be updated as the new test set, and returning to S22.

7. The method for predicting the life of a solenoid valve in a nuclear power plant according to any one of claims 1 to 6, characterized in that: The set characteristic data include the coil current, coil voltage, coil resistance, coil inductance, coil pull-in time and coil release time of the solenoid valve, as well as the opening time and closing time of the solenoid valve.

8. A terminal device, characterized in that: include: A control unit is used to run a computer program and implement the steps of the method for predicting the life of a solenoid valve in a nuclear power plant as described in any one of claims 1 to 6 when executing the computer program.

9. The terminal device according to claim 8, characterized in that: Also includes: An adjustable power supply, connected to the control unit, and configured to output a test power supply according to a power supply control signal output by the control unit; A multifunctional acquisition card, connected to the control unit, for measuring the coil resistance and coil inductance of the solenoid valve under test; The switching matrix is ​​connected to the adjustable power supply, the control unit, the multifunctional acquisition card and the solenoid valve under test, and is used to cooperate with the control unit, the adjustable power supply and the multifunctional acquisition card to test the coil current, coil voltage, coil resistance and coil inductance of the solenoid valve under test.

10. The terminal device according to claim 9, characterized in that: The adjustable power supply includes a digital potential circuit and a voltage stabilization control circuit; The digital potential circuit is connected to the control unit to output a regulation signal according to the power control signal, and the voltage stabilization control circuit is connected to the digital potential circuit to output an output voltage of the test power supply according to the regulation signal; The switching matrix includes a first relay K2, a second relay K19, a resistor R20, a resistor R21, a resistor R14, a resistor R38 and an optocoupler relay U10; The first common contact of the first relay K2 is connected to the first end of the photoreceptor of the optical coupling relay U10, the first normally closed contact is connected to the positive pole of the test power supply of the voltage stabilization control circuit, the second common contact is connected to the second normally closed contact of the second relay K19 and the multifunctional acquisition card through the resistor R38, the second normally closed contact is grounded, the positive end of the excitation coil is connected to the first DC voltage, and the negative end of the excitation coil is connected to the control unit, the second end of the photoreceptor of the optical coupling relay U10 is connected to the second common contact of the first relay K2 through the resistor R20 and the resistor R21, and the resistor R20 The connection node of the resistor R21 is connected to the multifunctional acquisition card, the anode of the light-emitting body of the optical coupling relay U10 is connected to the first DC voltage, and the cathode of the light-emitting body is connected to the control unit via the resistor R14, the first common contact of the second relay K19 is connected to the first end of the coil of the electromagnetic valve under test, the first normally closed contact is connected to the second end of the light-receiving body of the optical coupling relay U10, and the second common contact is connected to the second end of the coil of the electromagnetic valve under test, and the first common contact, the first normally open contact, the second common contact and the second normally open contact of the second relay K19 are also connected to the multifunctional acquisition card respectively; When measuring the coil current and the coil voltage, the control unit controls the optocoupler relay U10 to be turned on, the first relay K2 to be demagnetized, and the second relay K19 to be demagnetized, so as to measure the second normally closed contact voltage of the second relay K19 and the connection node voltage of the resistor R20 and the resistor R21 through the multi-functional acquisition card, and enables the control unit to calculate the coil current and the coil voltage according to the connection node voltage of the resistor R20 and the resistor R21 and the second normally closed contact voltage of the second relay K19; When measuring the coil resistance and the coil inductance, the control unit controls the optocoupler relay U10 to turn off, the first relay K2 to demagnetize, and the second relay K19 to be energized, so that the first common contact, the first normally open contact, the second common contact, and the second normally open contact of the second relay K19 are also respectively connected to the multifunctional acquisition card to measure the coil resistance and the coil inductance through the multifunctional acquisition card.

Images