Steel rail thermit welding seam quality evaluation method based on machine learning and multi-factor monitoring

Abstract

The invention is suitable for the technical field of rail traffic, and provides a steel rail thermit welding seam quality evaluation method based on machine learning and multi-factor monitoring, and the method comprises the following steps: collecting thermit welding process data, and storing the collected data in a field welding management system; based on the collected data, a random forest algorithm is utilized to construct an exothermic welding seam quality evaluation model; a historical thermit welding process data training model is used, a mapping relation between all factors in the welding process and weld defects is established, the importance of features is evaluated by calculating the splitting contribution degree of each feature in a decision tree, and the influence weight of all the factors is determined; and based on the influence degree weight matrix of each factor, determining a score deduction mechanism when each factor does not reach the standard in combination with expert experience according to an exothermic welding quality evaluation rule, and constructing a scientific welding seam quality evaluation system. According to the method, the welding multi-factor data is comprehensively considered, and the welding seam quality quantitative evaluation method based on data driving can be provided.

Description

Technical Field

[0001] The present invention relates to the field of rail transportation technology, and in particular to a method for evaluating rail thermite welding seam quality based on machine learning and multi-factor monitoring. Background Art

[0002] Rail welding is a critical process in railway construction and maintenance, and its quality directly impacts the safety and stability of railway operations. Currently, flash welding, gas pressure welding, and thermite welding are the primary methods used for rail welding. While thermite welding is widely used due to its ease of operation, its weld quality is affected by numerous factors, including basic conditions (such as weather, temperature, and rail wear), welding process parameters (such as rail gap, gas pressure, preheating time, and deburring time), and post-weld treatment (such as rail surface straightness).

[0003] Currently, traditional weld quality assessment methods rely primarily on manual inspection, confirming the presence of surface defects such as cracks and pores, or using ultrasonic waves to detect internal weld defects. These methods are significantly influenced by operator experience, and some weld defects cannot be detected immediately, resulting in numerous limitations.

[0004] In recent years, with the development of big data and artificial intelligence technologies, some studies have begun to use machine learning models to predict weld quality. For example, they consider the impact of factors such as weld gap or welding temperature on weld quality. However, these methods are mostly based on data from a single factor, ignoring the combined influence of multiple factors, resulting in limited evaluation results.

[0005] The existing technology lacks a data-driven quantitative weld quality evaluation method that comprehensively considers multiple welding factors. Therefore, developing a rail thermite weld quality evaluation method based on machine learning and multi-factor monitoring has important practical significance and application value. Summary of the Invention

[0006] The present invention provides a rail thermite welding seam quality evaluation method based on machine learning and multi-factor monitoring to solve the technical problems mentioned in the background technology.

[0007] A method for evaluating rail thermite welding seam quality based on machine learning and multi-factor monitoring, the method comprising:

[0008] S1. Collect thermite welding process data and store the collected data in the on-site welding management system;

[0009] S2. Based on the collected data, a random forest algorithm is used to construct a thermite welding seam quality evaluation model. The model is trained using historical thermite welding process data to establish a mapping relationship between factors such as weather, rail temperature, rail gap, gas pressure, preheating time, and calming time during the welding process and weld defects. The importance of each factor is evaluated by calculating its split contribution in the decision tree, determining its degree of dominance over weld defects and thus determining the influence weight of each factor.

[0010] S3. Based on the weight matrix of the influence degree of each factor, according to the aluminothermic welding quality evaluation regulations, combined with expert experience, determine the deduction mechanism when each factor does not meet the standard, and build a scientific weld quality evaluation system.

[0011] As a further technical solution of the present invention, in step S1, the collected data is stored in the on-site welding management system. The system creates multiple data tables based on the Oracle database to store and manage welding records and corresponding welding information. Each welding record has a unique welding record identifier, which facilitates improving query performance and problem playback and tracing.

[0012] As a further technical solution of the present invention, step S2 includes model selection, data processing, model training and feature weight generation;

[0013] Model selection:

[0014] The random forest algorithm is used to model thermite welding process data and weld quality. The random forest algorithm assigns higher weight coefficients to key indicators such as gas pressure and preheating time, providing feature importance assessment.

[0015] Data processing:

[0016] A dataset was created based on historical thermite welding process data and the following processing was performed:

[0017] Continuous variables were standardized to ensure that the data were compared on a uniform scale;

[0018] Perform one-hot encoding on categorical variables to facilitate quantitative evaluation indicators;

[0019] Model training:

[0020] Five-fold cross validation was used to divide the data into five equal parts, four of which were selected as training sets and the remaining one as the test set.

[0021] A grid search strategy was used to optimize the random forest hyperparameters, including the number of decisions, the depth of the decision tree, and the minimum number of samples required for node splitting.

[0022] Feature weight generation:

[0023] The importance of factors is evaluated by calculating the split contribution of different factors in the decision number through information gain, and the weight ratio of each factor is obtained by normalization to obtain the weight matrix.

[0024] As a further technical solution of the present invention, step S4 includes:

[0025] Based on the weight matrix of the influence degree of each factor, according to the regulations for thermite welding quality evaluation, and combined with expert experience, a deduction mechanism is determined when each factor does not meet the standard;

[0026] The full score for thermite welding evaluation is set to 100 points. Based on the normalized and adjusted weight matrix, the corresponding deduction points when each operation is not satisfied are determined to quantitatively evaluate the weld quality.

[0027] Monitor the quality of thermite welding based on the quality evaluation scores, and focus on welds with hidden dangers.

[0028] As a further technical solution of the present invention, step S4 also includes result output and visualization:

[0029] Provides a thermite welding seam evaluation report, including quality evaluation results: welds are divided into high, medium, and low quality grades, and low-quality welds that require special attention are listed; major deduction items: The main deduction items for welds in different regions or teams are displayed to facilitate subsequent welding process optimization and welding management;

[0030] Provide visual data reports to show the comparison of weld quality at different times, regions, and teams, analyze welding operation rules, and provide a scientific basis for operation and maintenance decisions.

[0031] Beneficial effects achieved by the present invention:

[0032] On the technical level, by comprehensively considering various factors in the thermite welding process and combining the ability of the random forest model to capture nonlinear relationships, we can explore the mapping relationship between various factors in the welding process and defects in the welds, quantify the weight of the influence of each factor on the weld quality, and optimize the evaluation system based on expert experience to achieve real-time monitoring and early warning of the welding process, which can promote the intelligent development of rail welding technology.

[0033] In terms of economic benefits, this invention provides a strong basis for improving the quality of thermite welding seams and preventing and controlling risks, playing a significant role in enhancing the competence of welding teams and ensuring welding quality. Through a scientific weld quality evaluation system, welding processes and resource allocation can be optimized, reducing repair and replacement costs. Real-time weld quality evaluation can promptly identify potential problems, mitigate safety risks, and reduce unexpected large expenditures, supporting long-term planning for railway construction and maintenance and significantly improving economic benefits.

[0034] In terms of social benefits, the method of the present invention helps improve railway operation safety. By monitoring various operations during the welding process, it can optimize the welding process and improve weld quality, reduce safety accidents caused by weld defects, and ensure the safety and stability of train operations.

[0035] In summary, the rail thermite welding weld quality evaluation method based on machine learning and multi-factor monitoring proposed in the present invention can effectively improve the scientificity and practicality of rail welding quality evaluation, provide strong support for railway construction and maintenance, and has broad application prospects and social value. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] Figure 1 This is a flow chart of a rail thermite welding seam quality evaluation method based on machine learning and multi-factor monitoring provided by the present invention. DETAILED DESCRIPTION

[0037] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Therefore, the following detailed description of the embodiments of the present invention provided in the drawings is not intended to limit the scope of the invention claimed for protection, but merely represents selected embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0038] In the description of the present invention, the terms "first" and "second" are used for descriptive purposes only and should not be understood to indicate or imply relative importance or implicitly specify the number of the technical features indicated. Therefore, a feature specified as "first" or "second" may explicitly or implicitly include one or more of the features. In the description of the present invention, "plurality" means two or more, unless otherwise specifically defined.

[0039] See also Figure 1 The embodiment of the present invention provides a method for evaluating the quality of rail thermite welding seams based on machine learning and multi-factor monitoring, the method comprising:

[0040] S1. Collecting thermite welding process data and storing the collected data in the on-site welding management system. The thermite welding process data includes multiple factors such as weather, rail temperature, rail gap, gas pressure, preheating time, and quiet time during the thermite welding process;

[0041] S2. Based on the collected data, a random forest (RF) algorithm was used to construct a thermite welding quality evaluation model. The model was trained using historical thermite welding process data to establish a mapping relationship between factors such as weather, rail temperature, rail gap, gas pressure, preheating time, and quiet time during the welding process and weld defects. The importance of each factor was evaluated by calculating its split contribution (such as the Gini index or information gain) in the decision tree. The degree of dominance of each factor on weld defects was determined, thereby determining the influence weight of each factor.

[0042] S3. Based on the weight matrix of the degree of influence of each factor, in accordance with the regulations on thermite welding quality evaluation, and combined with expert experience, determine the deduction mechanism when each factor does not meet the standard. For example, for key indicators such as the flux does not match the rail to be welded, the gas pressure does not meet the standard, the preheating time does not meet the standard, the mold is removed or the nodules are pushed out in advance, and the rust removal of the rail ends is not carried out in accordance with the operating instructions, the deduction ratio should be increased by 50% to establish a differentiated distinction from other factors and build a scientific weld quality evaluation system.

[0043] In step S1 of this embodiment, the collected data is stored in the on-site welding management system. The system creates multiple data tables based on the Oracle database to store and manage welding records and corresponding welding information. Each welding record has a unique welding record identifier, which facilitates improving query performance and problem playback and tracing.

[0044] In this embodiment, step S2 includes model selection, data processing, model training and feature weight generation;

[0045] Model selection:

[0046] The Random Forest algorithm is used to model thermite welding process data and weld quality. The Random Forest algorithm can process high-dimensional data, capture nonlinear relationships, and assign higher weights to key indicators such as gas pressure and preheating time. (After normalization, the weights for indicators such as weather, rail end grinding, and rail gap distance are approximately 1-1.5, while the weights for key indicators such as gas pressure and preheating time are approximately 1.5-2.5.) This provides feature importance assessment.

[0047] Data processing:

[0048] A dataset was created based on historical thermite welding process data and the following processing was performed:

[0049] Standardize continuous variables (such as track temperature and rail gap) (Z-Score standardization) to ensure that the data can be compared on a unified scale;

[0050] One-hot encoding is used for categorical variables (such as gas type and weather) to facilitate quantitative evaluation indicators.

[0051] Model training:

[0052] Five-fold cross validation was used to divide the data into five equal parts, four of which were selected as training sets and the remaining one as a test set to ensure stable and reliable results.

[0053] A grid search strategy was used to optimize the random forest hyperparameters, including the number of decisions, the depth of the decision tree, and the minimum number of samples required for node splitting.

[0054] Feature weight generation:

[0055] The importance of factors is evaluated by calculating the split contribution of different factors in the decision number through information gain, and the weight ratio of each factor is obtained by normalization to obtain the weight matrix.

[0056] In this embodiment, step S4 includes:

[0057] Based on the weight matrix of the influence of each factor, in accordance with the regulations for thermite welding quality evaluation, and combined with expert experience, a deduction mechanism will be determined when each factor does not meet the standard. For example, for key indicators such as the mismatch between the flux and the rail to be welded, the non-standard gas pressure, the non-standard preheating time, the premature removal of the mold or the pushing of the nodules, and the failure to perform rail end rust removal in accordance with the operating instructions, the deduction rate should be increased by 50% to establish differentiation from other factors;

[0058] The full score for thermite welding evaluation is set to 100 points. Based on the normalized and adjusted weight matrix, the corresponding deduction points when each operation is not satisfied are determined to quantitatively evaluate the weld quality.

[0059] Monitor the quality of thermite welding based on the quality evaluation scores, and focus on welds with hidden dangers.

[0060] Step S4 of this embodiment also includes result output and visualization:

[0061] Provides a detailed evaluation report on thermite welding seams, including quality evaluation results: welds are divided into high, medium, and low quality grades, and low-quality welds that require special attention are listed; major deduction items: The main deduction items for welds in different regions or teams are displayed to facilitate subsequent welding process optimization and welding management;

[0062] Provide visual data reports to show the comparison of weld quality at different times, regions, and teams, analyze welding operation rules, and provide a scientific basis for operation and maintenance decisions.

[0063] It should be noted that, in this document, the term "comprises" or any other variant thereof is intended to cover non-exclusive inclusion, so that a process, method, article, or apparatus that includes a series of elements includes not only those elements, but also includes other elements not explicitly listed, or also includes elements inherent to such process, method, article, or apparatus. In the absence of further limitations, an element defined by the phrase "comprises a ..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes the element.

[0064] The above are only preferred embodiments of the present invention and are not intended to limit the patent scope of the present invention. Any equivalent structure or equivalent process transformation made using the contents of the present invention description and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present invention.

Claims

1. A method for evaluating rail thermite welding seam quality based on machine learning and multi-factor monitoring, characterized in that: The method comprises: S1. Collect thermite welding process data and store the collected data in the on-site welding management system; S2. Based on the collected data, a random forest algorithm was used to construct a thermite welding seam quality evaluation model. The model was trained using historical thermite welding process data to establish a mapping relationship between factors in the welding process and weld defects. Factors included weather, rail temperature, rail gap, gas pressure, preheating time, and quiet time during welding. The importance of each factor was evaluated by calculating its split contribution in the decision tree, determining its degree of dominance over weld defects and thus determining the influence weight of each factor. S3. Based on the weight matrix of the influence degree of each factor, according to the aluminothermic welding quality evaluation regulations, combined with expert experience, determine the deduction mechanism when each factor does not meet the standard, and build a scientific weld quality evaluation system.

2. The method for evaluating rail thermite welding seam quality based on machine learning and multi-factor monitoring according to claim 1, characterized in that: In step S1, the collected data is stored in the on-site welding management system. The system creates multiple data tables based on the Oracle database to store and manage welding records and corresponding welding information. Each welding record has a unique welding record identifier, which facilitates improving query performance and problem playback and tracing.

3. The method for evaluating rail thermite welding seam quality based on machine learning and multi-factor monitoring according to claim 1, characterized in that: Step S2 includes model selection, data processing, model training and feature weight generation; Model selection: The random forest algorithm is used to model thermite welding process data and weld quality. The random forest algorithm assigns higher weight coefficients to key factors and provides feature importance assessment. Key factors include gas pressure and preheating time. Data processing: A dataset was created based on historical thermite welding process data and the following processing was performed: Continuous variables were standardized to ensure that the data were compared on a uniform scale; Perform one-hot encoding on categorical variables to facilitate quantitative evaluation indicators; Model training: Five-fold cross validation was used to divide the data into five equal parts, four of which were selected as training sets and the remaining one as the test set. A grid search strategy was used to optimize the random forest hyperparameters, including the number of decisions, the depth of the decision tree, and the minimum number of samples required for node splitting. Feature weight generation: The importance of factors is evaluated by calculating the split contribution of different factors in the decision number through information gain, and the weight ratio of each factor is obtained by normalization to obtain the weight matrix.

4. The method for evaluating rail thermite welding seam quality based on machine learning and multi-factor monitoring according to claim 1, characterized in that: Step S4 includes: Based on the weight matrix of the influence degree of each factor, according to the regulations for thermite welding quality evaluation, and combined with expert experience, a deduction mechanism is determined when each factor does not meet the standard; The full score for thermite welding evaluation is set to 100 points. Based on the normalized and adjusted weight matrix, the corresponding deduction points when each operation is not satisfied are determined to quantitatively evaluate the weld quality. Monitor the quality of thermite welding based on the quality evaluation scores, and focus on welds with hidden dangers.

5. The method for evaluating rail thermite welding seam quality based on machine learning and multi-factor monitoring according to claim 1, characterized in that: Step S4 also includes result output and visualization: Provides a thermite welding seam evaluation report, including quality evaluation results: welds are divided into high, medium, and low quality grades, and low-quality welds that require special attention are listed; major deduction items: The main deduction items for welds in different regions or teams are displayed to facilitate subsequent welding process optimization and welding management; Provide visual data reports to show the comparison of weld quality at different times, regions, and teams, analyze welding operation rules, and provide a scientific basis for operation and maintenance decisions.

Images