Intelligent debugging system for steel pipe hydrostatic pressure valve

Abstract

The invention relates to the technical field of hydrostatic pressure valve debugging, and discloses a steel pipe hydrostatic pressure valve intelligent debugging system, which comprises a valve opening and closing test module for testing to obtain the opening and closing duration of a steel pipe hydrostatic pressure valve; the valve hydrostatic pressure testing module is used for testing to obtain the pressure stabilization coefficient of the steel pipe hydrostatic pressure valve; the valve dynamic adjustment test module is used for predicting the valve opening through a valve opening prediction model according to the first pipeline parameter; the feature matrix generation module is used for constructing a feature matrix according to the second pipeline parameters; the fault prediction module predicts a fault point distance through a fault prediction model according to the feature matrix; according to the method, the nonlinear mapping relation between the first pipeline parameter and the valve opening is established through the valve opening prediction model, so that the pipeline water pressure is safely and quickly adjusted, and the nonlinear mapping relation between the second pipeline parameter and the fault point distance is established through the fault prediction model, so that the fault point prediction function is realized.

Description

Technical Field

[0001] The present invention relates to the technical field of hydrostatic valve commissioning, and more specifically, it relates to an intelligent commissioning system for steel pipe hydrostatic valves. Background Technique

[0002] Hydrostatic pressure refers to the pressure generated by the height of a liquid column in a static state. Excessive hydrostatic pressure may cause leakage or rupture at pipe joints, sealing components, etc. A steel pipe hydrostatic valve is a valve device specifically used to control and regulate the hydrostatic pressure in a pipeline system. Its main functions include: pressure control, pressure maintenance, and flow regulation, to ensure the safety and stability of the pipeline under the action of hydrostatic pressure.

[0003] The existing commissioning methods for steel pipe hydrostatic valves mainly include: valve opening and closing test, hydrostatic test, and dynamic adjustment test; the valve opening and closing test refers to testing the opening and closing actions of the valve and recording the completion degree of the valve's opening and closing actions; the hydrostatic test refers to applying hydrostatic pressure to the pipeline and monitoring the pressures on both sides of the valve in real time to determine whether there is a leakage in the pipeline; the dynamic adjustment test refers to adjusting the pipeline water pressure to the target pressure by controlling the opening of the valve and recording the adjustment duration to evaluate the response speed and adjustment accuracy of the valve.

[0004] However, the above solutions mainly rely on manual operation, with a complex and time-consuming process. With the development of industrial automation and intelligence, the existing technology can automatically complete the valve opening and closing test, hydrostatic test, and dynamic adjustment test through electronic instructions, thereby improving the commissioning efficiency. However, in the commissioning process and subsequent use process of the hydrostatic valve, the existing technology lacks the ability to predict faults, making it difficult to detect fault points in a timely manner, resulting in out-of-control pipeline pressure and increased maintenance costs. Summary of the Invention

[0005] The present invention provides an intelligent commissioning system for steel pipe hydrostatic valves to solve the technical problems in the above background technique.

[0006] The present invention provides an intelligent commissioning system for steel pipe hydrostatic valves, including:

[0007] A valve opening and closing test module, which is used to test and obtain the opening and closing duration of the steel pipe hydrostatic valve;

[0008] First, send an opening instruction to the steel pipe hydrostatic valve in the closed state through an electronic controller until the water pressures on both sides of the valve are the same. Then, send a closing instruction to the steel pipe hydrostatic valve in the open state through the electronic controller until the water pressure downstream of the valve is 0, and record the above operation duration as the opening and closing duration;

[0009] A valve hydrostatic test module, which is used to test and obtain the pressure stabilization coefficient of the steel pipe hydrostatic valve;

[0010] Gradually increase the water pressure on the steel pipe hydrostatic valve in the closed state until the water pressure upstream of the valve reaches the first target water pressure. Record the duration of the above operation, and record the water pressure fluctuation difference upstream of the valve and the seepage flow rate downstream of the valve within the first preset time period, and calculate the obtained pressure stabilization coefficient;

[0011] The valve dynamic adjustment test module is used to collect the first pipeline parameters and input them into the valve opening prediction model. The output value represents the valve opening when the pipeline water pressure reaches the second target water pressure. Adjust the valve opening according to the value output by the valve opening prediction model, and record the duration when the pipeline water pressure reaches the second target water pressure;

[0012] The first pipeline parameters include: the current pipeline water pressure, the second target water pressure, the pipeline length, the pipeline cross-sectional area, the density of the liquid in the pipeline, and the flow rate of the liquid in the pipeline, where the second target water pressure is a user-defined parameter;

[0013] The feature matrix generation module is used to collect the second pipeline parameters at M points in the pipeline at a preset time interval within the second preset time period and construct them into a feature matrix;

[0014] The feature matrix includes M rows and N columns. The element value in the m-th row and n-th column represents the second pipeline parameter collected at the m-th point in the pipeline at the n-th time point, where 1 ≤ m ≤ M, 1 ≤ n ≤ N, and N is equal to the second preset time period divided by the preset time interval. Both the second preset time period and the preset time interval are user-defined parameters;

[0015] The second pipeline parameters include: pipeline water pressure, liquid flow rate in the pipeline, and temperature;

[0016] The fault prediction module is used to input the feature matrix into the fault prediction model, and the output value represents the Euclidean distance between the fault point and the first point in the pipeline.

[0017] Further, when performing the valve opening and closing test, when the water pressure downstream of the valve is 0, record the seepage flow rate downstream of the valve through the flow sensor. If the seepage flow rate is greater than or equal to the seepage flow rate threshold or the opening and closing duration is greater than or equal to the opening and closing duration threshold, it indicates that the closing of the steel pipe hydrostatic valve is abnormal, where both the seepage flow rate threshold and the opening and closing duration threshold are user-defined parameters.

[0018] Further, the pressure stabilization coefficient K stable The calculation formula is as follows:

[0019]

[0020] where P target represents the first target water pressure, ΔP represents the water pressure fluctuation difference upstream of the valve within the first preset time period, ΔP is equal to the maximum water pressure upstream of the valve minus the minimum water pressure within the first preset time period, Pflow Indicates the seepage flow downstream of the valve within the first preset time period, T target Indicates the duration when the water pressure upstream of the valve reaches the first target water pressure. u1 and u2 respectively represent the first weight coefficient and the second weight coefficient, and are custom parameters with a sum value equal to 1. Among them, the first target water pressure and the first preset time period are both custom parameters. sigmoid represents the sigmoid function, which is used to control the value range of the voltage stabilization coefficient between 0 and 1.

[0021] Furthermore, it is judged that when the voltage stabilization coefficient is greater than or equal to the first preset threshold, it indicates that the hydrostatic test result of the steel pipe hydrostatic valve is excellent. It is judged that when the voltage stabilization coefficient is greater than or equal to the second preset threshold and less than the first preset threshold, it indicates that the hydrostatic test result of the steel pipe hydrostatic valve is qualified. It is judged that when the voltage stabilization coefficient is less than the second preset threshold, it indicates that the hydrostatic test result of the steel pipe hydrostatic valve is unqualified. Among them, the first preset threshold and the second preset threshold are both custom parameters.

[0022] Furthermore, the valve opening prediction model includes: a first feature extraction layer, a second feature extraction layer, a feature fusion layer, and a first classifier;

[0023] The first feature extraction layer inputs the first pipeline parameter and outputs the first feature vector;

[0024] The second feature extraction layer inputs the first pipeline parameter and outputs the second feature vector;

[0025] The feature fusion layer is used to splice the first feature vector and the second feature vector to obtain the third feature vector;

[0026] The third feature vector is input into the first classifier, and the classification space of the first classifier represents the valve opening when the pipeline water pressure reaches the second target water pressure;

[0027] The calculation formula of the valve opening prediction model includes:

[0028] Feature1 = Swish(X × W1 + b1);

[0029] Feature2 = Swish(W2 × softmax(X T × W3 + b2));

[0030] Where Feature1 and Feature2 represent the first feature vector and the second feature vector respectively, X represents the first pipeline parameter input by the valve opening prediction model, W1, W2, and W3 represent the first weight parameter, the second weight parameter, and the third weight parameter respectively, b1 and b2 represent the first bias parameter and the second bias parameter respectively, T represents the transpose operation, Swish represents the Swish activation function, and softmax represents the softmax activation function.

[0031] Further, obtaining the sample labels of the training samples for training the valve opening prediction model includes the following steps:

[0032] Step S201, select A identical steel pipe hydrostatic pressure valves and collect the first pipeline parameters of the corresponding pipelines;

[0033] Where A is a custom parameter;

[0034] Step S202, set different valve openings for the A steel pipe hydrostatic pressure valves respectively;

[0035] The valve opening of each steel pipe hydrostatic pressure valve is a custom parameter;

[0036] Step S203, record the duration when the pipeline water pressure corresponding to the A steel pipe hydrostatic pressure valves reaches the second target water pressure and the maximum valve water pressure respectively, and calculate the comprehensive scores of the A steel pipe hydrostatic pressure valves;

[0037] The comprehensive score score of the a-th steel pipe hydrostatic pressure valve a is calculated as follows:

[0038]

[0039] Where 1 ≤ a ≤ A, time a and represent the duration when the pipeline water pressure corresponding to the a-th steel pipe hydrostatic pressure valve reaches the second target water pressure and the maximum valve water pressure respectively, represents the valve water pressure safety value of the a-th steel pipe hydrostatic pressure valve, u3 and u4 represent the third weight coefficient and the fourth weight coefficient respectively, and are custom parameters with a total value equal to 1, and max represents the maximum value function;

[0040] Step S204, take the valve opening corresponding to the maximum value of the comprehensive score as the sample label of a training sample, and take the first pipeline parameter corresponding to the maximum value of the comprehensive score as the sample data of a training sample;

[0041] Step S205, repeat steps S201 to S204 until B training samples are obtained;

[0042] Where B is a custom parameter.

[0043] Further, if the current pipeline water pressure is greater than the second target water pressure and the duration for which the pipeline water pressure reaches the second target water pressure is greater than or equal to the preset duration threshold, then the value output by the valve opening prediction model is reduced by 10% as the sample label of a training sample, and the corresponding first pipeline parameter is used as the sample data of a training sample; if the current pipeline water pressure is less than the second target water pressure and the duration for which the pipeline water pressure reaches the second target water pressure is greater than or equal to the preset duration threshold, then the value output by the valve opening prediction model is increased by 10% as the sample label of a training sample, and the corresponding first pipeline parameter is used as the sample data of a training sample, where the preset duration threshold is a custom parameter.

[0044] Further, the fault prediction model includes N hidden layers and 1 second classifier, and each hidden layer includes a first hidden unit and a second hidden unit;

[0045] The first hidden unit of the nth hidden layer inputs the element values from the 1st row to the Mth row of the nth column of the feature matrix and outputs a first updated vector;

[0046] The second hidden unit of the nth hidden layer inputs the first updated vector output by the first hidden unit of the nth hidden layer and outputs a second updated vector;

[0047] The second updated vector output by the second hidden unit of the Nth hidden layer is input into the second classifier, and the classification space of the second classifier represents the distance between the fault point and the 1st point in the pipeline.

[0048] Further, the first hidden unit is constructed based on the transformer model, and the second hidden unit is constructed based on the gated neural network.

[0049] Further, obtaining the sample label of the training sample for training the fault prediction model includes the following steps:

[0050] Step S301, construct a pipeline digital model through a simulation platform;

[0051] Step S302, randomly generate a fault point in the pipeline digital model and record the Euclidean distance between the fault point and the 1st point in the pipeline as the sample label of a training sample;

[0052] Step S303, within the second preset time period, collect the second pipeline parameters of M points in the pipeline at preset time intervals and construct them into a feature matrix as the sample data of a training sample;

[0053] Step S304, repeat steps S301 to S303 until C training samples are obtained;

[0054] Where C is a custom parameter.

[0055] The beneficial effects of the present invention are as follows: The present invention provides an automated debugging system for the hydrostatic pressure valve of steel pipes, and establishes a non-linear mapping relationship between the first pipeline parameter and the valve opening through the valve opening prediction model, so as to realize the safe and rapid adjustment of the pipeline water pressure. A non-linear mapping relationship between the second pipeline parameter and the distance of the fault point is established through the fault prediction model, so as to realize the function of fault point prediction, and the fault point can be detected in time, reducing the maintenance cost. BRIEF DESCRIPTION OF THE DRAWINGS

[0056] Figure 1 is a schematic diagram of an intelligent debugging system for the hydrostatic pressure valve of steel pipes according to the present invention;

[0057] Figure 2 is a flowchart of obtaining the sample label of the training sample for training the valve opening prediction model according to the present invention;

[0058] Figure 3 is a flowchart of obtaining the sample label of the training sample for training the fault prediction model according to the present invention.

[0059] In the figure: valve opening and closing test module 101, valve hydrostatic pressure test module 102, valve dynamic adjustment test module 103, feature matrix generation module 104, fault prediction module 105. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0060] Now, the subject matter described herein will be discussed with reference to example embodiments. It should be understood that discussing these embodiments is only to enable those skilled in the art to better understand and thus implement the subject matter described herein. Without departing from the protection scope of the content of this specification, changes can be made to the functions and arrangements of the elements discussed. Each example can omit, substitute or add various processes or components as needed. In addition, the features described relative to some examples can also be combined in other examples.

[0061] It should be noted that, unless otherwise defined, the technical terms or scientific terms used in one or more embodiments of the present invention should have the ordinary meanings understood by those of ordinary skill in the field to which the present invention pertains. In one or more embodiments of the present invention, words such as "first", "second" and the like do not denote any order, quantity or importance, but are only used to distinguish different components. Words such as "including" or "comprising" mean that the elements or objects appearing before this word cover the elements or objects listed after this word and their equivalents, without excluding other elements or objects. Words such as "connected" or "coupled" are not limited to physical or mechanical connections, but may include electrical connections, whether direct or indirect. "Upper", "lower", "left", "right", etc. are only used to represent relative position relationships. When the absolute position of the object being described changes, the relative position relationship may also change accordingly.

[0062] As Figures 1 to 3 shown, a static hydrostatic pressure valve intelligent debugging system for steel pipes includes:

[0063] A valve opening and closing test module 101, which is used to test and obtain the opening and closing duration of the static hydrostatic pressure valve of the steel pipe;

[0064] First, an opening instruction is sent to the static hydrostatic pressure valve of the steel pipe in the closed state through an electronic controller until the water pressures on both sides of the valve are the same. Then, a closing instruction is sent to the static hydrostatic pressure valve of the steel pipe in the open state through the electronic controller until the water pressure downstream of the valve is 0, and the operation duration of the above operations is recorded as the opening and closing duration;

[0065] A valve static hydrostatic pressure test module 102, which is used to test and obtain the pressure stabilization coefficient of the static hydrostatic pressure valve of the steel pipe;

[0066] The water pressure of the static hydrostatic pressure valve of the steel pipe in the closed state is gradually increased until the water pressure upstream of the valve reaches the first target water pressure. The operation duration of the above operations is recorded, and the water pressure fluctuation difference upstream of the valve and the seepage flow rate downstream of the valve are recorded within the first preset time period, and the pressure stabilization coefficient is calculated;

[0067] A valve dynamic adjustment test module 103, which is used to collect the first pipeline parameters and input them into the valve opening prediction model. The output value represents the valve opening when the pipeline water pressure reaches the second target water pressure. The valve opening is adjusted according to the output value of the valve opening prediction model, and the duration when the pipeline water pressure reaches the second target water pressure is recorded;

[0068] The first pipeline parameters include: the current pipeline water pressure, the second target water pressure, the pipeline length, the pipeline cross-sectional area, the density of the liquid in the pipeline, and the flow rate of the liquid in the pipeline, where the second target water pressure is a user-defined parameter;

[0069] A feature matrix generation module 104, which is configured to collect second pipeline parameters at M points in the pipeline at a preset time interval within a second preset time period and construct them into a feature matrix;

[0070] The feature matrix includes M rows and N columns. The element value of the m-th row and the n-th column represents the second pipeline parameter collected at the m-th point in the pipeline at the n-th time point, where 1 ≤ m ≤ M, 1 ≤ n ≤ N, and N is equal to the second preset time period divided by the preset time interval. Both the second preset time period and the preset time interval are user-defined parameters. Preferably, the second preset time period is set to 1 minute and the preset time interval is set to 10 seconds;

[0071] The second pipeline parameters include: pipeline water pressure, liquid flow rate in the pipeline, and temperature;

[0072] A fault prediction module 105, which is configured to input the feature matrix into a fault prediction model, and the output value represents the Euclidean distance between the fault point and the first point in the pipeline.

[0073] It should be noted that the present invention provides an automated debugging system for a steel pipe hydrostatic pressure valve, and a non-linear mapping relationship between the first pipeline parameters and the valve opening is established through a valve opening prediction model, so as to achieve safe and rapid adjustment of the pipeline water pressure. A non-linear mapping relationship between the second pipeline parameters and the fault point distance is established through a fault prediction model, so as to achieve the function of fault point prediction, and the fault point can be detected in time and the maintenance cost can be reduced.

[0074] In an embodiment of the present invention, when performing a valve opening and closing test, when the water pressure downstream of the valve is 0, the leakage flow rate downstream of the valve is recorded by a flow sensor. It is judged that when the leakage flow rate is greater than or equal to the leakage flow rate threshold or the opening and closing duration is greater than or equal to the opening and closing duration threshold, it means that the closing of the steel pipe hydrostatic pressure valve is abnormal, where both the leakage flow rate threshold and the opening and closing duration threshold are user-defined parameters.

[0075] In an embodiment of the present invention, the pressure stabilization coefficient K stable The calculation formula is as follows:

[0076]

[0077] Where P target represents the first target water pressure, ΔP represents the water pressure fluctuation difference upstream of the valve within a first preset time period, and ΔP is equal to the maximum water pressure minus the minimum water pressure upstream of the valve within the first preset time period. P flow represents the leakage flow rate downstream of the valve within the first preset time period, and T targetIndicates the duration when the water pressure upstream of the valve reaches the first target water pressure. u1 and u2 respectively represent the first weight coefficient and the second weight coefficient, and are custom parameters with a sum value equal to 1. Among them, the first target water pressure and the first preset time period are both custom parameters. sigmoid represents the sigmoid function, which is used to control the value range of the voltage stabilization coefficient between 0 and 1.

[0078] Preferably, u1 is set to 0.8, u2 is set to 0.2, the first target water pressure is set to 1.5 times the normal working water pressure of the steel pipe hydrostatic pressure valve, and the first preset time period is set to 10 minutes.

[0079] In an embodiment of the present invention, it is judged that when the voltage stabilization coefficient is greater than or equal to the first preset threshold, it indicates that the hydrostatic pressure test result of the steel pipe hydrostatic pressure valve is excellent. It is judged that when the voltage stabilization coefficient is greater than or equal to the second preset threshold and less than the first preset threshold, it indicates that the hydrostatic pressure test result of the steel pipe hydrostatic pressure valve is qualified. It is judged that when the voltage stabilization coefficient is less than the second preset threshold, it indicates that the hydrostatic pressure test result of the steel pipe hydrostatic pressure valve is unqualified. Among them, the first preset threshold and the second preset threshold are both custom parameters. Preferably, the first preset threshold is set to 0.95 and the second preset threshold is set to 0.8.

[0080] It should be noted that when the hydrostatic pressure test result of the steel pipe hydrostatic pressure valve is unqualified, it indicates that the seal of the steel pipe hydrostatic pressure valve may be aged, damaged or improperly installed. And the duration when the water pressure upstream of the valve reaches the first target water pressure is too long or the water pressure fluctuation difference upstream of the valve within the first preset time period is large, which may also indicate that the valve stem or valve core is stuck due to impurity accumulation.

[0081] In an embodiment of the present invention, the valve opening prediction model includes: a first feature extraction layer, a second feature extraction layer, a feature fusion layer and a first classifier;

[0082] The first feature extraction layer inputs the first pipeline parameter and outputs the first feature vector;

[0083] The second feature extraction layer inputs the first pipeline parameter and outputs the second feature vector;

[0084] The feature fusion layer is used to splice the first feature vector and the second feature vector to obtain the third feature vector;

[0085] The third feature vector is input into the first classifier, and the classification space of the first classifier represents the valve opening when the pipeline water pressure reaches the second target water pressure;

[0086] The calculation formula of the valve opening prediction model includes:

[0087] Feature1 = Swish(X × W1 + b1);

[0088] Feature2 = Swish(W2 × softmax(X T × W3 + b2));

[0089] Where Feature1 and Feature2 represent the first feature vector and the second feature vector respectively, X represents the first pipeline parameter input by the valve opening prediction model, W1, W2, and W3 represent the first weight parameter, the second weight parameter, and the third weight parameter respectively, b1 and b2 represent the first bias parameter and the second bias parameter respectively, T represents the transpose operation, Swish represents the Swish activation function, and softmax represents the softmax activation function.

[0090] It should be noted that the weight parameters and bias parameters in the valve opening prediction model are all learnable parameters. Before the forward calculation of the valve opening prediction model, the first pipeline parameter is normalized to eliminate the influence of dimensions. That is, the first pipeline parameter is a vector of size 1×6, and each dimension value corresponds to the current pipeline water pressure, the second target water pressure, the pipeline length, the pipeline cross-sectional area, the liquid density in the pipeline, and the liquid flow rate in the pipeline after normalization. The first weight parameter can be designed as a matrix of size 6×16, then the size of the first feature vector is 1×16. The third weight parameter can be designed as a vector of size 1×6. The product of the transpose of the first pipeline parameter and the third weight parameter can obtain a matrix of size 6×6. The second weight parameter can be designed as a vector of size 1×6, then the size of the second feature vector is 1×6. Then the size of the third feature vector is 1×(16 + 6) = 1×22. The activation function of the first classifier is the softmax activation function, which maps the third feature vector to the valve opening, and this will not be elaborated here.

[0091] In an embodiment of the present invention, as Figure 2 shown, obtaining the sample label of the training sample for training the valve opening prediction model includes the following steps:

[0092] Step S201, select A identical steel pipe hydrostatic pressure valves and collect the first pipeline parameters of the corresponding pipelines;

[0093] Where A is a custom parameter. Preferably, A is set to 5;

[0094] Step S202, set different valve openings for the A steel pipe hydrostatic pressure valves respectively;

[0095] The valve opening of each steel pipe hydrostatic pressure valve is a custom parameter;

[0096] Step S203: Record the duration when the pipeline water pressure corresponding to each of the A hydrostatic valves of steel pipes reaches the second target water pressure and the maximum valve water pressure respectively, and calculate the comprehensive score of the A hydrostatic valves of steel pipes;

[0097] The comprehensive score score of the a-th hydrostatic valve of steel pipe a is calculated as follows:

[0098]

[0099] where 1 ≤ a ≤ A, time a and respectively represent the duration when the pipeline water pressure corresponding to the a-th hydrostatic valve of steel pipe reaches the second target water pressure and the maximum valve water pressure, represents the valve water pressure safety value of the a-th hydrostatic valve of steel pipe, u3 and u4 respectively represent the third weight coefficient and the fourth weight coefficient, and are custom parameters with the sum value equal to 1, and max represents the maximum value function;

[0100] Preferably, u3 is set to 0.4 and u4 is set to 0.6, is set to 1.3 times the normal working water pressure of the hydrostatic valve of steel pipe;

[0101] Step S204: Take the valve opening corresponding to the maximum value of the comprehensive score as the sample label of a training sample, and take the first pipeline parameter corresponding to the maximum value of the comprehensive score as the sample data of a training sample;

[0102] Step S205: Repeat Step S201 to Step S204 until B training samples are obtained;

[0103] where B is a custom parameter, preferably, B is set to 1000.

[0104] It should be noted that one training sample corresponds to one sample label and one sample data. The B training samples can be divided into a training set, a validation set and a test set according to the ratio of 7:2:1, and the mean square error is specified as the loss function. The weight parameters and bias parameters of the valve opening prediction model are updated backward through the gradient descent algorithm (such as AdaGrad, RMSProp, etc.), which will not be elaborated here.

[0105] In an embodiment of the present invention, when the current pipeline water pressure is greater than the second target water pressure and the duration for which the pipeline water pressure reaches the second target water pressure is greater than or equal to the preset duration threshold, the value output by the valve opening prediction model is reduced by 10% as the sample label of a training sample, and the corresponding first pipeline parameter is used as the sample data of a training sample; when the current pipeline water pressure is less than the second target water pressure and the duration for which the pipeline water pressure reaches the second target water pressure is greater than or equal to the preset duration threshold, the value output by the valve opening prediction model is increased by 10% as the sample label of a training sample, and the corresponding first pipeline parameter is used as the sample data of a training sample, where the preset duration threshold is a custom parameter.

[0106] It should be noted that this automation provided by the present invention feeds the actual adjustment result back to the valve opening prediction model, which can greatly increase the number of training samples, improve the prediction accuracy of the model, and enhance the robustness of the model.

[0107] In an embodiment of the present invention, the fault prediction model includes N hidden layers and 1 second classifier, and each hidden layer includes a first hidden unit and a second hidden unit;

[0108] The first hidden unit of the nth hidden layer inputs the element values from the 1st row to the Mth row of the nth column of the feature matrix and outputs a first updated vector;

[0109] The second hidden unit of the nth hidden layer inputs the first updated vector output by the first hidden unit of the nth hidden layer and outputs a second updated vector;

[0110] The second updated vector output by the second hidden unit of the Nth hidden layer is input into the second classifier, and the classification space of the second classifier represents the distance between the fault point and the 1st point in the pipeline.

[0111] In an embodiment of the present invention, the first hidden unit is constructed based on the Transformer model or can also be constructed based on the self-attention mechanism, and the second hidden unit is constructed based on the gated neural network (GRU) or can also be constructed based on the recurrent neural network (RNN).

[0112] In an embodiment of the present invention, as Figure 3 shown, obtaining the sample label of the training sample for training the fault prediction model includes the following steps:

[0113] Step S301, constructing a pipeline digital model through a simulation platform;

[0114] Step S302, randomly generating a fault point in the pipeline digital model and recording the Euclidean distance between the fault point and the 1st point in the pipeline as the sample label of a training sample;

[0115] Step S303, within a second preset time period, collect second pipeline parameters at M points in the pipeline at preset time intervals, and construct them into a feature matrix as the sample data of a training sample;

[0116] Step S304, repeat Step S301 to Step S303 until C training samples are obtained;

[0117] where C is a custom parameter. Preferably, B is set to 1000.

[0118] It should be noted that the simulation platform can be ANSYS Fluent (a fluid dynamics simulation modeling software) or COMSOL Multiphysics (a multi-physics simulation modeling software). Constructing a pipeline digital model through the simulation platform belongs to conventional technical means and will not be elaborated here.

[0119] The above describes the embodiments of this embodiment, but this embodiment is not limited to the above specific implementation manners. The above specific implementation manners are merely illustrative rather than restrictive. Under the inspiration of this embodiment, those of ordinary skill in the art can also make many forms, all of which fall within the protection scope of this embodiment.

Claims

1. An intelligent debugging system for hydrostatic pressure valves of steel pipes, characterized in that, Including: A valve opening and closing test module, which is used to test and obtain the opening and closing duration of the steel pipe hydrostatic pressure valve; A valve hydrostatic pressure test module, which is used to test and obtain the pressure stabilization coefficient of the steel pipe hydrostatic pressure valve; A valve dynamic regulation test module, which is used to collect the first pipeline parameters and input them into the valve opening prediction model. The output value represents the valve opening when the pipeline water pressure reaches the second target water pressure. Adjust the valve opening according to the output value of the valve opening prediction model, and record the duration when the pipeline water pressure reaches the second target water pressure; The first pipeline parameters include: the current pipeline water pressure, the second target water pressure, the pipeline length, the pipeline cross-sectional area, the liquid density in the pipeline, and the liquid flow rate in the pipeline, where the second target water pressure is a user-defined parameter; A feature matrix generation module, which is used to collect the second pipeline parameters at M points in the pipeline at preset time intervals within the second preset time period and construct them into a feature matrix; The second pipeline parameters include: pipeline water pressure, liquid flow rate in the pipeline, and temperature; A fault prediction module, which is used to input the feature matrix into the fault prediction model, and the output value represents the Euclidean distance between the fault point and the first point in the pipeline.

2. The intelligent commissioning system for a steel pipe hydrostatic pressure valve according to claim 1, wherein, First, send an opening instruction to the steel pipe hydrostatic pressure valve in the closed state through the electronic controller until the water pressures on both sides of the valve are the same. Then, send a closing instruction to the steel pipe hydrostatic pressure valve in the open state through the electronic controller until the water pressure downstream of the valve is 0, and record the above operation duration as the opening and closing duration. When performing the valve opening and closing test, when the water pressure downstream of the valve is 0, record the seepage flow rate downstream of the valve through the flow sensor. If the seepage flow rate is greater than or equal to the seepage flow rate threshold or the opening and closing duration is greater than or equal to the opening and closing duration threshold, it means that the closing of the steel pipe hydrostatic pressure valve is abnormal, where both the seepage flow rate threshold and the opening and closing duration threshold are user-defined parameters.

3. The intelligent commissioning system for the hydrostatic pressure valve of a steel pipe according to claim 1, wherein, Gradually increase the water pressure to the steel pipe hydrostatic pressure valve in the closed state until the water pressure upstream of the valve reaches the first target water pressure, record the above operation duration, and record the water pressure fluctuation difference upstream of the valve and the seepage flow rate downstream of the valve within the first preset time period, and calculate to obtain the pressure stabilization coefficient; Voltage regulation coefficient K stable The calculation formula is as follows: Where P target represents the first target water pressure, ΔP represents the water pressure fluctuation difference upstream of the valve within the first preset time period, ΔP is equal to the maximum water pressure minus the minimum water pressure upstream of the valve within the first preset time period, P flow represents the seepage flow rate downstream of the valve within the first preset time period, T target represents the duration when the water pressure upstream of the valve reaches the first target water pressure, u1 and u2 respectively represent the first weight coefficient and the second weight coefficient, and are custom parameters with a sum value equal to 1. Among them, the first target water pressure and the first preset time period are both custom parameters, and sigmoid represents the sigmoid function, which is used to control the value range of the voltage stabilization coefficient between 0 and 1.

4. An intelligent debugging system for a hydrostatic pressure valve of a steel pipe according to claim 1, characterized in that, Judge that when the pressure stabilization coefficient is greater than or equal to the first preset threshold, it means that the hydrostatic pressure test result of the steel pipe hydrostatic pressure valve is excellent. Judge that when the pressure stabilization coefficient is greater than or equal to the second preset threshold and less than the first preset threshold, it means that the hydrostatic pressure test result of the steel pipe hydrostatic pressure valve is qualified. Judge that when the pressure stabilization coefficient is less than the second preset threshold, it means that the hydrostatic pressure test result of the steel pipe hydrostatic pressure valve is unqualified, where both the first preset threshold and the second preset threshold are user-defined parameters.

5. An intelligent debugging system for a hydrostatic pressure valve of a steel pipe according to claim 1, characterized in that, The valve opening prediction model includes: a first feature extraction layer, a second feature extraction layer, a feature fusion layer, and a first classifier; The first feature extraction layer inputs the first pipeline parameters and outputs a first feature vector; The second feature extraction layer inputs the first pipeline parameters and outputs a second feature vector; The feature fusion layer is used to splice the first feature vector and the second feature vector to obtain a third feature vector; The third feature vector is input into the first classifier, and the classification space of the first classifier represents the valve opening when the pipeline water pressure reaches the second target water pressure; The calculation formula of the valve opening prediction model includes: Feature1 = Swish(X × W1 + b1); Feature2 = Swish(W2 × softmax(X T × W3 + b2)); Among them, Feature1 and Feature2 represent the first feature vector and the second feature vector respectively, X represents the first pipeline parameter input by the valve opening prediction model, W1, W2, and W3 represent the first weight parameter, the second weight parameter, and the third weight parameter respectively, b1 and b2 represent the first bias parameter and the second bias parameter respectively, T represents the transpose operation, Swish represents the Swish activation function, and softmax represents the softmax activation function.

6. The intelligent commissioning system for a static hydrostatic pressure valve of a steel pipe according to claim 1, wherein, Obtaining the sample labels of the training samples for training the valve opening prediction model includes the following steps: Step S201: Select A identical steel pipe hydrostatic pressure valves and collect the first pipeline parameters of the corresponding pipelines; Where A is a custom parameter; Step S202: Set different valve openings for the A steel pipe hydrostatic pressure valves respectively; The valve opening of each steel pipe hydrostatic pressure valve is a custom parameter; Step S203: Record the duration when the pipeline water pressure corresponding to the A steel pipe hydrostatic pressure valves reaches the second target water pressure and the maximum value of the valve water pressure respectively, and calculate the comprehensive scores of the A steel pipe hydrostatic pressure valves; The comprehensive score score of the a-th steel pipe hydrostatic valve a The calculation formula is as follows: where 1 ≤ a ≤ A, time a and respectively represent the duration when the pipeline water pressure corresponding to the a-th steel pipe hydrostatic pressure valve reaches the second target water pressure and the maximum value of the valve water pressure, represents the valve water pressure safety value of the a-th steel pipe hydrostatic pressure valve, u3 and u4 respectively represent the third weight coefficient and the fourth weight coefficient, and are custom parameters with a total value equal to 1, max represents the maximum value function; Step S204: Take the valve opening corresponding to the maximum value of the comprehensive score as the sample label of a training sample, and take the first pipeline parameter corresponding to the maximum value of the comprehensive score as the sample data of a training sample; Step S205: Repeat steps S201 to S204 until B training samples are obtained; Where B is a custom parameter.

7. The intelligent commissioning system for a steel pipe hydrostatic pressure valve according to claim 6, characterized in that, When the current pipeline water pressure is greater than the second target water pressure and the duration when the pipeline water pressure reaches the second target water pressure is greater than or equal to the preset duration threshold, reduce the value output by the valve opening prediction model by 10% as the sample label of a training sample, and use the corresponding first pipeline parameter as the sample data of a training sample; when the current pipeline water pressure is less than the second target water pressure and the duration when the pipeline water pressure reaches the second target water pressure is greater than or equal to the preset duration threshold, increase the value output by the valve opening prediction model by 10% as the sample label of a training sample, and use the corresponding first pipeline parameter as the sample data of a training sample, where the preset duration threshold is a custom parameter.

8. An intelligent commissioning system for a hydrostatic pressure valve of a steel pipe according to claim 1, characterized in that, The feature matrix includes M rows and N columns. The element value of the m-th row and the n-th column represents the second pipeline parameter collected at the m-th point in the pipeline at the n-th time point, where 1 ≤ m ≤ M, 1 ≤ n ≤ N, N is equal to the second preset time period divided by the preset time interval, and both the second preset time period and the preset time interval are custom parameters; the fault prediction model includes N hidden layers and 1 second classifier, and each hidden layer includes a first hidden unit and a second hidden unit; The first hidden unit of the n-th hidden layer inputs the element values from the 1st row to the M-th row of the n-th column of the feature matrix and outputs the first updated vector; The second hidden unit of the n-th hidden layer inputs the first updated vector output by the first hidden unit of the n-th hidden layer and outputs the second updated vector; The second update vector output by the second hidden unit of the Nth hidden layer is input into the second classifier, and the classification space of the second classifier represents the distance between the fault point and the first point in the pipeline.

9. The intelligent commissioning system for a steel pipe hydrostatic pressure valve according to claim 1, characterized in that, The first hidden unit is constructed based on the transformer model, and the second hidden unit is constructed based on the gated neural network.

10. The intelligent commissioning system for a steel pipe hydrostatic pressure valve according to claim 1, characterized in that, Obtaining the sample labels of the training samples for training the fault prediction model includes the following steps: Step S301, constructing a pipeline digital model through a simulation platform; Step S302, randomly generating a fault point in the pipeline digital model, and recording the Euclidean distance between the fault point and the first point in the pipeline as the sample label of a training sample; Step S303, within a second preset time period, collecting the second pipeline parameters of M points in the pipeline at preset time intervals, and constructing them into a feature matrix as the sample data of a training sample; Step S304, repeating Step S301 to Step S303 until C training samples are obtained; where C is a custom parameter.

Images