Welding process tracing method based on multi-source data fusion

Abstract

The invention relates to the technical field of welding process tracing, and discloses a welding process tracing method based on multi-source data fusion. The method comprises the following steps: firstly, collecting welding multi-source data, cleaning the data, setting an entity type and a relationship type based on a knowledge extraction method of a rule base, and then matching an attribute type to output a welding process triple; secondly, technological parameter correlation mining is carried out, linear correlation coefficients among technological parameters are calculated through correlation analysis, a welding technology control relation is obtained through recognition, the weight of a quality relation is calculated, and a welding technology knowledge graph is established; a welding process quality prediction model is established based on the full-connection neural network, the welding process quality is predicted, and a welding process quality prediction result is output; and finally, determining a tracing priority based on a multi-path fusion and sorting algorithm, and searching a reverse tracing path. The welding multi-source data are fused through the knowledge graph, the purpose of welding process tracing is achieved, and the method is objective and accurate.

Description

Technical Field

[0001] The present invention relates to the technical field of welding process tracing, and in particular to a welding process tracing method based on multi-source data fusion. Background Art

[0002] Multi-source data fusion technology significantly improves perception and decision-making capabilities in industrial scenarios by collecting, correlating, and comprehensively analyzing heterogeneous data from different sensors, systems, and links. In the field of welding process traceability, this technology constructs a holographic digital image of the welding process by integrating multi-dimensional data streams such as welding current and voltage waveforms, machine vision weld pool images, infrared thermal imaging temperature fields, gas flow and pressure values, and equipment status logs. Currently, multi-source data fusion is still in its early stages of application in welding traceability: most systems only implement basic-level data collection and archival storage (such as single parameter out-of-limit alarms) and have not yet formed in-depth data association mining capabilities; cross-system and cross-level data barriers are common, and time series alignment and feature association technologies are not yet mature; real-time data processing and immediate quality control decisions still face challenges.

[0003] Traditional welding process traceability methods mainly rely on operators to manually fill out parameter forms such as welding voltage, current, and speed based on the readings of welding machine instruments. Quality inspectors independently record the location and type of welding process quality defects, use barcode labels to connect paper reports of weldments in different processes (such as assembly, welding, and flaw detection), and establish basic process associations by scanning barcodes to obtain independent data.

[0004] In traditional welding process traceability methods, multi-source data such as process parameters, equipment logs, and quality inspection reports during the welding process are stored independently. Time and space alignment requires manual matching, resulting in serious data silos. This makes it difficult to quantify and capture the core influencing factors of welding quality, and it is impossible to trace the root cause when quality problems occur. Summary of the Invention

[0005] In response to the problems in the related art, the present invention provides a welding process tracing method based on multi-source data fusion to overcome the technical problems existing in the existing related art.

[0006] To solve the above technical problems, the present invention is achieved through the following technical solutions: The present invention is a welding process tracing method based on multi-source data fusion, comprising the following steps: S1. Collect multi-source welding data and perform data cleaning. Based on the knowledge extraction method of the rule base, set the entity type and relationship type, introduce the similarity coefficient to identify entities and relationships, and then match the attribute type to output the welding process triple; S2. Based on the welding process triples, process parameter association mining is performed, linear correlations between process parameters are calculated through correlation analysis, welding process control relationships are identified, and quality relationship weights are calculated to establish a welding process knowledge graph; S3. Establish a welding process quality prediction model based on a fully connected neural network, predict the welding process quality, and output the welding process quality prediction result; S4. Set path search rules, determine the traceability priority based on the multi-path fusion and sorting algorithm in the welding process knowledge graph according to the welding process quality prediction results, find the reverse traceability path, and realize welding process traceability.

[0007] Preferably, collecting welding multi-source data and performing data cleaning includes: Collect welding multi-source data, and record the text data in the welding multi-source data as a text data set to be cleaned; The text data set to be cleaned is subjected to unit unification processing and time stamp alignment, and then a welding term library is set for adaptive word segmentation, stop word filtering, and entity type marking to generate a processed welding multi-source data set.

[0008] Preferably, the introducing similarity coefficients to identify entities and relationships includes: The processed welding multi-source data set is used to set a relationship type library based on a rule library to obtain relationship types; Establishing an entity type combination according to the entity type, obtaining entity type attributes, and calculating similarity coefficients of the entity type attributes; Set a similarity threshold. When the similarity coefficient is not greater than the similarity threshold, there is no relationship type between the entity type combinations, and the corresponding entity type combination is deleted. Otherwise, there is a relationship type between the entity type combinations, and the corresponding relationship type is marked between the entity type combinations. The entities and relationships are identified and a tuple is established.

[0009] Preferably, the re-matching attribute type output welding process triplet includes: According to the entity type attributes in the entity type combination, attribute matching is performed on the entities in the binary group until the binary group matching is completed to form a welding process triplet.

[0010] Preferably, calculating the linear correlation between process parameters by correlation analysis includes: According to the process parameter data corresponding to the entities in the welding process triplet, a standardized process parameter data set is obtained by standardization processing, and a linear correlation coefficient between the process parameters is obtained by correlation analysis calculation.

[0011] Preferably, calculating the quality relationship weight to establish the welding process knowledge graph includes: Set a correlation threshold. When the linear correlation coefficient is greater than the correlation threshold, the corresponding process parameters are recorded as positively correlated to form a positively correlated process parameter pair. Otherwise, they are recorded as negatively correlated to form a negatively correlated process parameter pair. Then, a forward control chain and a reverse control chain are established in sequence to generate a welding process control relationship. The entities in the welding process control relationship are used as knowledge graph nodes, the quality relationship weights are calculated, the relationships and attributes in the welding process control relationship are used as knowledge graph edges, the knowledge graph nodes are connected using knowledge graph edges, and the quality relationship weights are marked to establish a welding process knowledge graph.

[0012] Preferably, establishing a welding process quality prediction model based on a fully connected neural network includes: Re-acquire new weld samples, collect welding multi-source data and environmental parameters, and form a multi-source data fusion tensor; The welding process quality prediction model is set to include an input layer, a hidden layer and an output layer, wherein the hidden layer uses a three-layer dense layer structure, which are respectively recorded as the first dense layer, the second dense layer and the third dense layer. The gradient norm is restricted. The first dense layer adopts the ELU activation function and performs batch normalization. The Dropout ratio is set and transferred to the second dense layer. The second dense layer adopts the Leaky Relu activation function and is transferred to the third dense layer. The third dense layer adopts the Swish activation function to establish a welding process quality prediction model.

[0013] Preferably, the predicting of welding process quality includes: Dividing the multi-source data fusion tensor into a sample training set and a sample test set, inputting the tensor into a welding process quality prediction model for training, and obtaining a final welding process quality prediction model; The processed welding multi-source data set is combined with environmental parameters and input into a final welding process quality prediction model to output a welding process quality prediction result.

[0014] Preferably, determining the tracing priority based on a multi-path fusion and sorting algorithm includes: Set welding process quality constraints. When the welding process quality prediction result exceeds the welding process quality constraints, the welding process traceability trigger condition is met, and the path search rules are set to generate a first-level associated knowledge graph node. When the first-level associated knowledge graph node is the root node, the tracing path is output at this time, otherwise the knowledge graph node is traced back in reverse, the tracing path integral is calculated, and the tracing priority is obtained.

[0015] Preferably, the searching for a reverse tracing path includes: Select the traceability path corresponding to the highest traceability priority and record it as the reverse traceability path. Record all knowledge graph edges and knowledge graph nodes on the reverse traceability path. Perform spatial mapping on the reverse traceability path in chronological order to obtain the position of the knowledge graph node, determine the time when the welding process quality defects occur, and realize welding process traceability.

[0016] The present invention has the following beneficial effects: 1. This invention collects multi-source welding data and performs data cleaning to unify heterogeneous data into structured entities, thereby solving the problem of traditional welding data islands. It uses a knowledge extraction method based on a rule base to set entity types and relationship types, and introduces a similarity coefficient calculation and threshold judgment mechanism to match attribute types, breaking through the limitations of manual experience and achieving objective diagnosis of process anomalies. The use of knowledge graph technology facilitates subsequent rapid retrieval and reasoning.

[0017] 2. This invention conducts process parameter association mining, uses correlation analysis to calculate the linear correlation between process parameters, identifies the welding process control relationship, generates a forward / reverse control chain to achieve adaptive parameter adjustment, and establishes a welding process knowledge graph. Through the chain reasoning ability of the knowledge graph, compared with traditional manual analysis methods, the tracing time of welding defects (such as porosity and lack of fusion) is greatly reduced.

[0018] 3. This invention predicts the welding process quality by establishing a welding process quality prediction model based on a fully connected neural network. It can effectively capture nonlinear process characteristics, ensure that the model converges stably in dynamic data streams, and trigger a self-compensation mechanism, which greatly improves the quality prediction accuracy. It can achieve parallel high-precision prediction of the three key indicators of penetration depth, porosity, and hardness, output the welding process quality prediction results, and then feed back the welding process knowledge graph to form a continuously optimized intelligent decision-making closed loop.

[0019] 4. This invention sets path search rules, determines the traceability priority based on multi-path fusion and sorting algorithms, quickly locks the key path, reduces the time for troubleshooting welding process causes, automatically strengthens the weight of high-frequency fault paths, and accurately associates the time / space nodes of defects and equipment failures, thereby reducing the incidence of similar defects. Find the reverse traceability path, accurately trace the welding process, and complete the traceability closed-loop optimization.

[0020] Of course, any product implementing the present invention does not necessarily need to achieve all of the advantages described above at the same time. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] In order to more clearly illustrate the technical solutions of the embodiments of the invention, the following briefly introduces the drawings required for describing the embodiments. Obviously, the drawings described below are only some embodiments of the invention. For ordinary technicians in this field, they can also obtain drawings based on these drawings without paying any creative work.

[0022] Figure 1 The present invention provides a flow chart of a welding process tracing method based on multi-source data fusion; Figure 2 This is a schematic diagram of the process of establishing a welding process quality prediction model provided by the present invention. DETAILED DESCRIPTION

[0023] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making any creative efforts shall fall within the scope of protection of the present invention.

[0024] In traditional welding process traceability methods, multi-source data such as process parameters, equipment logs, and quality inspection reports during the welding process are stored independently. Time and space alignment requires manual matching, resulting in serious data silos. This makes it difficult to quantify and capture the core influencing factors of welding quality, and it is impossible to trace the root cause when quality problems occur.

[0025] To address the above technical issues, a specific embodiment is described in detail below: an energy equipment company's LNG pipeline automatic welding production line (with an annual output of 50 km of welds) uses a Fronius TPS-500 welding machine (with real-time monitoring of current, voltage, and wire feed speed). Environmental sensors are used to collect preheat temperature (with an accuracy of ±1°C) and ambient humidity (±3%RH). The quality inspection system uses an X-ray flaw detector (with a porosity detection accuracy of 0.1 mm). These provide a good foundation for the implementation of the embodiments of the present invention.

[0026] In the specific implementation process of the above embodiment, first, multi-source welding data is collected and data cleaning is performed. Based on the knowledge extraction method of the rule base, entity types and relationship types are set, similarity coefficients are introduced to identify entities and relationships, and then attribute types are matched to output welding process triples; timestamp alignment and terminology adaptive word segmentation are used to unify heterogeneous data into structured entities, solving the traditional welding data island problem. At the same time, similarity coefficient calculation and threshold judgment mechanism are introduced to break through the limitations of manual experience and achieve objective diagnosis of process anomalies. Knowledge graph technology is used to facilitate subsequent rapid retrieval and reasoning; secondly, process parameter association mining is performed, and the linear correlation between process parameters is calculated through correlation analysis to identify the welding process control relationship, and the quality relationship weight is calculated to establish a welding process knowledge graph; this method uses linear correlation to generate forward / reverse control chains to achieve parameter adaptive adjustment. Through the chain reasoning ability of the knowledge graph, compared with the traditional manual analysis method, it is much easier. The tracing time of welding defects (such as porosity and lack of fusion) is greatly reduced; then a welding process quality prediction model is established based on a fully connected neural network to predict the welding process quality and output the welding process quality prediction results; the model can effectively capture nonlinear process characteristics, ensure the stable convergence of the model in the dynamic data stream, and trigger the self-compensation mechanism, which greatly improves the quality prediction accuracy and realizes parallel high-precision prediction of the three key indicators of penetration depth, porosity and hardness. The prediction results feed back into the welding process knowledge graph to form a continuously optimized intelligent decision-making closed loop; finally, path search rules are set, and the tracing priority is determined based on the multi-path fusion and sorting algorithm, and the reverse tracing path is found. The key path is quickly locked through the multi-path fusion and sorting algorithm, which reduces the time for troubleshooting the causes of the welding process, automatically strengthens the weight of the high-frequency fault path, accurately associates the time / space nodes of defects and equipment failures, reduces the incidence of similar defects, accurately realizes welding process traceability, and completes the traceability closed loop optimization.

[0027] Furthermore, in order to better introduce the technical solutions of the embodiments of the present invention, Figure 1 As shown, an embodiment of the present invention provides a welding process tracing method based on multi-source data fusion, which specifically includes the following contents: S1. Collect multi-source welding data and perform data cleaning. Based on the knowledge extraction method of the rule base, set the entity type and relationship type, introduce the similarity coefficient to identify entities and relationships, and then match the attribute type to output the welding process triple; Said S1 comprises the following steps: S11. Label different weld passes during the welding process to form weld pass IDs, and then collect welding multi-source data, wherein the welding multi-source data includes process parameter data, process specification text, quality inspection reports, equipment logs, etc., wherein the process parameter data includes welding current, arc voltage, wire feed speed, etc., and record the text data in the welding multi-source data as the text data set to be cleaned; The text data set to be cleaned is subjected to unit unification processing and time stamp alignment, and then a welding term library is set up. The text data set to be cleaned is adaptively segmented, the text data is segmented and synonyms are merged to obtain a segmented text data set; stop word filtering is then performed and entity type marking is performed. The entity types include process parameters, equipment entities, material attributes, etc., to generate a processed welding multi-source data set; S12. For the processed welding multi-source data set, a relationship type library is set based on a rule library, wherein the relationship type library includes causal influence, impact range, equipment association, etc., to obtain relationship types; then, entity type combinations are established based on the entity types, and similarity coefficients are introduced to determine the relationship types between the entity type combinations, thereby identifying entities and relationships. The specific steps are as follows: S121. Obtain entity type attributes of the entity type combination, wherein the entity type attributes include a standard range, a level, specific data, etc., which are respectively recorded as a first entity type attribute set and a second entity type attribute set. Calculate the intersection and union of the first entity type attribute set and the second entity type attribute set, then assign an attribute weight to the intersection of the first entity type attribute set and the second entity type attribute set, and calculate a similarity coefficient. ,in Represents the first entity type attribute set, Represents the second entity type attribute set, Represents the first entity type attribute set and the second entity type attribute set a The attribute weight of the intersection of S122. Set a similarity threshold. When the similarity coefficient is not greater than the similarity threshold, there is no relationship type between the entity type combinations, and the corresponding entity type combination is deleted. Otherwise, there is a relationship type between the entity type combinations, and the corresponding relationship type is marked between the entity type combinations. Entities and relationships are identified and a tuple is established. S13. According to the entity type attributes in the entity type combination, attribute matching is performed on the entities in the binary group according to the entity type attributes, until the binary group matching is completed, thereby forming a {entity, relationship, attribute} triple, which is recorded as a welding process triple; In this embodiment, multi-source welding data is collected and cleaned. Based on the knowledge extraction method of the rule base, entity types and relationship types are set, similarity coefficients are introduced to identify entities and relationships, and then attribute types are matched to output welding process triples. Timestamp alignment and terminology adaptive word segmentation are used to unify heterogeneous data into structured entities, solving the traditional welding data island problem. At the same time, similarity coefficient calculation and threshold judgment mechanism are introduced to break through the limitations of manual experience and achieve objective diagnosis of process anomalies. Knowledge graph technology is used to facilitate subsequent rapid retrieval and reasoning. Specifically, for example, a weld ID is marked for a stainless steel pipe welding task, and weld IDs are collected. Connect to multiple source data, including process parameters: current 150A, voltage 22V, wire feed speed 12cm / min, process specification text: use ER308L welding wire, interpass temperature ≤ 150℃, quality inspection report: weld bead #W107 has 0.8mm air holes, equipment log: abnormal vibration of wire feeder (2025-05-10 14:22); unit standardization: 12cm / min → 120mm / min, original text: "interpass temperature must be strictly controlled", after word segmentation: ["interpass temperature", "need", "strict", "control"], synonym merging: "welding wire ER308L" → "filler material: E R308L", stop word filtering: remove "need", "strict", "control", etc.; among them, current 150A → process parameters, wire feeder → equipment entity, ER308L → material properties; set relationship type library: causal influence (such as voltage fluctuation → pores), equipment association (such as wire feeder → wire feeding speed), first entity type attribute set (welding current attribute set) and second entity type attribute set (penetration attribute set), among which welding current attribute set: {value, unit (A), stability}, penetration attribute set: {depth value, unit (mm), uniformity, current association}, intersection: {unit} → cardinality 1, union: {value, unit (A), stability, depth value, unit (mm), uniformity, current association} → cardinality 7; set attribute weights based on domain importance and calculate the similarity coefficient; obtain welding multi-source data generated by 500 welding tasks, compare the similarities between different entity types, and find that when the similarity is 0.3, there is an association between entity types. Label the corresponding relationship type and perform attribute matching to form a triple {entity, relationship, attribute}, such as {voltage 22V, causal influence, abnormal fluctuation ±2V}, {wire feeder XYZ200, equipment association, rated wire feeding speed 130mm / min}, etc. S2. Based on the welding process triples, process parameter association mining is performed, linear correlations between process parameters are calculated through correlation analysis, welding process control relationships are identified, and quality relationship weights are calculated to establish a welding process knowledge graph; The S2 comprises the following steps: S21, according to the process parameter data corresponding to the entity in the welding process triplet, according to the weld ID Record the process parameter data in sequence to obtain the initial process parameter data set, calculate the mean and range of the process parameter data for each weld pass, subtract the mean of the process parameter data from the process parameter data and divide it by the range of the process parameter data to obtain the standardized process parameter data set; Select different types of process parameter data from the standardized process parameter data set, record them as welding current data set and arc voltage data set respectively, calculate the mean of welding current data set and the mean of arc voltage data set, calculate the linear correlation coefficient between process parameters through correlation analysis, and obtain the linear correlation coefficient ,in Indicates the number of the welding current data set. i Welding current data, Indicates the arc voltage data set i Arc voltage data, represents the mean value of the welding current data set, represents the mean value of the arc voltage data set, m Indicates the number of weld bead IDs, ; S22, calculating linear correlation coefficients for all types of process parameter data in the standardized process parameter data set in sequence to generate a correlation matrix; setting a correlation threshold, and when the linear correlation coefficient is greater than the correlation threshold, recording the corresponding process parameters as positively correlated to form a positively correlated process parameter pair; otherwise, recording the corresponding process parameters as negatively correlated to form a negatively correlated process parameter pair; establishing a forward control chain between the positively correlated process parameter pairs, and establishing a reverse control chain between the negatively correlated process parameter pairs, to convert the welding process triples into a welding process control relationship; S23. Calculate the quality relationship weight based on the forward control chain and the reverse control chain in the welding process control relationship and establish a welding process knowledge graph. The specific steps are as follows: S231. Entities in the welding process control relationship are used as knowledge graph nodes. Any knowledge graph node is selected and recorded as a key knowledge graph node. Knowledge graph nodes that have a forward control chain and a reverse control chain with the key knowledge graph node are found. A random forest is used to calculate the quality relationship weight by aggregating the feature contributions of multiple decision trees. S232. For the welding process triples in the welding process control relationship, use the relationships and attributes in the welding process control relationship as knowledge graph edges, and connect the knowledge graph nodes using the knowledge graph edges; extract all welding process triples in sequence, connect the knowledge graph nodes and mark the quality relationship weights, to establish a welding process knowledge graph; In this embodiment, process parameter association mining is performed, and the linear correlation between process parameters is calculated through correlation analysis, the welding process control relationship is identified, and the quality relationship weight is calculated to establish a welding process knowledge graph; this method uses linear correlation to generate a forward / reverse control chain to achieve parameter adaptive adjustment. Through the chain reasoning ability of the knowledge graph, the tracing time of welding defects (such as porosity and lack of fusion) is greatly reduced compared with the traditional manual analysis method; specifically, for example, the marked weld IDs are: #W101-#W110 (a total of 10 welds), where the current average is 152A, the range is 20A, and the voltage average is 10A. The value is 23V, the range is 5V, the mean wire feeding speed is 118mm / min, the range is 30mm / min, etc.; taking the current mean 152A and the voltage mean 23V as an example, the linear correlation coefficient is calculated to be -0.85. The linear correlation coefficients between the process parameters are calculated in turn for the welding multi-source data generated by 500 welding jobs. The analysis shows that only when the linear correlation coefficient is greater than 0.8, there is correlation between the process parameters. Therefore, the correlation threshold is set to 0.8, and it is concluded that current and voltage are a positively correlated process parameter pair; according to the linear correlation coefficients between power, voltage and wire feeding speed, the correlation matrix is ​​established: , generating a control chain, where the forward control chain is: current increase → forward drive → wire feeding speed increase, and the reverse control chain is: voltage increase → reverse inhibition → current decrease, wire feeding speed increase → reverse inhibition → voltage decrease. Using random forest weight calculation, the contribution of voltage fluctuation features is 0.38, and the contribution of wire feeding anomaly features is 0.29. A knowledge graph is established: nodes represent entities (current, voltage, wire feeding speed, porosity defects), and edges represent control relationships and quality weights. S3. Establish a welding process quality prediction model based on a fully connected neural network, predict the welding process quality, and output the welding process quality prediction result; The S3 includes the following steps: S31. Reacquire a new weld sample, collect welding multi-source data and environmental parameters, wherein the environmental parameters include preheating temperature, ambient humidity, working temperature, etc., obtain a time series segment of the collected welding multi-source data, compose a multi-source data fusion tensor, and set welding multi-source data constraints; S32, establishing a welding process quality prediction model based on a fully connected neural network, and performing model training, predicting the welding process quality according to the processed welding multi-source data set, and outputting the welding process quality prediction result, such as Figure 2 The specific steps for establishing the welding process quality prediction model are as follows: S321. Set a welding process quality prediction model including an input layer, a hidden layer, and an output layer, wherein the hidden layer uses a three-layer dense layer structure, which are respectively recorded as the first dense layer, the second dense layer, and the third dense layer, and restrict the gradient norm. The first dense layer uses the ELU activation function and performs batch normalization. The Dropout ratio is set and the data is transferred to the second dense layer. The second dense layer uses the LeakyRelu activation function and is transferred to the third dense layer. The third dense layer uses the Swish activation function to establish a welding process quality prediction model. S322. Divide the multi-source data fusion tensor into a sample training set and a sample test set, input the sample training set into the welding process quality prediction model for training until the welding process quality prediction model converges, and obtain a trained welding process quality prediction model; then input the sample test set into the trained welding process quality prediction model, set an accuracy threshold, and when the accuracy of the output result is greater than the accuracy threshold, stop the iteration to obtain a final welding process quality prediction model; otherwise, adjust the weight until the accuracy of the output result is greater than the accuracy threshold; S323. Input the processed welding multi-source data set into a final welding process quality prediction model in combination with environmental parameters, and output a welding process quality prediction result, wherein the welding process quality prediction result includes a welding penetration prediction value, a porosity prediction value, and a maximum hardness prediction value of the welding heat-affected zone; In this embodiment, a welding process quality prediction model is established based on a fully connected neural network to predict the welding process quality and output the welding process quality prediction results. The model can effectively capture nonlinear process characteristics, ensure the stable convergence of the model in a dynamic data stream, and trigger a self-compensation mechanism, which greatly improves the quality prediction accuracy and realizes parallel high-precision prediction of the three key indicators of penetration depth, porosity, and hardness. The prediction results feed back into the welding process knowledge graph to form a continuously optimized intelligent decision-making closed loop. Specifically, for example, the weld ID: #W201-#W250 is selected to obtain welding multi-source data (current, voltage, wire feed speed) and environmental parameters (preheating temperature, ambient humidity, etc.). The fusion tensor dimensions are [50 welds] × [120 time points] × [6 parameters], with a current fluctuation range of ±15A, a humidity safety threshold of ≤70%RH, and a penetration qualification standard of ≥5.0mm. The model is trained with the training set of 30 welds #W201-#W230 and the test set of 20 welds #W231-#W250, obtaining the final welding process quality prediction model. For example, the multi-source welding data and environmental parameters corresponding to weld #W107 are input, and the predicted weld penetration is 5.72mm, the porosity is 1.25%, and the maximum hardness of the heat-affected zone is 254HV. S4. Setting a path search rule, determining the tracing priority based on the welding process quality prediction result in the welding process knowledge graph and a multi-path fusion and sorting algorithm, and finding a reverse tracing path to realize welding process tracing; The S4 comprises the following steps: S41. According to the welding process quality prediction result, a welding process quality constraint is set. When the welding penetration prediction value, the porosity prediction value, or the maximum hardness prediction value of the welding heat-affected zone exceeds the welding process quality constraint, the welding process tracing trigger condition is met. A path search rule is set, a dynamic weighted path search is performed, and a tracing path integral is calculated. The specific steps are as follows: S411. After the welding process traceability trigger condition is met, a knowledge graph node corresponding to the welding process quality defect is determined in the welding process knowledge graph as the starting point of the traceability path; a maximum traceability level of the knowledge graph node is set, a weight filtering threshold is set, and knowledge graph node edges corresponding to the knowledge graph nodes whose quality relationship weights are greater than the weight filtering threshold are retained to obtain a first-level associated knowledge graph node; S412. When the first-level associated knowledge graph node is a root node, output the traceability path. Otherwise, continue to trace the knowledge graph node in reverse, determine whether the associated knowledge graph node is a root node, until the maximum traceability level of the knowledge graph node is reached, determine the total number of traceability paths, and obtain several traceability paths to form a traceability path set; at this time, calculate the product of the quality relationship weights of all associated knowledge graph nodes on the traceability path to obtain the traceability path score; S42. Find the traceability path score corresponding to the traceability path in the traceability path set, calculate the data completeness rate of the traceability path in the traceability path set, and calculate the time from the starting point of the traceability path to the key point of the traceability path in the traceability path, which is recorded as the timeliness closeness; assign weights to the traceability path score, the data completeness rate, and the timeliness closeness to calculate the total score of the traceability path; sort the traceability paths in descending order according to the total score, corresponding to the traceability priority in turn; select the traceability path corresponding to the highest traceability priority, record it as the reverse traceability path, record all knowledge graph edges and knowledge graph nodes on the reverse traceability path, spatially map the reverse traceability path in chronological order, obtain the position of the knowledge graph node, determine the occurrence time of the welding process quality defect, and realize welding process traceability; In this embodiment, a path search rule is set, and the tracing priority is determined based on a multi-path fusion and sorting algorithm. The reverse tracing path is found, and the key path is quickly locked through the multi-path fusion and sorting algorithm, which reduces the time for troubleshooting the welding process causes, automatically strengthens the weight of the high-frequency fault path, accurately associates the time / space nodes of the defect and the equipment failure, reduces the incidence of similar defects, and accurately realizes the welding process traceability. Specifically, for example, the prediction result of weld bead #W300 is: porosity = 3.5% > constraint threshold (2.0%), defect node positioning: porosity defect node in the knowledge graph, with QP300 as the starting point of the tracing path, setting the maximum tracing level to 3 levels to limit the search depth, and the weight filtering threshold to 0.5, retaining the key influencing path; generating a path set: Path 1: QP300←[weight filtering threshold 0.76] voltage fluctuation (level 1), Path 2: QP300←[weight filtering threshold 0.76] voltage fluctuation← wire feeding speed abnormality (level 2), Path 3: QP300←[weight filtering threshold 0.76] voltage fluctuation← current fluctuation (2 Level), Path 4: QP300←[weight filtering threshold 0.76] voltage fluctuation←abnormal wire feeding speed←[weight filtering threshold 0.58] wire feeder vibration (Level 3), calculate the traceability path integrals in sequence (for example, path 1 integral: 0.76, path 4 integral: 0.76×1.0×0.58=0.441, the default value is 1.0 when there is no weight filtering threshold); multi-path fusion sorting: Path 1: data completeness rate 100%, timeliness closeness (hours) 0.8, Path 4: data completeness rate 85%, timeliness closeness ( hours) 3.2, assign weights to the traceability path integral, data completeness and timeliness closeness, establish a new knowledge graph based on the welding multi-source data generated by 500 welding jobs, analyze the traceability time of each path in the new knowledge graph, and quantify the influence of the traceability path integral, data completeness and timeliness closeness on the path traceability time, with weights of 0.6, 0.2 and 0.2 respectively; at this time, the total integral of path 1 is 0.906 (the highest priority), and the root cause is located: voltage fluctuation, the spatial position is obtained, and the welding process traceability is realized.

[0028] Throughout this specification, references to terms such as "one embodiment," "example," or "specific example" indicate that the specific features, structures, materials, or characteristics described in conjunction with that embodiment or example are included in at least one embodiment or example of the invention. In this specification, schematic representations of these terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in any one or more embodiments or examples.

[0029] The preferred embodiments of the invention disclosed above are intended only to help illustrate the invention. These preferred embodiments do not exhaust all details, nor do they limit the invention to the specific embodiments described. Obviously, many modifications and variations are possible based on the content of this specification. These embodiments are selected and described in detail in this specification to better explain the principles and practical applications of the invention, thereby enabling those skilled in the art to better understand and utilize the invention.

Claims

1. A welding process tracing method based on multi-source data fusion, characterized in that: include: S1. Collect multi-source welding data and perform data cleaning. Based on the knowledge extraction method of the rule base, set the entity type and relationship type, introduce the similarity coefficient to identify entities and relationships, and then match the attribute type to output the welding process triple; S2. Based on the welding process triples, process parameter association mining is performed, linear correlations between process parameters are calculated through correlation analysis, welding process control relationships are identified, and quality relationship weights are calculated to establish a welding process knowledge graph; S3. Establish a welding process quality prediction model based on a fully connected neural network, predict the welding process quality, and output the welding process quality prediction result; S4. Set path search rules, determine the traceability priority based on the multi-path fusion and sorting algorithm in the welding process knowledge graph according to the welding process quality prediction results, find the reverse traceability path, and realize welding process traceability.

2. The welding process tracing method based on multi-source data fusion according to claim 1 is characterized in that: The collecting of welding multi-source data and data cleaning includes: Collect welding multi-source data, and record the text data in the welding multi-source data as a text data set to be cleaned; The text data set to be cleaned is subjected to unit unification processing and time stamp alignment, and then a welding term library is set for adaptive word segmentation, stop word filtering, and entity type marking to generate a processed welding multi-source data set.

3. The welding process tracing method based on multi-source data fusion according to claim 2 is characterized in that: The introduction of similarity coefficients to identify entities and relationships includes: The processed welding multi-source data set is used to set a relationship type library based on a rule library to obtain relationship types; Establishing an entity type combination according to the entity type, obtaining entity type attributes, and calculating similarity coefficients of the entity type attributes; Set a similarity threshold. When the similarity coefficient is not greater than the similarity threshold, there is no relationship type between the entity type combinations, and the corresponding entity type combination is deleted. Otherwise, there is a relationship type between the entity type combinations, and the corresponding relationship type is marked between the entity type combinations. The entities and relationships are identified and a tuple is established.

4. The welding process tracing method based on multi-source data fusion according to claim 3 is characterized in that: The re-matching attribute type output welding process triplet includes: According to the entity type attributes in the entity type combination, attribute matching is performed on the entities in the binary group until the binary group matching is completed to form a welding process triplet.

5. The welding process tracing method based on multi-source data fusion according to claim 4 is characterized in that: Calculating the linear correlation between process parameters by correlation analysis includes: According to the process parameter data corresponding to the entities in the welding process triplet, a standardized process parameter data set is obtained by standardization processing, and a linear correlation coefficient between the process parameters is obtained by correlation analysis calculation.

6. The welding process tracing method based on multi-source data fusion according to claim 5 is characterized in that: The calculation of the quality relationship weight to establish the welding process knowledge graph includes: Set a correlation threshold. When the linear correlation coefficient is greater than the correlation threshold, the corresponding process parameters are recorded as positively correlated to form a positively correlated process parameter pair. Otherwise, they are recorded as negatively correlated to form a negatively correlated process parameter pair. Then, a forward control chain and a reverse control chain are established in sequence to generate a welding process control relationship. The entities in the welding process control relationship are used as knowledge graph nodes, the quality relationship weights are calculated, the relationships and attributes in the welding process control relationship are used as knowledge graph edges, the knowledge graph nodes are connected using knowledge graph edges, and the quality relationship weights are marked to establish a welding process knowledge graph.

7. The welding process tracing method based on multi-source data fusion according to claim 6 is characterized in that: The welding process quality prediction model based on the fully connected neural network includes: Re-acquire new weld samples, collect welding multi-source data and environmental parameters, and form a multi-source data fusion tensor; The welding process quality prediction model is set to include an input layer, a hidden layer and an output layer, wherein the hidden layer uses a three-layer dense layer structure, which are respectively recorded as the first dense layer, the second dense layer and the third dense layer. The gradient norm is restricted. The first dense layer adopts the ELU activation function and performs batch normalization. The Dropout ratio is set and transferred to the second dense layer. The second dense layer adopts the Leaky Relu activation function and is transferred to the third dense layer. The third dense layer adopts the Swish activation function to establish a welding process quality prediction model.

8. The welding process tracing method based on multi-source data fusion according to claim 7 is characterized in that: The prediction of welding process quality includes: Dividing the multi-source data fusion tensor into a sample training set and a sample test set, inputting the tensor into a welding process quality prediction model for training, and obtaining a final welding process quality prediction model; The processed welding multi-source data set is combined with environmental parameters and input into a final welding process quality prediction model to output a welding process quality prediction result.

9. The welding process tracing method based on multi-source data fusion according to claim 8 is characterized in that: Determining the tracing priority based on the multi-path fusion and sorting algorithm includes: Set welding process quality constraints. When the welding process quality prediction result exceeds the welding process quality constraints, the welding process traceability trigger condition is met, and the path search rules are set to generate a first-level associated knowledge graph node. When the first-level associated knowledge graph node is the root node, the tracing path is output at this time, otherwise the knowledge graph node is traced back in reverse, the tracing path integral is calculated, and the tracing priority is obtained.

10. The welding process tracing method based on multi-source data fusion according to claim 9, characterized in that: The searching for the reverse tracing path includes: Select the traceability path corresponding to the highest traceability priority and record it as the reverse traceability path. Record all knowledge graph edges and knowledge graph nodes on the reverse traceability path. Perform spatial mapping on the reverse traceability path in chronological order to obtain the position of the knowledge graph node, determine the time when the welding process quality defects occur, and realize welding process traceability.

Images