Copper bar welding reliability temperature rise verification method

Abstract

The invention discloses a temperature rise verification method for welding reliability of a copper bar, relates to the technical field of copper bars, and provides the following scheme aiming at the problems that potential and developing hidden defects proposed in the background technology cannot be identified and potential safety hazards are buried due to the fact that obvious temperature abrupt change is possibly not caused in the test period. Comprising the steps of sample selection and pretreatment: selecting a to-be-tested copper bar welding part, performing nondestructive testing (such as ultrasonic testing or X-ray testing) on the sample, observing defects such as air holes, slag inclusion, cracks and incomplete penetration, removing obviously unqualified samples, and pretreating the sample, including surface cleaning and stress relief annealing. According to the invention, the artificial intelligence technology is deeply integrated into the traditional temperature rise test process, so that the characteristic mode can be identified when a tiny hot spot is just formed and the overall temperature rise is not caused to exceed the standard, the early intelligent diagnosis of hidden defects is realized, the fault hidden danger is eliminated in the bud state, and the accuracy of the detection result is improved at the same time.

Description

Technical Field

[0001] This invention relates to the field of copper busbar technology, and in particular to a method for verifying the temperature rise of copper busbar welding reliability. Background Technology

[0002] Power busbars are made of copper or aluminum and are long conductors with a rectangular or chamfered (rounded) rectangular cross-section. Busbars made of aluminum are called aluminum busbars (also known as aluminum busbars, aluminum busbars, or grounding aluminum busbars), while those made of copper are called copper busbars (also known as copper busbars, copper busbars, or grounding copper busbars). They are used in circuits to transmit current and connect electrical equipment.

[0003] As a major component of generators, the quality of welding between copper busbars directly affects the safe and stable operation of the entire generator. If the welding is unreliable, the weld joint is prone to overheating due to excessive contact resistance during long-term energized operation, which can lead to generator failure or even safety accidents. Therefore, it is crucial to effectively test the reliability of copper busbar welding.

[0004] Temperature rise testing is one of the important means to evaluate the reliability of electrical connectors. By testing the temperature change of the connector under energized conditions, its conductivity and connection reliability can be reflected. Currently, most existing copper busbar temperature rise testing methods focus on the overall heating test of the copper busbar. It can only be discovered when the defect has become so serious that it causes obvious and stable temperature rise anomalies. For potential and developing latent defects (such as early micro cracks), they may not have caused significant temperature changes during the test, so they cannot be identified and may pose safety hazards. Summary of the Invention

[0005] This invention provides a method for verifying the temperature rise reliability of copper busbar welding, which solves the problem in the prior art that potential, developing latent defects (such as initial microcracks) may not have caused significant temperature changes during the test, thus failing to be identified and creating potential safety hazards.

[0006] To achieve the above objectives, the present invention adopts the following technical solution: A method for verifying the reliability of copper busbar welding by temperature rise includes the following steps: S1: Sample selection and pretreatment: Select the copper busbar welded parts to be tested, perform non-destructive testing on the samples (such as ultrasonic testing or X-ray testing), observe defects such as porosity, slag inclusion, cracks, and incomplete penetration, and remove obviously unqualified samples. Sample pretreatment includes surface cleaning (using alcohol or ultrasonic cleaning) and stress-relieving annealing. S2: Sample Installation: Qualified samples are fixed on the test platform, which includes a power module, temperature control module, data acquisition module (temperature sensor and data logger), infrared thermal imager, fixing device, and intelligent analysis center. The fixing device is designed to simulate the actual installation state and can apply controllable mechanical stress (e.g., applying a tightening torque of 5-10 N·m using a torque wrench) to ensure that the stress state of the copper busbar welded parts is consistent with the actual situation. The sensor configuration includes thermocouple sensors at the center of the weld, and sensors at non-welded locations 5cm and 10cm away from the weld. The infrared thermal imager is used to monitor the temperature distribution of the entire copper busbar in real time and assist in verifying the temperature measurement data at the verification points. The intelligent analysis center includes a data interface module that receives massive amounts of data from all thermocouple sensors and the infrared thermal imager in real time; a feature engineering module that extracts key features from the raw data; an AI core model library with built-in trained machine learning models; and a decision and output module that generates predictions, diagnostic reports, and judgment results. The AI ​​core model library includes a combination of Long Short-Term Memory Network (LSTM) and Convolutional Neural Network (CNN) with Support Vector Machine (SVM). S3: Parameter Setting: Set the output current and voltage based on the rated operating current (I_rated) of the copper busbar. The output current is in a multi-level loading mode: the first level is the rated current (1.0×I_rated) for 30 minutes; the second level is the overload current (1.2×I_rated) for 60 minutes; the third level is the transient overcurrent (1.5×I_rated, lasting 5 minutes), simulating actual fault conditions. The voltage parameters match the current to ensure the normal operation of the copper busbar. S4: Intelligent Temperature Rise Test: Power on and temperature control module are activated to run the copper busbar welding components under set conditions. Data acquisition: Temperature sensors collect data in real time at a sampling frequency of once per minute; Infrared thermal imager collects full-image data every 5 minutes. Dynamic testing phase: After the temperature stabilizes (defined as a temperature change ≤1K / 10 minutes as stable), at least 3 temperature cycles are performed (each cycle includes heating and cooling phases), with a total test duration of no less than 4 hours. Sensor calibration: Before testing, all temperature sensors are calibrated in a standard constant temperature bath with an error not exceeding ±0.5℃. Test monitoring: Temperature data is monitored in real time. S5: Real-time prediction and real-time diagnosis: Data flows into the intelligent analysis center in real time for real-time prediction. The LSTM model runs every 5 minutes and outputs a prediction of the future temperature. If the predicted final stable temperature will exceed the safety threshold, the system will issue an early alarm. Real-time diagnosis: The CNN model analyzes the infrared thermal image in real time. Once an abnormal temperature concentration area is detected (such as a "crescent-shaped" hot spot, which may indicate that the edge is not fully soldered), the system will highlight it on the operation interface and indicate the suspected defect type. S6: Data Recording and In-Depth Analysis: After the test, the system automatically generates a multi-dimensional analysis report, including a prediction accuracy report, a comparison of the AI ​​prediction curve with the actual measurement curve, an assessment of model confidence, a health status score (based on all features, the AI ​​gives a comprehensive health score of 0-100), a defect probability analysis (giving the probability of various defects such as "porosity" and "cracks" at the weld point), and a lifespan prediction (the AI ​​model combines the temperature rise data of this test (especially the performance under overload conditions) with the material aging model to output an "expected lifespan" (e.g., under rated operating conditions, the reliable operating time of this weld point is expected to be >15 years). S7: Intelligent Reliability Judgment: Based on AI comprehensive judgment, qualified, AI health status score > 85 points, all defect probabilities < 5%, and predicted life meets design requirements; critical warning, health score between 70-85 points, probability of a certain defect between 5%-15%, it is recommended to focus on monitoring or conduct re-inspection; unqualified, health score < 70 points, or predicted life does not meet the standard, or a high probability of serious defects is found.

[0007] Preferably, the stress-relief annealing temperature in S1 is 200℃±10℃, held for 1 hour, and then cooled in the furnace to eliminate the influence of processing stress on the test.

[0008] Preferably, the Long Short-Term Memory (LSTM) network processes time-series data and can accurately predict subsequent temperature changes and the final stable value based on the previous temperature rise trend; the Convolutional Neural Network (CNN) is used to analyze the two-dimensional temperature distribution map provided by the infrared thermal imager and identify abnormal hot spot patterns; and the Support Vector Machine (SVM) is used to classify the extracted features into defects (such as normal, porosity, incomplete penetration, and cracks).

[0009] Preferably, the data input to the AI ​​model by the feature engineering module in S2 includes the original temperature value and the deep features extracted from it, such as temporal features, temperature rise rate, second derivative of temperature curve, local fluctuation variance, and time required to reach a specific temperature; spatial features, maximum temperature difference between welded and non-welded areas, standard deviation of temperature distribution, and direction of temperature gradient in thermal image; and comprehensive features, temperature response sensitivity under different combinations of current load and ambient temperature.

[0010] Preferably, the model training data source for the intelligent analysis center in S2 is the temperature rise test conducted on a large number of copper busbar welded parts in a laboratory with known conditions (including samples with various artificially created defects and qualified samples), forming a labeled training dataset. The training objective is to enable the Long Short-Term Memory Network (LSTM) model to learn to predict the temperature rise curve and stable temperature for the next 120 minutes based on the data from the previous 60 minutes, and to enable the Convolutional Neural Network (CNN) combined with the Support Vector Machine (SVM) model to learn to associate specific temperature distribution patterns with specific welding defect types.

[0011] Preferably, in step S4, when the temperature at the welding point suddenly rises by more than 10K / minute, a safety protection mechanism (such as cutting off the power supply) is automatically triggered.

[0012] The beneficial effects of this invention are as follows: 1. By deploying time-series prediction models such as LSTM, the final stable temperature and temperature rise curve can be accurately predicted based on the temperature rise dynamics in the early stages of testing. This allows potential defective products to be identified before testing is completed, reducing the judgment time by an average of 30%-50%, greatly improving quality inspection efficiency, and is suitable for rapid screening on the production line. At the same time, by using CNN models to perform real-time analysis of infrared thermal images, it is possible to identify the characteristic patterns of tiny hot spots when they have just formed and before they cause the overall temperature rise to exceed the standard, realizing early intelligent diagnosis of hidden defects and eliminating potential faults in their infancy.

[0013] 2. An AI correlation model was established between short-term temperature rise test data and long-term operational reliability. This model can specifically test the temperature rise of copper busbar welds. By comparing the temperature data of welded and non-welded parts, it can accurately reflect the impact of welding quality on the heating of the copper busbar. Based on accelerated test data, the expected service life of the copper busbar welds under rated operating conditions can be extrapolated. By fusing multi-sensor data (point temperature and thermal image) and using AI models for noise reduction and pattern recognition, misjudgments caused by poor sensor contact or random fluctuations at a single measuring point are effectively avoided, making the test results more accurate.

[0014] In summary, this invention, by deeply integrating artificial intelligence technology into the traditional temperature rise testing process, can identify the characteristic patterns of tiny hot spots as soon as they form and before they cause the overall temperature rise to exceed the standard. This enables early intelligent diagnosis of hidden defects, eliminates potential faults in their infancy, and improves the accuracy of the test results. Attached Figure Description

[0015] Figure 1 This is a schematic diagram of the process structure of a method for verifying the reliability of copper busbar welding temperature rise proposed in this invention. Detailed Implementation

[0016] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments.

[0017] Example 1, referring to Figure 1 A method for verifying the reliability of copper busbar welding by temperature rise includes the following steps: S1: Sample selection and pretreatment: Select the copper busbar welded parts to be tested, perform non-destructive testing on the samples (such as ultrasonic testing or X-ray testing), observe defects such as porosity, slag inclusion, cracks, and incomplete penetration, and remove obviously unqualified samples. Sample pretreatment includes surface cleaning (using alcohol or ultrasonic cleaning) and stress relief annealing. The stress relief annealing temperature is 200℃±10℃, held for 1 hour, and cooled with the furnace to eliminate the influence of processing stress on the test. S2: Sample Installation: Secure the qualified sample onto the test platform. The platform includes a power module, temperature control module, data acquisition module (temperature sensor and data logger), infrared thermal imager, fixing device, and intelligent analysis center. The fixing device is designed to simulate actual installation conditions and can apply controllable mechanical stress (e.g., 5-10 g using a torque wrench). (Tightening torque of N·m) to ensure the stress state of the copper busbar welded parts is consistent with the actual situation. Sensor configuration: Thermocouple sensors are used for temperature, with one sensor each at the center of the weld, 5cm and 10cm away from the weld in the non-welded area. Infrared thermal imager is used to monitor the temperature distribution of the entire copper busbar in real time and to assist in verifying the temperature measurement data at the verification points. The intelligent analysis center includes a data interface module that receives massive amounts of data from all thermocouple sensors and infrared thermal imager in real time. Feature engineering module extracts key features from the raw data. The data input to the AI ​​model for feature engineering module includes the raw temperature value and the deep features extracted from it, such as temporal features (temperature rise rate, second derivative of temperature curve, local fluctuation variance, time required to reach a specific temperature); spatial features (maximum temperature difference between welded and non-welded areas, standard deviation of temperature distribution, temperature gradient direction in thermal image); and comprehensive features (temperature response sensitivity under different current load and ambient temperature combinations). AI core model library: Built-in trained machine learning models. Decision and output module: Generates predictions, diagnostic reports and judgment results. The core AI model library includes a Long Short-Term Memory (LSTM) network and a Convolutional Neural Network (CNN) combined with a Support Vector Machine (SVM). The LSTM network processes time-series data and can accurately predict subsequent temperature changes and the final stable value based on the previous temperature rise trend. The CNN network is used to analyze the two-dimensional temperature distribution map provided by the infrared thermal imager and identify abnormal hot spot patterns. The SVM network is used to classify the extracted features into defects (such as normal, porosity, incomplete penetration, and cracks). The model training data of the intelligent analysis center comes from temperature rise tests on a large number of copper busbar welded parts in known states (including samples with various artificially created defects and qualified samples) in the laboratory, forming a labeled training dataset. The training objective is to enable the LSTM network model to predict the temperature rise curve and stable temperature for the next 120 minutes based on the data of the previous 60 minutes, and to enable the CNN network combined with the SVM network model to associate specific temperature distribution patterns with specific welding defect types. S3: Parameter Setting: Set the output current and voltage based on the rated operating current (I_rated) of the copper busbar. The output current is in a multi-level loading mode: the first level is the rated current (1.0×I_rated) for 30 minutes; the second level is the overload current (1.2×I_rated) for 60 minutes; the third level is the transient overcurrent (1.5×I_rated, lasting 5 minutes), simulating actual fault conditions. The voltage parameters match the current to ensure the normal operation of the copper busbar. S4: Intelligent Temperature Rise Test: Power on and temperature control module are activated to run the copper busbar welding components under set conditions. Data acquisition: Temperature sensors collect data in real time at a sampling frequency of once per minute; Infrared thermal imager collects full-image data every 5 minutes. Dynamic testing phase: After the temperature stabilizes (defined as a temperature change ≤1K / 10 minutes as stable), at least 3 temperature cycles are performed (each cycle includes heating and cooling phases), with a total test duration of no less than 4 hours. Sensor calibration: Before testing, all temperature sensors are calibrated in a standard constant temperature bath with an error not exceeding ±0.5℃. Test monitoring: Temperature data is monitored in real time. When the temperature at the welding point suddenly rises by more than 10K / minute, safety protection (such as power cut-off) is automatically triggered. S5: Real-time prediction and real-time diagnosis: Data flows into the intelligent analysis center in real time for real-time prediction. The LSTM model runs every 5 minutes and outputs a prediction of the future temperature. If the predicted final stable temperature will exceed the safety threshold, the system will issue an early alarm. Real-time diagnosis: The CNN model analyzes the infrared thermal image in real time. Once an abnormal temperature concentration area is detected (such as a "crescent-shaped" hot spot, which may indicate that the edge is not fully soldered), the system will highlight it on the operation interface and indicate the suspected defect type. S6: Data Recording and In-Depth Analysis: After the test, the system automatically generates a multi-dimensional analysis report, including a prediction accuracy report, a comparison of the AI ​​prediction curve with the actual measurement curve, an assessment of model confidence, a health status score (based on all features, the AI ​​gives a comprehensive health score of 0-100), a defect probability analysis (giving the probability of various defects such as "porosity" and "cracks" at the weld point), and a lifespan prediction (the AI ​​model combines the temperature rise data of this test (especially the performance under overload conditions) with the material aging model to output an "expected lifespan" (e.g., under rated operating conditions, the reliable operating time of this weld point is expected to be >15 years). S7: Intelligent Reliability Judgment: Based on AI comprehensive judgment, qualified, AI health status score > 85 points, all defect probabilities < 5%, and predicted life meets design requirements; critical warning, health score between 70-85 points, probability of a certain defect between 5%-15%, it is recommended to focus on monitoring or conduct re-inspection; unqualified, health score < 70 points, or predicted life does not meet the standard, or a high probability of serious defects is found.

[0018] By deploying time-series prediction models such as LSTM, the final stable temperature and temperature rise curve can be accurately predicted based on the temperature rise dynamics in the early stages of testing. This allows potential non-conforming products to be identified before testing is completed, reducing the judgment time by an average of 30%-50%, greatly improving quality inspection efficiency, and is suitable for rapid screening on the production line. At the same time, by using CNN models to perform real-time analysis of infrared thermal images, it is possible to identify the characteristic patterns of tiny hot spots when they are just formed and before they cause the overall temperature rise to exceed the standard, realizing early intelligent diagnosis of hidden defects and eliminating potential faults in their infancy. An AI correlation model between short-term temperature rise test data and long-term operational reliability has been established, which can specifically perform temperature rise tests on copper busbar welds. By comparing the temperature data of welded and non-welded parts, it can accurately reflect the impact of welding quality on the heating of copper busbars. Based on accelerated test data, the expected service life of copper busbar welds under rated operating conditions can be extrapolated. By fusing multi-sensor data (spot temperature and thermal images) and using AI models for noise reduction and pattern recognition.

[0019] In the description of this invention, it should be understood that the terms "center," "longitudinal," "lateral," "length," "width," "thickness," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," "outer," "clockwise," and "counterclockwise," etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are only for the convenience of describing this invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on this invention.

[0020] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature. In the description of this invention, "a plurality of" means two or more, unless otherwise explicitly specified.

[0021] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A method for verifying the reliability of copper bar welding by temperature rise, characterized in that, Comprising the following steps: S1: Sample selection and pretreatment: Select the copper bar welding parts to be tested, and perform non-destructive testing (such as ultrasonic testing or X-ray testing) on the sample to observe defects such as pores, slag inclusions, cracks, and incomplete penetration, and remove obviously unqualified samples. The sample is pretreated, including surface cleaning (using alcohol or ultrasonic cleaning) and stress relief annealing; S2: Sample installation: Fix the qualified sample on the test platform, which includes a power module, a temperature control module, a data acquisition module (temperature sensor and data logger), an infrared thermal imager, a fixing device, and an intelligent analysis hub. The fixing device is designed to simulate the actual installation state, and the fixing device can apply controllable mechanical stress (such as applying a tightening torque of 5-10 N·m through a torque wrench) to ensure that the stress state of the copper bar welding part is consistent with the actual state. The sensor configuration uses thermocouple sensors at the center of the weld, 5 cm and 10 cm away from the weld, and one sensor at each non-welding position. The infrared thermal imager is used to monitor the temperature distribution of the entire copper bar in real time, and to assist in verifying the point temperature data. The intelligent analysis hub includes a data interface module that receives massive data from all thermocouple sensors and infrared thermal imagers in real time, a feature engineering module that extracts key features from raw data, an AI core model library that includes trained machine learning models, a decision and output module that generates prediction, diagnosis reports and decision results. The AI core model library includes a long short-term memory network (LSTM) and a convolutional neural network (CNN) combined with a support vector machine (SVM); S3: Parameter setting: Set the output current and voltage based on the rated working current (I_rated) of the copper bar. The output current is in multi-level loading mode: the first level is rated current (1.0×I_rated) running for 30 minutes; the second level is overload current (1.2×I_rated) running for 60 minutes; the third level is transient overcurrent (1.5×I_rated, lasting for 5 minutes), simulating actual fault conditions, and the voltage parameter matches the current to ensure the normal operation of the copper bar; S4: Intelligent temperature rise test: Start the power supply and temperature control module to make the copper bar welding part run under the set conditions. Data acquisition: temperature sensors collect data in real time with a sampling frequency of once per minute; the infrared thermal imager collects full image data every 5 minutes. In the dynamic test phase, after the temperature stabilizes (defined as a temperature change ≤1K / 10 minutes), at least 3 temperature cycle periods (each period includes a heating and cooling phase) are performed, and the total test duration is not less than 4 hours. Sensor calibration: before testing, all temperature sensors are calibrated in a standard constant temperature bath with an error of not more than ±0.5℃. Test monitoring: real-time monitoring of temperature data; S5: Real-time prediction and real-time diagnosis: data flows into the intelligent analysis hub in real time, real-time prediction, LSTM model runs every 5 minutes, outputs the prediction of future temperature, if the predicted final stable temperature will exceed the safety threshold, the system will give an early warning, real-time diagnosis, CNN model analyzes infrared thermal image in real time, once detects abnormal temperature concentration area (such as "crescent" hot spot may indicate edge under-penetration), the system highlights and prompts the suspected defect type on the operation interface; S6: Data recording and deep analysis: after the test is completed, the system automatically generates a multi-dimensional analysis report, including prediction accuracy report, comparison of AI prediction curve and actual measurement curve, evaluation of model confidence, health status score, based on all features, AI gives a comprehensive health score of 0-100, defect probability analysis, gives the probability of existence of "porosity", "crack" and other defects in this welding point, life prediction, AI model combines the temperature rise data of this test (especially the performance under overload condition) with the material aging model, outputs a "expected service life" (for example, this welding point is expected to run reliably for >15 years under rated operating conditions); S7: Intelligent reliability judgment: based on AI comprehensive judgment, qualified, AI health status score > 85, all defect probabilities < 5%, and predicted life meets design requirements; critical warning, health score between 70-85, certain defect probability between 5%-15%, suggest intensive monitoring or retest; unqualified, health score < 70, or predicted life does not meet the requirements, or high probability of serious defects is found.

2. The copper bar welding reliability temperature rise verification method of claim 1, wherein, The stress relief annealing temperature in S1 is 200℃±10℃, and the holding time is 1 hour, and the furnace is cooled to eliminate the influence of processing stress on the test.

3. The copper bar welding reliability temperature rise verification method of claim 1, wherein, The long short-term memory network (LSTM) processes time series data, can accurately predict the subsequent temperature change and final stable value according to the previous temperature rise trend, the convolutional neural network (CNN) is used to analyze the two-dimensional temperature distribution map provided by the infrared thermal imager to identify abnormal hot spot patterns; support vector machine (SVM) is used for defect classification (such as normal, porosity, under-penetration, crack) of extracted features.

4. The copper bar welding reliability temperature rise verification method of claim 1, wherein, The feature engineering module in S2 inputs data into the AI model, including original temperature values and deep features extracted therefrom, such as time series features, temperature rise rate, second derivative of temperature curve, local fluctuation variance, time required to reach a certain temperature; Spatial features, maximum temperature difference between welding area and non-welding area, standard deviation of temperature distribution, temperature gradient direction in thermal image; Comprehensive features, temperature response sensitivity under different current load and environmental temperature combinations.

5. The copper bar welding reliability temperature rise verification method of claim 1, wherein, The model training data source of the intelligent analysis hub in S2 is obtained by testing a large number of copper bar welding pieces with known states (including samples with various defects artificially manufactured and qualified samples) in a laboratory to form a labeled training data set, and the training target is to let the long short-term memory network (LSTM) model learn to predict the temperature rise curve and stable temperature in the next 120 minutes according to the data in the previous 60 minutes, and let the convolutional neural network (CNN) combined with the support vector machine (SVM) model learn to associate a specific temperature distribution pattern with a specific welding defect type.

6. The copper bar welding reliability temperature rise verification method of claim 1, wherein, When the temperature at the welding position suddenly rises by more than 10K / min in S4, the safety protection (such as cutting off the power supply) is automatically triggered.

Images