Quality monitoring method for metal welding process of thermal power plant, storage medium, equipment, computer program product and device

Abstract

The invention discloses a thermal power plant metal welding process quality monitoring method, a storage medium, equipment, a computer program product and a device, and the method comprises the steps: collecting key parameters in the thermal power plant metal welding process, and marking the welding quality according to the key parameters; performing noise removal and normalization processing operation on the collected key parameters, and forming a key parameter sequence according to a time sequence; constructing a thermal power plant metal welding quality prediction model, sequentially inputting the key parameter sequences into the thermal power plant metal welding quality prediction model, and training the thermal power plant metal welding quality prediction model by taking the marked welding quality as a label until a mean square error loss function is converged, completing the training of the thermal power plant metal welding quality prediction model; key parameters in the thermal power plant metal welding process are collected in real time and input into the trained thermal power plant metal welding quality prediction model, the metal welding quality is predicted, and real-time monitoring of the metal welding process is achieved.

Description

Technical Field

[0001] The present invention relates to the technical field of thermal power plant welding, and in particular to a quality monitoring method, storage medium, equipment, computer program product and device for a metal welding process in a thermal power plant. Background Art

[0002] In the equipment construction, overhaul and maintenance process of thermal power plants, metal welding is one of the key processes, and the welding quality is directly related to the operation safety and service life of the equipment. Traditional welding quality monitoring methods mainly rely on manual inspection and post-nondestructive testing methods, such as X-ray flaw detection and ultrasonic testing. Although these methods can detect defects after welding, they have the following problems:

[0003] (1) Detection hysteresis: Traditional metal welding quality inspection is usually carried out after welding is completed, which cannot achieve real-time monitoring. When problems are found, rework is often required, resulting in time and economic losses.

[0004] (2) Strong dependence on manual inspection: Manual inspection has errors and cannot fully monitor every detail of the welding process;

[0005] (3) High testing cost: The cost of later non-destructive testing is high and the cycle is long.

[0006] In order to solve the above technical problems, there is an urgent need for an intelligent monitoring system and method that can monitor the metal welding process in real time, automatically identify welding defects and provide feedback, so as to achieve automatic optimization and quality control of the welding process. Summary of the invention

[0007] In response to the problems existing in the prior art, the present invention provides a quality monitoring method, storage medium, equipment, computer program product and device for the metal welding process in a thermal power plant, which can monitor the metal welding process in real time, automatically identify the welding quality and provide feedback to achieve automatic optimization and quality control of the metal welding process.

[0008] To achieve the above technical objectives, the present invention adopts the following technical solution: a quality monitoring method for metal welding process in a thermal power plant, specifically comprising the following steps:

[0009] Step S1, collecting key parameters in the metal welding process of the thermal power plant, including: current, voltage, welding temperature, welding speed and molten pool state, and marking the welding quality according to the key parameters;

[0010] Step S2, performing noise removal and normalization processing operations on the collected key parameters to form a key parameter sequence in time sequence;

[0011] Step S3, constructing a thermal power plant metal welding quality prediction model, inputting the key parameter sequence into the thermal power plant metal welding quality prediction model in sequence, using the marked welding quality as a label, training the thermal power plant metal welding quality prediction model until the mean square error loss function converges, and completing the training of the thermal power plant metal welding quality prediction model;

[0012] Step S4: collect key parameters of the metal welding process of the thermal power plant in real time, input them into the trained thermal power plant metal welding quality prediction model, and predict the metal welding quality.

[0013] Furthermore, in step S3, the metal welding quality prediction model of a thermal power plant is composed of an LSTM layer, a self-attention layer, a fully connected layer and an SVM classifier. The LSTM layer can capture the long-term dependencies in the key parameter sequence and generate a hidden state vector; the self-attention layer captures important information from the hidden state vector and generates a feature vector of a fixed length through the fully connected layer; the SVM classifier is used to predict the metal welding quality based on the feature vector.

[0014] Furthermore, step S3 includes the following sub-steps:

[0015] Step S3.1, input the key parameter sequence into the LSTM layer of the metal welding quality prediction model of the thermal power plant, capture the long-term dependency in the key parameter sequence, and generate a hidden state vector;

[0016] Step S3.2: Input the hidden state vector into the self-attention layer to calculate the attention score, then perform softmax normalization to obtain the attention weight, perform weighted summation of the attention score and the attention weight to capture the feature vector;

[0017] Step S3.3, input the captured feature vector into the fully connected layer, and output a feature vector of fixed length;

[0018] Step S3.4, inputting the fixed-length feature vector into the SVM classifier, and predicting the metal welding quality through the polynomial kernel function;

[0019] Step S3.5, calculate the mean square error loss function of the predicted metal welding quality and the label, repeat steps S3.1-S3.4 until the mean square error loss function converges, and complete the training of the thermal power plant metal welding quality prediction model.

[0020] Furthermore, the LSTM layer in step S3.1 is composed of an input gate, a forget gate, and an output gate, wherein the input gate is used to determine the information i added to the cell state by the key parameter sequence. t =σ(U i ·r t-1 +W i ·ht +b i ), and also generate new candidate memories through the tanh function The forget gate is used to decide whether to discard the information of the key parameter sequence in the cell state. t =σ(U f ·r t-1 +W f ·h t +b f ), and updates the cell state according to the output of the forget gate and the input gate The output gate is used to determine whether to output the information in the updated cell state to the hidden state r at the next moment. t =o t ⊙tanhc t , where h t represents the key parameter sequence at time t, σ represents the sigmoid function, tanh represents the hyperbolic sine function, represents the candidate storage state, c t represents the internal state, U and W both represent weight matrices, b represents the bias, ⊙ represents the dot product of the matrix, o t represents the activation function of the output gate, o t =σ(U o ·r t-1 +W o ·h t +b o ).

[0021] Furthermore, the process of capturing the feature vector in step S3.2 is:

[0022] s=score(q a ,k j )×attention

[0023] Among them, score(q a ,k j ) represents the attention score of the hidden state vector, q a represents the a-th query vector, k j represents the jth key vector; attention represents score(q a ,k j ) is normalized by softmax to obtain the attention weight, attention = softmax(score(q i ,k j ))v j , v j Represents the j-th value vector.

[0024] Furthermore, the predicted metal welding quality is qualified or unqualified. If the predicted metal welding quality is unqualified, the key parameters of the metal welding are adjusted; if the predicted metal welding quality is still unqualified after multiple adjustments of the key parameters, the operator is reminded to stop the machine for inspection.

[0025] Furthermore, the present invention also provides a computer-readable storage medium storing a computer program, wherein the computer program enables a computer to execute the quality monitoring method for the metal welding process in a thermal power plant.

[0026] Furthermore, the present invention also provides an electronic device, comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, the quality monitoring method for the metal welding process in a thermal power plant is implemented.

[0027] Furthermore, the present invention also provides a computer program product, including a computer program, and when the computer program is executed by a processor, the quality monitoring method of the metal welding process in a thermal power plant is implemented.

[0028] Furthermore, the present invention also provides a quality monitoring device for a metal welding process in a thermal power plant, comprising: a data acquisition unit, a data processing unit, a display and feedback unit, and an alarm and control unit;

[0029] The data acquisition unit is used to collect key parameters in the metal welding process of the thermal power plant, including: current, voltage, welding temperature, welding speed and molten pool state;

[0030] The data processing unit performs noise removal and normalization processing on the collected key parameters and then inputs them into the metal welding quality prediction model of the thermal power plant to predict the metal welding quality;

[0031] The display and feedback unit is used to display the predicted metal welding quality. If the metal welding quality is unqualified, key parameters are adjusted through the alarm and control unit.

[0032] Compared with the prior art, the present invention has the following beneficial effects: the quality monitoring method of the metal welding process in a thermal power plant of the present invention realizes real-time monitoring of the metal welding quality by constructing a metal welding quality prediction model for a thermal power plant and acquiring key parameters of metal welding, thereby avoiding manual dependence; at the same time, the metal welding quality prediction model for a thermal power plant of the present invention captures the long-term dependency in the key parameter sequence through LSTM, generates a hidden state vector with rich information, and then captures the feature vector through the self-attention mechanism to capture the global dependency, and by making full use of the advantages of LSTM and the self-attention mechanism, improves the extraction of feature vectors in the metal welding process, thereby realizing accurate metal welding quality prediction. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] Figure 1 It is a flow chart of the quality monitoring method of the metal welding process in a thermal power plant of the present invention;

[0034] Figure 2 It is a schematic diagram of the metal welding quality prediction model for thermal power plants in the present invention. DETAILED DESCRIPTION

[0035] The technical solution of the present invention is further explained below in conjunction with the accompanying drawings.

[0036] like Figure 1 The present invention provides a quality monitoring method for a metal welding process in a thermal power plant, which specifically comprises the following steps:

[0037] Step S1, collecting key parameters in the metal welding process of the thermal power plant, including: current, voltage, welding temperature, welding speed and molten pool state, and marking the welding quality according to the key parameters;

[0038] Step S2, performing noise removal and normalization processing operations on the collected key parameters to form a key parameter sequence in time sequence;

[0039] Step S3, constructing a thermal power plant metal welding quality prediction model, inputting the key parameter sequence into the thermal power plant metal welding quality prediction model in sequence, using the marked welding quality as a label, training the thermal power plant metal welding quality prediction model until the mean square error loss function converges, and completing the training of the thermal power plant metal welding quality prediction model;

[0040] like Figure 2 In the present invention, the metal welding quality prediction model of a thermal power plant is composed of an LSTM layer, a self-attention layer, a fully connected layer and an SVM classifier. The LSTM layer can capture the long-term dependencies in the key parameter sequence and generate a hidden state vector; the self-attention layer captures important information from the hidden state vector, captures the global dependencies, and generates a fixed-length feature vector through the fully connected layer; the SVM classifier is used to predict the metal welding quality according to the feature vector. The present invention makes full use of the advantages of LSTM and self-attention mechanisms to improve the extraction of feature vectors in the metal welding process, thereby achieving accurate metal welding quality prediction.

[0041] Step S3 specifically includes the following sub-steps:

[0042] Step S3.1, input the key parameter sequence into the LSTM layer of the thermal power plant metal welding quality prediction model, capture the long-term dependency in the key parameter sequence, and generate a hidden state vector; specifically, the LSTM layer consists of an input gate, a forget gate, and an output gate. The input gate is used to determine the information i added to the cell state by the key parameter sequence. t =σ(Ui ·r t-1 +W i ·h t +b i ), and also generate new candidate memories through the tanh function The forget gate is used to decide whether to discard the information of the key parameter sequence in the cell state. t =σ(U f ·r t-1 +W f ·h t +b f ), and updates the cell state according to the output of the forget gate and the input gate The output gate is used to determine whether to output the information in the updated cell state to the hidden state r at the next moment. t =o t ⊙tanhc t , where h t represents the key parameter sequence at time t, σ represents the sigmoid function, tanh represents the hyperbolic sine function, represents the candidate storage state, c t represents the internal state, U and W both represent weight matrices, b represents the bias, ⊙ represents the dot product of the matrix, o t represents the activation function of the output gate, o t =σ(U o ·r t-1 +W o ·h t +b o ). LSTM can flexibly control the storage and deletion of information in the key parameter sequence through the three gating structures of forget gate, input gate and output gate, and can better process long sequence data and learn long-term dependencies, thereby improving the prediction accuracy of metal welding quality.

[0043] Step S3.2: Input the hidden state vector into the self-attention layer to calculate the attention score, and then perform softmax normalization to obtain the attention weight. Perform weighted summation of the attention score and the attention weight to capture the feature vector. The self-attention mechanism does not rely on any external information source, but directly focuses on the interaction between the internal elements of the hidden state vector. By calculating the correlation between the internal elements of the hidden state vector, it captures complex dependencies to enhance the expressiveness of the metal welding prediction model of thermal power plants and improve the prediction accuracy.

[0044] s=score(q a ,k j )×attention

[0045] Among them, score(q a ,k j) represents the attention score of the hidden state vector, q a represents the a-th query vector, k j represents the jth key vector; attention represents score(q a ,k j ) is normalized by softmax to obtain the attention weight, attention = softmax(score(q i ,k j ))v j , v j Represents the j-th value vector.

[0046] Step S3.3: Input the captured feature vector into the fully connected layer and output a feature vector of fixed length.

[0047] Step S3.4, input the fixed-length feature vector into the SVM classifier, and predict the metal welding quality through the polynomial kernel function. The polynomial kernel function can handle complex nonlinear relationships and achieve high-precision prediction of metal welding quality. The SVM classifier has excellent generalization ability and can achieve accurate prediction of metal welding quality.

[0048] Step S3.5, calculate the mean square error loss function of the predicted metal welding quality and the label, repeat steps S3.1-S3.4 until the mean square error loss function converges, and complete the training of the thermal power plant metal welding quality prediction model.

[0049] Step S4: collect key parameters of the metal welding process of the thermal power plant in real time, input them into the trained thermal power plant metal welding quality prediction model, and predict the metal welding quality.

[0050] The predicted metal welding quality is qualified or unqualified. If the predicted metal welding quality is unqualified, the key parameters of the metal welding are adjusted; if the predicted metal welding quality is still unqualified after adjusting the key parameters for multiple times, the operator is reminded to stop the machine for inspection.

[0051] The quality monitoring method of the metal welding process in a thermal power plant of the present invention constructs a metal welding quality prediction model for the thermal power plant, realizes real-time monitoring of the metal welding quality by acquiring key parameters of metal welding, automatically identifies the welding quality and provides feedback, so as to realize automatic optimization and quality control of the metal welding process, avoid manual dependence, and be able to monitor the metal welding process in real time.

[0052] In a technical solution of the present invention, a computer-readable storage medium is also provided, storing a computer program, wherein the computer program enables a computer to execute the quality monitoring method for the metal welding process in a thermal power plant.

[0053] In a technical solution of the present invention, an electronic device is also provided, comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the quality monitoring method for the metal welding process in a thermal power plant is implemented.

[0054] In a technical solution of the present invention, a computer program product is also provided, including a computer program, and when the computer program is executed by a processor, the quality monitoring method of the metal welding process in a thermal power plant is implemented.

[0055] In a technical solution of the present invention, a quality monitoring device for a metal welding process in a thermal power plant is also provided, comprising: a data acquisition unit, a data processing unit, a display and feedback unit, and an alarm and control unit;

[0056] The data acquisition unit is used to collect key parameters in the metal welding process of the thermal power plant, including: current, voltage, welding temperature, welding speed and molten pool state;

[0057] The data processing unit performs noise removal and normalization processing on the collected key parameters and then inputs them into the metal welding quality prediction model of the thermal power plant to predict the metal welding quality;

[0058] The display and feedback unit is used to display the predicted metal welding quality. If the metal welding quality is unqualified, key parameters are adjusted through the alarm and control unit.

[0059] In the embodiments disclosed in the present application, the computer storage medium can be a tangible medium that can contain or store a program for use by or in conjunction with an instruction execution system, device or equipment. The computer storage medium can include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, devices or equipment, or any suitable combination of the above. More specific examples of computer storage media can include electrical connections based on one or more lines, portable computer disks, hard disks, random access memories (RAM), read-only memories (ROM), erasable programmable read-only memories (EPROM or flash memory), optical fibers, portable compact disk read-only memories (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination of the above.

[0060] Those of ordinary skill in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed in this application can be implemented in electronic hardware or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can 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 this application.

[0061] The above are only preferred embodiments of the present invention, and the protection scope of the present invention is not limited to the above embodiments. All technical solutions under the concept of the present invention belong to the protection scope of the present invention. It should be pointed out that for ordinary technicians in this technical field, some improvements and modifications without departing from the principle of the present invention should be regarded as the protection scope of the present invention.

Claims

1. A quality monitoring method for metal welding process in a thermal power plant, characterized in that: The specific steps include: Step S1, collecting key parameters in the metal welding process of the thermal power plant, including: current, voltage, welding temperature, welding speed and molten pool state, and marking the welding quality according to the key parameters; Step S2, performing noise removal and normalization processing operations on the collected key parameters to form a key parameter sequence in time sequence; Step S3, constructing a thermal power plant metal welding quality prediction model, inputting the key parameter sequence into the thermal power plant metal welding quality prediction model in sequence, using the marked welding quality as a label, training the thermal power plant metal welding quality prediction model until the mean square error loss function converges, and completing the training of the thermal power plant metal welding quality prediction model; Step S4: collect key parameters of the metal welding process of the thermal power plant in real time, input them into the trained thermal power plant metal welding quality prediction model, and predict the metal welding quality.

2. The quality monitoring method of the metal welding process in a thermal power plant according to claim 1 is characterized in that: In step S3, the metal welding quality prediction model of a thermal power plant consists of an LSTM layer, a self-attention layer, a fully connected layer and an SVM classifier. The LSTM layer can capture the long-term dependencies in the key parameter sequence and generate a hidden state vector; the self-attention layer captures important information from the hidden state vector and generates a fixed-length feature vector through the fully connected layer; the SVM classifier is used to predict the metal welding quality based on the feature vector.

3. The quality monitoring method of the metal welding process in a thermal power plant according to claim 2 is characterized in that: Step S3 includes the following sub-steps: Step S3.1, input the key parameter sequence into the LSTM layer of the metal welding quality prediction model of the thermal power plant, capture the long-term dependency in the key parameter sequence, and generate a hidden state vector; Step S3.2: Input the hidden state vector into the self-attention layer to calculate the attention score, then perform softmax normalization to obtain the attention weight, perform weighted summation of the attention score and the attention weight to capture the feature vector; Step S3.3, input the captured feature vector into the fully connected layer, and output a feature vector of fixed length; Step S3.4, inputting the fixed-length feature vector into the SVM classifier, and predicting the metal welding quality through the polynomial kernel function; Step S3.5, calculate the mean square error loss function of the predicted metal welding quality and the label, repeat steps S3.1-S3.4 until the mean square error loss function converges, and complete the training of the thermal power plant metal welding quality prediction model.

4. The quality monitoring method of the metal welding process in a thermal power plant according to claim 3 is characterized in that: In step S3.1, the LSTM layer consists of an input gate, a forget gate, and an output gate. The input gate is used to determine the information i added to the cell state by the key parameter sequence. t =σ(U i ·r t-1 +W i ·h t +b i ), and also generate new candidate memories through the tanh function The forget gate is used to decide whether to discard the information of the key parameter sequence in the cell state. t =σ(U f ·r t-1 +W f ·h t +b f ), and updates the cell state according to the output of the forget gate and the input gate The output gate is used to determine whether to output the information in the updated cell state to the hidden state r at the next moment. t =o t ⊙tanhc t , where h t represents the key parameter sequence at time t, σ represents the sigmoid function, tanh represents the hyperbolic sine function, represents the candidate storage state, c t represents the internal state, U and W both represent weight matrices, b represents the bias, ⊙ represents the dot product of the matrix, o t represents the activation function of the output gate, o t =σ(U o ·r t-1 +W o ·h t +b o ).

5. The quality monitoring method for metal welding process in a thermal power plant according to claim 3 is characterized in that: The process of capturing the feature vector in step S3.2 is: s=score(q a ,k j )×attention Among them, score(q a ,k j ) represents the attention score of the hidden state vector, q a represents the a-th query vector, k j represents the jth key vector; attention represents score(q a ,k j ) is normalized by softmax to obtain the attention weight, attention = softmax(score(q i ,k j ))v j , v j Represents the j-th value vector.

6. The quality monitoring method of metal welding process in a thermal power plant according to claim 1, characterized in that: The predicted metal welding quality is qualified or unqualified. If the predicted metal welding quality is unqualified, the key parameters of the metal welding are adjusted; if the predicted metal welding quality is still unqualified after adjusting the key parameters for multiple times, the operator is reminded to stop the machine for inspection.

7. A computer-readable storage medium storing a computer program, characterized in that: The computer program enables the computer to execute the quality monitoring method for the metal welding process in a thermal power plant as described in any one of claims 1-6.

8. An electronic device, characterized in that: include: A memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the quality monitoring method for the metal welding process in a thermal power plant as described in any one of claims 1 to 6 is implemented.

9. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the quality monitoring method for the metal welding process in a thermal power plant described in any one of claims 1 to 6 is implemented.

10. A quality monitoring device for metal welding process in a thermal power plant, characterized in that: include: Data acquisition unit, data processing unit, display and feedback unit and alarm and control unit; The data acquisition unit is used to collect key parameters in the metal welding process of the thermal power plant, including: current, voltage, welding temperature, welding speed and molten pool state; The data processing unit performs noise removal and normalization processing on the collected key parameters and then inputs them into the metal welding quality prediction model of the thermal power plant to predict the metal welding quality; The display and feedback unit is used to display the predicted metal welding quality. If the metal welding quality is unqualified, key parameters are adjusted through the alarm and control unit.

Images