Simulation modeling method, device and equipment for multi-way valve

Abstract

The invention discloses a simulation modeling method, device and equipment for a multi-way valve, and belongs to the technical field of engineering machine.The method comprises the steps that a pressure flow characteristic test is conducted on a target multi-way valve, and a test data set is obtained; preprocessing all test data in the test data set to obtain a sample data set; constructing a multi-way valve simulation model, wherein the multi-way valve simulation model comprises a multi-way valve flow model and a multi-way valve pressure model; and respectively training the multi-way valve flow model and the multi-way valve pressure model through the sample data set to obtain a final multi-way valve simulation model. The multi-way valve simulation model is divided into the multi-way valve flow model and the multi-way valve pressure model, the multi-way valve flow model and the multi-way valve pressure model are trained respectively, the output of the multi-way valve flow model serves as the input of the multi-way valve pressure model in model application, the model application efficiency and real-time performance are guaranteed, and the accuracy of the training model is also guaranteed.

Description

Technical Field

[0001] The present invention relates to the technical field of engineering machinery, and in particular to a simulation modeling method, device and equipment for a multi-way valve. Background Art

[0002] At present, in the process of development of construction machinery, construction machinery is constantly developing towards refinement, greening and diversification of operations. Modeling and simulation testing has become an indispensable step in the research and development optimization of construction machinery. Due to the harsh working environment and complex movements of construction machinery, the modeling and simulation of the whole machine needs to take into account both model accuracy and simulation real-time performance. As the key hub of construction machinery motion control, multi-way valves especially need to be accurately and real-timely simulated in research and development.

[0003] Due to the diversity of types and complex movements of construction machinery, the internal structure of the multi-way valve, which is the hub of construction machinery movement control, is complex and diverse. The standard hydraulic library of the currently popular hydraulic system simulation software only has common functional simulation models. For complex models, it is necessary to establish simulation models for the basic structural components of the multi-way valve separately, and then combine and superimpose the simulation models of each component to obtain the overall model of the multi-way valve. This leads to a complex modeling method process, too many parameters required for modeling, a large workload, and a large resource consumption for subsequent application calculations. While ensuring the accuracy of the simulation model, the model simulation efficiency and real-time performance cannot be taken into account. Summary of the invention

[0004] The purpose of the present invention is to overcome the deficiencies in the prior art and provide a simulation modeling method, device and equipment for a multi-way valve to solve the technical problems that the existing multi-way valve modeling process is complex, resource consumption is large, and simulation accuracy, simulation efficiency and real-time performance cannot be taken into account.

[0005] To achieve the above object, the present invention is implemented by adopting the following technical solutions: In a first aspect, the present invention provides a simulation modeling method for a multi-way valve, comprising: Conduct pressure and flow characteristic tests on the target multi-way valve to obtain a test data set; Preprocessing all test data in the test data set to obtain a sample data set; Construct a multi-way valve simulation model, including a multi-way valve flow model and a multi-way valve pressure model; The multi-way valve flow model and the multi-way valve pressure model are trained respectively through the sample data set to obtain a final multi-way valve simulation model.

[0006] Optionally, the pressure-flow characteristic test on the target multi-way valve includes: testing on a multi-way valve test bench and / or testing on a multi-way valve simulation test platform.

[0007] Optionally, the test data includes: pilot pressure , valve inlet flow rate , pressure before the valve , flow rate at the working port of the valve and pressure at the working port of the valve .

[0008] Optionally, the preprocessing includes: performing time alignment processing, normalization processing, outlier processing, missing value processing, and filtering processing on the test data.

[0009] Optionally, both the multi-way valve flow model and the multi-way valve pressure model are network models based on the encoder-decoder architecture of a recurrent neural network, and an attention mechanism layer is built between the encoder layer and the decoder layer; the basic units of both the encoder and the decoder are LSTM neural units. The encoder is used to convert the input sequence into a hidden state vector, the decoder is used to convert the hidden state vector into a target sequence, and the attention mechanism layer combines the encoder output at the current time step and the decoder output at the previous time step, evaluates their alignment degree using weight coefficients, and obtains the importance degree of the encoder information.

[0010] Optionally, the training of both the multi-way valve flow model and the multi-way valve pressure model includes: Using Glorot to initialize the hyperparameters, and determining the distribution range of the randomly initialized parameters according to the number of inputs and outputs of each layer; Using Bayesian optimization to select hyperparameters: learning rate, number of neurons in each layer; Using mini-batch stochastic descent and momentum optimization for model optimization; Introducing dropout regularization in each LSTM neural unit to prevent overfitting during training.

[0011] Optionally, the multi-way valve flow model takes the pilot pressure , the valve inlet flow rate as inputs, and takes the flow rate at the working port of the valve as the output; The multi-way valve pressure model takes the pilot pressure , the flow rate at the working port of the valve and the pressure at the working port of the valve as inputs, and takes the pressure before the valve as the output.

[0012] Optionally, the training of both the multi-way valve flow model and the multi-way valve pressure model further includes: The sample data set is divided into a training set, a validation set, and a test set in proportion by means of random sampling. The training set is used for model training, the validation set is used to cross-validate the accuracy of the model during training and iteratively optimize the model; the test set is used to finally test the generalization ability of the model.

[0013] In a second aspect, the present invention provides a simulation modeling device for a multi-way valve, including: A data acquisition module configured to perform a pressure-flow characteristic test on a target multi-way valve to obtain a test data set; A data processing module configured to preprocess all test data in the test data set to obtain a sample data set; A model construction module configured to construct a multi-way valve simulation model, including a multi-way valve flow model and a multi-way valve pressure model; A model training module configured to train the multi-way valve flow model and the multi-way valve pressure model respectively through the sample data set to obtain a final multi-way valve simulation model.

[0014] In a third aspect, the present invention provides an electronic device, including a processor and a storage medium; The storage medium is used to store instructions; The processor is used to operate according to the instructions to execute the steps according to the above method.

[0015] Compared with the prior art, the beneficial effects achieved by the present invention: The present invention provides a simulation modeling method, device and equipment for a multi-way valve. Compared with the prior art, firstly, compared with the existing multi-way valve modeling method based on the combination and superposition of basic components, the modeling method of the present invention effectively reduces the mapping work of the complex internal structure of the multi-way valve before modeling, greatly reducing the workload and cost before modeling, and due to the diversification of the training model encapsulation method, the simulation efficiency and real-time performance of the training model are greatly improved; secondly, compared with the current situation of the simulation accuracy not being high enough in the model training method for the whole multi-way valve, the present invention proposes to establish a flow training model and a pressure training model for the flow and pressure of the multi-way valve respectively, where the output of the flow training model is used as the input of the pressure training model during model application, which not only ensures the model application efficiency and real-time performance, but also ensures the accuracy of the training model. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 is a schematic flow chart of the simulation modeling method for a multi-way valve provided by an embodiment of the present invention; Figure 2 is a schematic diagram of the working principle of a multi-way valve provided by an embodiment of the present invention; Figure 3It is a working schematic diagram of the flow model of the multi-way valve provided by an embodiment of the present invention; Figure 4 It is a working schematic diagram of the pressure model of the multi-way valve provided by an embodiment of the present invention; Figure 5 It is a working schematic diagram of the simulation model of the multi-way valve provided by an embodiment of the present invention. Specific embodiments

[0017] The present invention will be further described below with reference to the accompanying drawings. The following embodiments are only used to more clearly illustrate the technical solutions of the present invention and cannot be used to limit the protection scope of the present invention.

[0018] Embodiment 1: As Figure 1 shown, an embodiment of the present invention provides a simulation modeling method for a multi-way valve, including the following steps: Step S1: Conduct a pressure-flow characteristic test on the target multi-way valve to obtain a test data set.

[0019] Specifically, in this embodiment, conducting a pressure-flow characteristic test on the target multi-way valve includes: conducting tests on a multi-way valve test bench and / or on a multi-way valve simulation test platform.

[0020] In the case where a large number of experiments cannot be carried out, by combining tests on the test bench and on the simulation test platform, the number of samples is expanded to provide sufficient sample data for subsequent model training.

[0021] Specifically, the test requirements are as follows: 1) Conduct a pressure-flow characteristic test on each stage valve of the multi-way valve. The test variables include the pressure at the working port of the valve, the inlet flow rate of the valve, and the pilot pressure. Batch tests are carried out according to the gradient setting, and the gradient interval should be as small as possible and the test data should be as many as possible.

[0022] 2) The collected test data includes: pilot pressure , inlet flow rate of the valve , pressure in front of the valve , flow rate at the working port of the valve and the pressure at the working port of the valve , and record and store the test data.

[0023] Step S2: Preprocess all the test data in the test data set to obtain a sample data set.

[0024] Preprocessing refers to performing a series of operations on the input data before inputting it so that it is more suitable for training by a neural network. Its main purposes include reducing the noise and variations of the data, reducing the data dimension, and ensuring that different features are on the same scale, so that the model can better learn the features and improve the performance of the model.

[0025] Specifically in this embodiment, the preprocessing includes: performing time alignment processing, normalization processing, outlier processing, missing value processing, and filtering processing on the test data.

[0026] The test data is converted into a time series with a fixed time step through time alignment processing. Inconsistent or incomplete data is deleted through outlier processing to remove invalid data. Missing values are filled using the mean, median, or other methods, or specialized techniques (such as interpolation) are used to process missing values. The data is filtered using a filter to remove high-frequency noise in the test data.

[0027] Step S3: Construct a multi-way valve simulation model, including a multi-way valve flow model and a multi-way valve pressure model.

[0028] During the operation of the hydraulic system, the pilot signal controls the opening of the multi-way valve port, thereby adjusting the output flow of the working port and acting on the hydraulic actuator for movement. The external load acts on the working port of the multi-way valve in the form of load pressure through the hydraulic actuator, and the load pressure is superimposed on the pressure difference of the multi-way valve and acts in front of the valve. As Figure 2 shown, according to the causal relationship, the training objectives of the multi-way valve are set as the output flow of the working port of the multi-way valve and the pressure in front of the multi-way valve. The training of the multi-way valve model is carried out separately according to two routes of flow and pressure. Therefore, two network models, namely a multi-way valve flow model and a multi-way valve pressure model, are constructed.

[0029] Specifically in this embodiment, both the multi-way valve flow model and the multi-way valve pressure model are network models based on the encoder-decoder architecture of the recurrent neural network, and an attention mechanism layer is built between the encoder layer and the decoder layer; its input is a sequence, and the output is also a sequence. The encoder converts a variable-length signal sequence into a fixed-length encoded encoder vector, and the decoder converts this fixed-length vector into a variable-length target signal sequence.

[0030] Encoder part: The basic unit of the encoder is the LSTM neuron unit. The multi-way valve flow and pressure control has typical time series characteristics. The output of the traditional fully connected network only depends on the input features at the current moment of pressure, while the LSTM can use the neuron state at the previous moment as the input of the neuron at the current moment, enabling the state variable information at an earlier moment to also be used for calculating the pressure and flow characteristics at subsequent moments, which has great advantages in dealing with flow sequence problems.

[0031] Decoder part: The basic unit of the decoder is also the LSTM neuron unit. Predicting the multi-way valve pressure and flow sequence is obtained step by step as the decoder decodes. Each decoding uses the memory information output by the encoder and the input at this moment of the output of the previous time series to jointly perform decoding, thereby obtaining the target output.

[0032] Attention mechanism part: The attention mechanism layer receives all the outputs from the encoder and all the hidden states from the decoder at the previous time series, and uses weight coefficients to evaluate their alignment degree, so that the importance of each encoder information can be obtained, and its information is multiplied by the corresponding weight coefficient and then output. Finally, all scores pass through the softmax layer to obtain the final weights of each encoder output.

[0033] Step S4: Train the multi-way valve flow model and the multi-way valve pressure model respectively through the sample data set to obtain the final multi-way valve simulation model.

[0034] Before training, the sample data set is divided into a training set, a validation set and a test set in proportion (6:2:2) by means of random sampling. The training set is used for model training, the validation set is used to cross-validate the accuracy of the model during training, and the model is iteratively optimized; the test set is used to finally test the generalization ability of the model.

[0035] During training, Glorot is used to initialize the hyperparameters, and the distribution range of randomly initialized parameters is determined according to the number of inputs and outputs of each layer; since the model structure complexity is not high, the tanh activation function is selected. Although the training cost is slightly higher than RELU, the smooth non-linear accuracy is higher.

[0036] Use Bayesian optimization to select hyperparameters: learning rate, number of neurons in each layer.

[0037] For example, in the single working port test of the multi-way valve, a total of 10 valve working port pressures and 20 pilot pressures (10 square wave signals, 5 ramp signals and 5 random signals) are adopted, a total of 200 test conditions, and each test condition is tested 3 times with an average test time of 20s. The training parameters set for the constructed model are: batch size 64, maximum number of iterations 500, loss function selected as the MSE mean square error function, optimizer Adam, dropout probability 0.1, initial learning rate 0.005.

[0038] Use mini-batch stochastic descent and momentum optimization to optimize the model.

[0039] Gradient descent method is one of the most common methods to update model weights. Using mini-batch gradient descent method, a small batch can be randomly selected from the samples for training each time instead of a group. This can avoid falling into local optima and converge to the global optimum faster. Momentum optimization introduces a momentum term on the basis of the gradient descent method to accelerate the convergence of the gradient descent method in the relevant direction and suppress oscillations. The momentum term is based on the exponentially weighted average of the previous gradients, making the parameter updates smoother and more stable.

[0040] Introduce dropout regularization in each LSTM neuron to prevent overfitting during training.

[0041] As Figure 3 shown, the multi-way valve flow model takes the pilot pressure , the valve inlet flow rate as inputs, and the valve working port flow rate as the output.

[0042] As Figure 4 shown, the multi-way valve pressure model takes the pilot pressure , the valve working port flow rate and the valve working port pressure as inputs, and the valve pre-pressure as the output.

[0043] As Figure 5 shown, after the multi-way valve flow model and the multi-way valve pressure model are trained, the final multi-way valve simulation model is constructed. In the final multi-way valve simulation model, the valve working port flow rate output by the multi-way valve flow model is used as the input of the multi-way valve pressure model, and the multi-way valve pressure model no longer obtains the valve working port flow rate from the outside. The inputs of the final multi-way valve simulation model are: pilot pressure , valve inlet flow rate , valve working port pressure , and the output is the valve pre-pressure .

[0044] Accuracy verification of the multi-way valve simulation model: Input the pilot pressure and the valve inlet flow rate in the test set part into the multi-way valve flow model for prediction; input the pilot pressure , the valve working port flow rate and the valve working port pressure in the test set part into the multi-way valve pressure model for prediction. Then compare the output of the training model with the corresponding output of the real test set. The test results show that the model accuracy is as high as 91.5%.

[0045] Model saving and encapsulation: After building and saving the multi-way valve simulation model, encapsulate and convert this model into an FMU or ONNX model. In this way, this model can be used on other simulation platforms. At the same time, it can also be encapsulated and converted into an S-Function model to speed up the running speed of the model.

[0046] In summary, a simulation modeling method for a multi-way valve provided in this embodiment conducts pressure-flow characteristic tests on a multi-way valve test bench and / or a multi-way valve simulation test platform, and obtains the working state variables and pressure-flow information of the multi-way valve through the tests; preprocesses the data to obtain a training set, a validation set, and a test set; designs and builds the neural network model structure of the training model; trains the multi-way valve flow model and the multi-way valve pressure model respectively, and verifies the model accuracy; combines the two models to obtain a multi-way valve simulation model, and saves and packages it. Aiming at the problem of poor overall training accuracy of the multi-way valve, the simulation modeling method for training the multi-way valve model is optimized, which has the advantages of higher accuracy and faster calculation speed.

[0047] Embodiment 2:

[0048] An embodiment of the present invention provides a simulation modeling device for a multi-way valve, including: A data acquisition module, configured to conduct pressure-flow characteristic tests on a target multi-way valve to obtain a test data set; A data processing module, configured to preprocess all the test data in the test data set to obtain a sample data set; A model construction module, configured to construct a multi-way valve simulation model, including a multi-way valve flow model and a multi-way valve pressure model; A model training module, configured to train the multi-way valve flow model and the multi-way valve pressure model respectively through the sample data set to obtain a final multi-way valve simulation model.

[0049] Embodiment 3: Based on the simulation modeling method for a multi-way valve provided in Embodiment 1, an embodiment of the present invention provides an electronic device, including a processor and a storage medium; The storage medium is used to store instructions; The processor is used to operate according to the instructions to execute the steps of the above method.

[0050] Those skilled in the art should understand that the embodiments of the present invention can be provided as a method, a system, or a computer program product. Therefore, the present invention can adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present invention can adopt the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0051] The present invention is described with reference to the flowcharts and / or block diagrams of methods, apparatuses (systems), and computer program products according to embodiments of the present invention. It should be understood that each flow and / or block in the flowcharts and / or block diagrams, and combinations of flows and / or blocks in the flowcharts and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to the processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing devices to produce a machine, such that the instructions executed by the processor of the computer or other programmable data processing devices produce a means for realizing the functions specified in one flow Figure 1 one flow or multiple flows and / or blocks Figure 1 or multiple blocks.

[0052] These computer program instructions can also be stored in a computer-readable memory that can direct a computer or other programmable data processing device to work in a specific manner, such that the instructions stored in the computer-readable memory produce a manufactured article including an instruction means that realizes the functions specified in one flow Figure 1 one flow or multiple flows and / or blocks Figure 1 or multiple blocks.

[0053] These computer program instructions can also be loaded onto a computer or other programmable data processing device, such that a series of operation steps are executed on the computer or other programmable device to produce a computer-implemented process, and thus the instructions executed on the computer or other programmable device provide steps for realizing the functions specified in one flow Figure 1 one flow or multiple flows and / or blocks Figure 1 or multiple blocks.

[0054] The above is only the preferred embodiment of the present invention. It should be noted that for those of ordinary skill in the art, without departing from the technical principle of the present invention, several improvements and modifications can be made, and these improvements and modifications should also be regarded as the protection scope of the present invention.

Claims

1. A simulation modeling method for a multi-way valve, characterized in that: include: Conduct pressure and flow characteristic tests on the target multi-way valve to obtain a test data set; Preprocessing all test data in the test data set to obtain a sample data set; Construct a multi-way valve simulation model, including a multi-way valve flow model and a multi-way valve pressure model; The multi-way valve flow model and the multi-way valve pressure model are trained respectively through the sample data set to obtain a final multi-way valve simulation model.

2. The simulation modeling method of a multi-way valve according to claim 1, characterized in that: The pressure flow characteristic test on the target multi-way valve includes: testing on a multi-way valve test bench and / or testing on a multi-way valve simulation test platform.

3. The simulation modeling method of a multi-way valve according to claim 1, characterized in that: The test data include: pilot pressure , Valve inlet flow , Valve front pressure , Valve working port flow And the valve working port pressure .

4. The simulation modeling method of a multi-way valve according to claim 1, characterized in that: The preprocessing includes: performing time alignment processing, normalization processing, outlier processing, missing value processing and filtering processing on the test data.

5. The simulation modeling method of a multi-way valve according to claim 1, characterized in that: The multi-way valve flow model and the multi-way valve pressure model are both network models of the encoder-decoder architecture based on a recurrent neural network, and an attention mechanism layer is built between the encoder layer and the decoder layer; the basic units of the encoder and the decoder are both LSTM neural units, the encoder is used to convert the input sequence into a hidden state vector, and the decoder is used to convert the hidden state vector into a target sequence, and the attention mechanism layer combines the encoder output of the current time step and the decoder output of the previous time step, and uses the weight coefficient to evaluate the degree of alignment to obtain the importance of the encoder information.

6. The simulation modeling method of a multi-way valve according to claim 5, characterized in that: The training of the multi-way valve flow model and the multi-way valve pressure model both include: Use Glorot to initialize hyperparameters, and determine the distribution range of random initialization parameters based on the number of inputs and outputs of each layer; Use Bayesian optimization to select hyperparameters: learning rate, number of neurons in each layer; Model optimization using mini-batch stochastic descent and momentum optimization; Dropout regularization is introduced in each LSTM neural unit to prevent overfitting in training.

7. The simulation modeling method of a multi-way valve according to claim 2, characterized in that: The multi-way valve flow model is based on the pilot pressure , the valve inlet flow As input, take the valve working port flow As output; The multi-way valve pressure model is based on the pilot pressure , the valve working port flow And the valve working port pressure As input, take the valve upstream pressure as output.

8. The simulation modeling method of a multi-way valve according to claim 1, characterized in that: The training of the multi-way valve flow model and the multi-way valve pressure model also includes: The sample data set is divided into a training set, a validation set and a test set in proportion by random sampling. The training set is used for model training, the validation set is used to cross-validate the accuracy of the model during the training process and iteratively optimize the model; the test set is used to finally test the generalization ability of the model.

9. A simulation modeling device for a multi-way valve, characterized in that: include: A data acquisition module is configured to perform a pressure and flow characteristic test on a target multi-way valve to obtain a test data set; A data processing module is configured to preprocess all test data in the test data set to obtain a sample data set; A model building module is configured to build a multi-way valve simulation model, including a multi-way valve flow model and a multi-way valve pressure model; The model training module is configured to train the multi-way valve flow model and the multi-way valve pressure model respectively through the sample data set to obtain a final multi-way valve simulation model.

10. An electronic device, characterized in that: including processor and storage medium; The storage medium is used to store instructions; The processor is configured to operate according to the instructions to execute the steps of the method according to any one of claims 1-8.

Images