Electric welding machine remote monitoring and running state diagnosis method and system

Abstract

The invention relates to the technical field of welding equipment monitoring, in particular to a remote monitoring and running state diagnosis method and system for an electric welding machine, and the method comprises the steps: synchronously collecting multi-dimensional parameters of the electric welding machine during working; carrying out data fusion processing on the collected multi-dimensional parameters; inputting the extracted features into a fault diagnosis model trained based on a machine learning algorithm for operation state diagnosis; transmitting the diagnosis result, the data acquired in real time and a digital twinborn body which is constructed on the basis of the data and reflects the real-time state of the electric welding machine to a remote monitoring end for display and monitoring, and receiving a parameter adjustment instruction from the remote monitoring end when the diagnosis result indicates abnormity or fault; and operation parameters of the electric welding machine are correspondingly adjusted. And remote monitoring and intelligent adjusting functions are introduced, so that the intelligent level of the electric welding machine is further enhanced, the requirement of human intervention is reduced, and the welding operation is more efficient and safer.

Description

Technical Field

[0001] This invention relates to the field of welding equipment monitoring technology, and in particular to a method and system for remote monitoring and operational status diagnosis of electric welding machines. Background Technology

[0002] Electric welding machines, as indispensable key equipment in modern manufacturing, are widely used in many important industrial fields such as machinery manufacturing, shipbuilding, bridge construction, and automobile repair. Their reliability and stability directly affect production efficiency and product quality. With the continuous development of industrial automation and intelligent technologies, remote monitoring and operational status diagnosis technologies for electric welding machines are gradually becoming important means to improve welding quality management.

[0003] However, existing remote monitoring systems for welding machines still have several technical limitations that urgently need to be addressed in practical applications. These problems seriously restrict their further promotion and full realization of their effectiveness:

[0004] Data acquisition is often limited to a single dimension, lacking simultaneous monitoring of multiple parameters: Most existing systems primarily collect basic electrical parameters such as voltage and current. However, welding quality is also significantly affected by environmental parameters (such as temperature and humidity) and equipment mechanical parameters (such as wire feed speed and welding angle). Currently, there is a lack of technical means to simultaneously and comprehensively collect these multi-dimensional parameters, resulting in an inability to provide a complete data foundation for subsequent in-depth analysis and diagnosis. For example, abnormal fluctuations in ambient humidity can lead to porosity in the weld, while instability in wire feed speed directly affects the weld formation quality; monitoring a single electrical parameter is insufficient to capture these defects caused by complex factors.

[0005] The lack of multi-parameter coupling analysis leads to low accuracy in fault diagnosis: Traditional monitoring methods struggle with multi-parameter coupling analysis due to the failure to establish correlation models between electrical, mechanical, and environmental parameters. Existing technologies typically only perform independent threshold judgments or simple comparisons of each parameter, failing to delve into the intrinsic relationships and interactions between them. This results in low accuracy in diagnosing equipment operating status, frequent false alarms and missed alarms, and an inability to predict and warn of potential faults early. Welding is a complex process involving multiple coupled factors, and this lack of analytical capability often causes maintenance strategies to lag behind actual fault occurrences.

[0006] Insufficient remote diagnostic capabilities and limited intelligence: Many existing systems are primarily limited to remote data transmission and centralized display, essentially functioning as "data dashboards" rather than "intelligent diagnostic systems." They lack mechanisms for intelligent analysis, diagnosis, and proactive early warning based on real-time data. Equipment status assessment and fault handling still heavily rely on manual analysis by experienced back-end personnel, which is not only inefficient and slow to respond but also hinders the refined and automated management of large-scale equipment clusters. This prevents the full realization of the value of remote monitoring and fails to meet the urgent needs of modern intelligent manufacturing for predictive maintenance and unmanned operation.

[0007] Therefore, developing a remote monitoring and operational status diagnosis method for welding machines that can overcome the above limitations and achieve multi-dimensional data fusion and collection, intelligent analysis and diagnosis, and accurate early warning is of great necessity and urgency for ensuring welding production quality, improving equipment management efficiency, and promoting the evolution of welding operations from digitalization to intelligence. Summary of the Invention

[0008] To achieve the above objectives, the present invention provides a method and system for remote monitoring and operational status diagnosis of an electric welding machine. The method for remote monitoring and operational status diagnosis of an electric welding machine includes the following steps:

[0009] The system synchronously collects multi-dimensional parameters of the welding machine during operation, including electrical parameters, mechanical parameters, and environmental parameters.

[0010] The collected multi-dimensional parameters are fused, including time synchronization of data from different sensors using a unified timestamp, filtering of the synchronized data to eliminate noise, and then extracting time-domain features, frequency-domain features, and correlation features reflecting the relationship between parameters from the preprocessed data.

[0011] The extracted features are input into a fault diagnosis model trained based on machine learning algorithms for operational status diagnosis. The fault diagnosis model is trained using historical data of the welding machine under different known working conditions and can output diagnostic results representing the current health status of the welding machine.

[0012] The diagnostic results, real-time collected data, and a digital twin reflecting the real-time status of the welding machine, constructed based on these data, are transmitted to a remote monitoring terminal for display and monitoring. When the diagnostic results indicate an abnormality or fault, parameter adjustment instructions are received from the remote monitoring terminal to adjust the operating parameters of the welding machine accordingly.

[0013] Preferably, the specific process of synchronously acquiring multi-dimensional parameters includes: acquiring welding voltage and welding current through voltage sensors and current sensors, wherein the voltage signal is rectified and filtered, and the current signal is acquired and converted into a voltage signal through non-contact sensing.

[0014] The speed signals of the wire feeding motor and the traveling mechanism are collected by encoders installed on the wire feeding motor and the traveling mechanism, and the actual wire feeding speed and welding speed are calculated based on the gear radius of the motor.

[0015] The welding angle between the welding torch and the workpiece is collected using an angle sensor.

[0016] The ambient temperature, humidity, and carbon dioxide concentration data of the welding machine's working area are collected using temperature sensors, humidity sensors, and gas sensors, respectively.

[0017] All sensor data acquisition is synchronized through a unified timing control mechanism to ensure data consistency over time.

[0018] Preferably, the data fusion processing includes filtering the data by employing a recursive estimation-based filtering algorithm to smooth the welding voltage and welding current signals in order to suppress random noise.

[0019] For mechanical parameter data, a filtering algorithm based on median calculation is used to process the data to eliminate sudden interference.

[0020] For environmental parameter data, a filtering algorithm based on linear averaging is used to process the data in order to maintain its stability.

[0021] The feature extraction process includes: calculating the arithmetic mean, variance, peak value, and fluctuation coefficient of the welding voltage and welding current signals as time-domain features;

[0022] Spectral transformation is performed on welding voltage and welding current signals to extract the main oscillation frequency components and frequency band energy distribution as frequency domain features;

[0023] The synergistic relationship between welding current and wire feed speed, and the mutual influence between welding voltage and ambient temperature were analyzed, and their statistical correlation strength was calculated as a multi-parameter correlation feature.

[0024] Preferably, the training process of the fault diagnosis model includes: collecting historical multi-dimensional parameter data of the welding machine under normal conditions, multiple known abnormal conditions, and fault conditions to form a training sample set;

[0025] Feature extraction is performed on the data in the training sample set to form a corresponding set of feature vectors;

[0026] The model is trained using a machine learning algorithm capable of handling multi-classification problems. The internal parameters of the model are adjusted by optimizing the algorithm so that the model can accurately distinguish different working states.

[0027] The model's performance is evaluated using cross-validation, and the model's structural parameters are adjusted based on the evaluation results until the model reaches the predetermined performance indicators.

[0028] During real-time diagnosis, the fault diagnosis model calculates the confidence level of its diagnosis results. When the confidence level is lower than the threshold determined by statistical analysis of historical operation data, a collaborative diagnosis mechanism based on a combination of multiple machine learning algorithms is activated to verify and synthesize the diagnosis results.

[0029] Preferably, the process of constructing the digital twin includes: establishing a three-dimensional geometric model of the welding machine and a digital representation of its electrical and control systems;

[0030] The real-time collected electrical, mechanical, and environmental parameters are mapped onto the digital model, driving the corresponding state variables in the model to be updated, so that the digital twin can reflect the actual working state of the welding machine in real time, including the motion state of each component, the changing trend of electrical parameters, and the influence of environmental parameters.

[0031] The remote monitoring terminal displays the real-time operating parameter curves of the welding machine, a visual display of the operation status diagnostic results, a list of early warning information prompts, and a historical data query interface.

[0032] Preferably, the process of receiving parameter adjustment instructions and adjusting the operating parameters of the welding machine includes: the remote monitoring terminal generating control instructions to adjust the welding voltage, welding current or wire feeding speed based on the diagnostic results;

[0033] Control commands are transmitted to the local control system of the welding machine via a wireless communication network;

[0034] The local control system parses the instructions and converts them into drive signals, which drive the actuators to adjust the working parameters of the welding machine.

[0035] After the parameters are adjusted, the working data of the welding machine is collected again and transmitted to the remote monitoring terminal to verify the adjustment effect.

[0036] Preferably, the method further includes an online optimization process for the fault diagnosis model: collecting data on the differences between actual fault handling results and model diagnosis results;

[0037] When a model diagnosis error or low confidence is found, the correct state label and corresponding data feature are added to the training dataset.

[0038] An algorithm capable of incremental learning is adopted to periodically update the parameters of the fault diagnosis model using newly added training data, enabling the model to adapt to changes in the working state of the welding machine and new fault modes.

[0039] Preferably, during the process of synchronously acquiring multi-dimensional parameters, the acquisition of the welding angle is specifically achieved by an angle sensor installed on the welding torch, which measures the tilt angle of the welding torch in multiple axial directions.

[0040] Carbon dioxide concentration is collected using a gas sensor based on the principle of infrared absorption, which calculates the gas concentration by measuring the absorption intensity at a specific infrared wavelength.

[0041] Preferably, the machine learning algorithm used in the training of the fault diagnosis model is a multi-class support vector machine, and its kernel function parameters and penalty factors are optimized through a grid search strategy, that is, a system search is performed within a predefined range of parameter values, and the optimal parameter combination is selected based on the cross-validation results.

[0042] Accordingly, embodiments of the present invention provide a remote monitoring and operation status diagnosis system for an electric welding machine, including a memory configured to store instructions, a processor configured to call the instructions from the memory, and capable of implementing any of the remote monitoring and operation status diagnosis methods for an electric welding machine described in any of the embodiments of the present invention when executing the instructions.

[0043] The beneficial effects of this invention are:

[0044] 1. This invention overcomes the limitations of traditional systems that only monitor single parameters such as voltage and current by simultaneously collecting electrical, mechanical, and environmental parameters of the welding machine. In particular, the simultaneous acquisition technology of multi-dimensional parameters such as voltage, current, temperature, humidity, carbon dioxide concentration, and wire feed motor speed ensures that the data comprehensively and accurately reflects the working status of the welding machine. The comprehensive data collected in real time provides a reliable foundation for subsequent fault diagnosis and welding quality analysis, avoiding the potential impact of factors such as abnormal environmental humidity or unstable wire feed speed on welding quality. This multi-dimensional data acquisition provides strong support for in-depth analysis of equipment operating status and fault causes.

[0045] 2. This invention, through data fusion and multi-parameter correlation feature extraction, effectively uncovers the intrinsic relationships between electrical, mechanical, and environmental parameters. The fault diagnosis model trained using machine learning algorithms can deeply analyze the mutual influences between various parameters. This multi-parameter coupling analysis significantly improves the accuracy of fault diagnosis, avoids false alarms and missed alarms, and enables early warning of potential faults. This improvement significantly enhances the timeliness and accuracy of equipment maintenance.

[0046] 3. This invention overcomes the limitations of existing systems that only possess data transmission and display functions, further realizing an intelligent diagnosis and fault early warning mechanism based on real-time data. By combining real-time monitoring data and a machine learning-based fault diagnosis model, this invention can provide real-time status monitoring, diagnostic results, and early warning information for welding machines at the remote monitoring end, greatly improving the intelligence level of remote monitoring. When the system diagnoses abnormal equipment status or faults, it can proactively generate adjustment commands and adjust parameters through the remote control system. This intelligent function does not rely on manual analysis, has a fast response speed, and enables refined management of large-scale equipment clusters, effectively supporting the needs of intelligent manufacturing for predictive maintenance and unmanned operation. Attached Figure Description

[0047] To more clearly illustrate the technical solutions in this invention or the prior art, the accompanying drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, those skilled in the art can obtain other drawings based on these drawings without creative effort.

[0048] Figure 1 This is a flowchart of the steps of the method of the present invention;

[0049] Figure 2 This is a flowchart illustrating the steps involved in constructing a digital twin using the method of the present invention.

[0050] Figure 3 This is a flowchart illustrating the steps of the online optimization process of the fault diagnosis model in the method of the present invention. Detailed Implementation

[0051] The present invention will now be described in detail with reference to the accompanying drawings and specific embodiments. It should also be noted that, to make the embodiments more comprehensive, the following embodiments are the best and preferred embodiments; those skilled in the art can also use other alternative methods to implement some well-known technologies.

[0052] Furthermore, the accompanying drawings are only for describing the embodiments in more detail and are not intended to specifically limit the invention.

[0053] Please see Figures 1-3This invention provides a method for remote monitoring and operational status diagnosis of a welding machine. During the operation of the welding machine, the system first synchronously collects electrical, mechanical, and environmental parameters. Electrical parameters include voltage, current, and power; mechanical parameters include wire feed speed and welding angle; and environmental parameters include temperature and humidity. The collection of multi-dimensional parameters ensures a comprehensive reflection of the welding machine's operational status, avoiding the inaccurate diagnosis of complex problems in the welding process, such as unstable wire feed or fluctuations in ambient humidity, caused by a single electrical parameter. This synchronous collection provides a rich data foundation for subsequent data processing and analysis.

[0054] The collected multi-dimensional data comes from different sensors. Data fusion processing uses a unified timestamp for time synchronization to ensure temporal consistency of data from different sensors. The synchronized data undergoes filtering to eliminate environmental noise interference and ensure accuracy. The preprocessed data will also have time-domain features, frequency-domain features, and correlation features reflecting the relationships between parameters extracted. These features reveal dynamic changes during the welding process, effectively capturing potential anomalies and ensuring the accuracy of the analysis results.

[0055] After feature extraction, the data is input into a fault diagnosis model trained using machine learning algorithms. This model is trained using historical data from the welding machine under different known operating conditions, enabling it to predict the machine's current operating status and output diagnostic results. This method improves the accuracy of fault diagnosis by identifying complex relationships between multi-dimensional parameters. Compared to traditional single-parameter judgment models, it avoids the risks of false alarms and missed alarms, and can predict potential equipment faults in a timely manner, providing early warnings.

[0056] Ultimately, the diagnostic results, real-time collected data, and the digital twin of the welding machine built upon this data are transmitted to the remote monitoring terminal. This digital twin accurately reflects the real-time operating status of the welding machine, which can be viewed through a data visualization interface at the monitoring terminal. When the diagnostic results indicate abnormal equipment status or malfunction, the system can receive instructions from the remote monitoring terminal to promptly adjust the welding machine's operating parameters and perform fault repair or optimization operations. This function reduces the operational burden on on-site personnel and enables remote and intelligent equipment management.

[0057] This invention solves the problems of single data and independent parameter judgment in traditional systems by synchronously collecting multi-dimensional parameters, performing data fusion processing and feature extraction, thus making equipment diagnosis more comprehensive and accurate. Simultaneously, the machine learning-based diagnostic model improves fault prediction capabilities, reduces false alarms and missed alarms, and enhances maintenance efficiency. The introduction of remote monitoring and intelligent adjustment functions further enhances the intelligence level of the welding machine, reduces the need for human intervention, and makes welding operations more efficient and safer. This method not only optimizes the operation monitoring of welding machines but also provides strong support for precise equipment management in future intelligent manufacturing.

[0058] In one possible implementation, voltage and current data from the welding machine during the welding process are first collected using voltage and current sensors. The voltage signal undergoes rectification and filtering, a process that eliminates noise and ensures stable and consistent voltage data. The current signal is detected using a non-contact sensing method, which acquires the current signal through magnetic field induction and then converts it into a voltage signal for subsequent analysis. This approach avoids potential safety hazards associated with direct contact and ensures the accuracy of signal transmission.

[0059] Encoders mounted on the wire feed motor and the traveling mechanism are used to acquire the rotational speed signals of these two motors in real time. The encoder's function is to accurately measure the motor's rotational speed and calculate the actual wire feed speed and welding speed based on the radius of the motor gears. This speed data is crucial for the stability of the welding process, reflecting the dynamic changes in wire feeding and welding in real time, and avoiding fluctuations in welding quality caused by speed instability.

[0060] An angle sensor can accurately capture the welding angle between the welding torch and the workpiece. The welding angle directly affects the quality and stability of the weld. Therefore, accurately acquiring welding angle data helps to monitor the welding process in real time and make corresponding adjustments based on angle changes, ensuring weld quality.

[0061] Temperature, humidity, and carbon dioxide concentration data of the welding machine's working area are collected using temperature, humidity, and gas sensors. Environmental factors such as changes in temperature, humidity, and gas concentration directly affect welding performance and safety. Therefore, real-time monitoring of these environmental parameters is crucial to promptly identify potential environmental problems, such as excessively high temperatures or humidity levels, or elevated concentrations of harmful gases, allowing for appropriate preventative measures to be taken.

[0062] All sensor data acquisition is synchronized through a unified timing control mechanism. This synchronization mechanism ensures that data collected by different sensors has consistent timestamps, avoiding misdiagnosis or incorrect analysis caused by inconsistent data timestamps. Through data synchronization, the system can integrate various parameters for accurate analysis and judgment.

[0063] This multi-dimensional data synchronous acquisition method comprehensively and accurately reflects various operating parameters of the welding machine during operation, avoiding the limitations of relying on single parameters. Synchronous acquisition of voltage, current, wire feed motor speed, welding angle, and environmental parameters allows the system to understand all dynamic changes during the welding process in real time, enabling real-time monitoring and rapid response. Furthermore, the data synchronization mechanism ensures the accuracy and consistency of data from various sensors, providing a solid foundation for subsequent data fusion processing, fault diagnosis, and remote adjustment. Therefore, this method significantly improves the safety, stability, and intelligent management level of the welding machine.

[0064] In one possible implementation, during the data fusion process, filtering algorithms are used to suppress various types of noise and interference, ensuring that the collected data accurately reflects the true state of the welding machine.

[0065] To smooth the welding voltage and current signals, a filtering algorithm based on recursive estimation is used. This algorithm can eliminate random noise by continuously updating the estimated values, making the signal more stable and thus improving the accuracy of subsequent analysis.

[0066] For mechanical parameters such as wire feed speed and motor speed, a filtering algorithm based on median calculation is used. This algorithm can effectively remove sudden interference in the data, such as instantaneous sensor errors or external interference, thereby ensuring the accuracy of the mechanical parameters.

[0067] Environmental parameter data, such as temperature and humidity, are filtered using an algorithm based on linear averaging. Linear averaging filtering can smooth these continuously changing data, keeping environmental parameters stable during changes and avoiding the impact of abrupt changes on the overall trend.

[0068] Feature extraction is a key step in the condition analysis of welding machines, which can transform complex raw data into representative and diagnostic feature values.

[0069] For welding voltage and current signals, calculate their arithmetic mean, variance, peak value, and fluctuation coefficient. The arithmetic mean represents the normal level of the signal, the variance reflects the degree of signal fluctuation, and the peak value and fluctuation coefficient reveal the intensity of signal fluctuation. These time-domain characteristics allow us to understand the stability of voltage and current during the welding process, providing a basis for subsequent fault diagnosis.

[0070] Spectral transformation is performed on welding voltage and current signals to extract the main oscillation frequency components and frequency band energy distribution. Spectral analysis can reveal periodic fluctuations or oscillations that may exist in the welding machine during operation. These frequency domain characteristics can help identify abnormal frequencies that may be generated by the welding machine under high load, overload, or equipment aging conditions, and have important diagnostic value.

[0071] By analyzing the synergistic relationship between welding current and wire feed speed, and the mutual influence between welding voltage and ambient temperature, their statistical correlation strength can be further extracted. These multi-parameter correlation characteristics can reveal the intrinsic relationships between various parameters, such as whether the coordination between current and wire feed speed is appropriate, and whether the changes in voltage and temperature are reasonable. By calculating these correlation characteristics, equipment failures or performance degradation caused by the combined effects of multiple factors can be identified.

[0072] Through the filtering and feature extraction steps described above, various noises and interferences can be effectively eliminated, ensuring the accuracy and stability of the data. The diversity of filtering algorithms, employing different algorithms for different data types, allows each data source to be processed appropriately, resulting in more accurate and reliable results. The time-domain, frequency-domain, and multi-parameter correlation feature extraction processes enable the system to comprehensively and deeply analyze the welding machine's operating status. It can not only identify problems from single parameter changes but also reveal complex fault mechanisms through the interrelationships between various parameters.

[0073] This method greatly improves the fault diagnosis capability and operation monitoring accuracy of welding machines, provides a scientific basis for remote monitoring and real-time adjustment, and ensures efficient, safe and continuous stable operation of welding operations.

[0074] In one possible implementation, constructing a training sample set is fundamental to model training. First, it's necessary to collect multi-dimensional parameter data of the welding machine under normal, known abnormal, and fault conditions. This data typically comes from various sensors on the welding machine, such as welding current, voltage, temperature, and wire feed speed. The collected multi-dimensional data for each condition will constitute a complete sample set. By ensuring the sample set covers parameters for both normal operation and different fault conditions, the model can learn the typical characteristics of different operating states during training, enhancing its ability to identify complex fault situations.

[0075] Feature extraction is the process of transforming raw data into effective information that machine learning algorithms can process. For each training sample set, key features need to be extracted, such as time-domain features (mean, variance, fluctuation coefficient), frequency-domain features (oscillation frequency, energy distribution), and correlation features between multiple parameters. These features are extracted through deep analysis of the data, forming a corresponding set of feature vectors, which serve as input for subsequent machine learning model training. The feature extraction process ensures efficient data representation, removes noise and irrelevant information, and improves the accuracy of model training.

[0076] During training, machine learning algorithms capable of handling multi-class classification problems are selected, such as decision trees, support vector machines (SVM), random forests, or neural networks. These algorithms can effectively distinguish different operating states (e.g., normal, minor faults, severe faults), and the internal parameters of the model (e.g., weights, biases) are adjusted through optimization algorithms to enable the model to classify accurately. During training, the model is repeatedly optimized to ensure it can provide high-accuracy predictions under each operating state.

[0077] To ensure the model's generalization ability and stability, cross-validation is used for evaluation. Cross-validation avoids the risk of overfitting by dividing the dataset into multiple subsets and using different subsets for training and testing in turn. During the evaluation process, based on the model's performance in cross-validation, its structural parameters (such as the depth of the decision tree and the kernel function of the support vector machine) are adjusted until the model can achieve the predetermined performance indicators, ensuring its reliability and accuracy in practical applications.

[0078] In actual operation, the fault diagnosis model calculates the confidence level of the diagnostic results in real time. Confidence level is a measure of the model's certainty regarding its diagnostic results, representing the model's degree of confidence in the outcome. When the confidence level of the diagnostic results falls below a threshold determined through statistical analysis of historical operational data, the model activates a collaborative diagnosis mechanism based on a combination of multiple machine learning algorithms. This mechanism verifies and synthesizes the diagnostic results through the collaborative work of multiple models, thereby improving the reliability and accuracy of the diagnosis. By combining the advantages of different algorithms, the collaborative diagnosis mechanism overcomes the limitations of a single algorithm, further reducing the risk of misdiagnosis and missed diagnosis, and ensuring effective fault diagnosis even in complex situations.

[0079] By combining and optimizing various machine learning algorithms, efficient and accurate diagnosis of different states of welding machines can be achieved. By collecting rich, multi-dimensional parameter data and extracting effective features, the model can make accurate judgments under various operating conditions. Cross-validation and collaborative diagnosis mechanisms further improve the robustness and reliability of the model, especially when faced with incomplete or inaccurate data, enabling multiple algorithms to work together to compensate for the shortcomings of a single model. Ultimately, this method not only enhances the fault detection capabilities of welding machines but also enables remote monitoring and real-time early warning, reducing maintenance costs, extending equipment lifespan, and improving welding quality and work efficiency.

[0080] In one possible implementation, a three-dimensional geometric model of the welding machine needs to be constructed first. This model includes the spatial layout and shape of the various physical components of the welding machine. These components include the welding current controller, motor, sensors, welding torch, etc. Modern 3D modeling tools and software are used to accurately recreate the appearance and internal structure of the welding machine, ensuring consistency between the digital model and the actual machine in terms of shape and movement.

[0081] Simultaneously, the electrical and control systems of the welding machine also need to be digitally represented. The electrical system includes power supplies, current regulation modules, voltage sensors, etc., while the control system encompasses hardware components such as PLC control units and embedded processors. These systems accurately reproduce the electrical control logic and signal flow of the welding machine through digital models, ensuring real-time monitoring and adjustment of the electrical and control systems.

[0082] During the operation of the welding machine, real-time collected electrical parameters (such as voltage, current, and power), mechanical parameters (such as rotational speed, temperature, and welding speed), and environmental parameters (such as humidity, temperature, and air pressure) are mapped to corresponding variables in the digital twin. This data is typically collected in real-time by sensors and monitoring equipment, and the system transmits these parameters to the digital model via an interface.

[0083] Digital twins are data-driven, updating corresponding state variables in the model in real time. Each time the sensors collect new data, the state variables in the model will automatically adjust according to a preset mapping relationship. For example, if the temperature of the welding machine exceeds the set range, the temperature value in the model will be updated in real time and may trigger an early warning mechanism, indicating that the welding machine is about to enter a fault state.

[0084] Through digital twins, the actual working status of welding machines can be accurately displayed. The digital twin will show the movement status of each component of the welding machine in real time, dynamically displaying changes such as motor rotation and welding gun position, helping operators remotely monitor the machine's operation. Furthermore, the changing trends of electrical parameters (such as current fluctuations and power changes) will also be reflected in the model, facilitating user monitoring of the overall operating status of the welding machine. Simultaneously, the impact of environmental parameters, such as the effect of external temperature changes on the welding machine's working efficiency, is also dynamically mapped.

[0085] The remote monitoring terminal is the platform for end users to obtain real-time operating information of the welding machine. This port will display several important pieces of information, including:

[0086] It displays the electrical, mechanical, and environmental parameter curves of the welding machine in real time, helping users to clearly see the trend of parameter changes.

[0087] Based on the analysis of real-time data and fault diagnosis models, visual charts of fault diagnosis results are generated to help operators quickly understand the health status of the welding machine.

[0088] If the operating parameters of the welding machine exceed the safe range, the system will automatically generate an early warning message and display it on the remote monitoring terminal, prompting the operator to take timely action.

[0089] Users can view the historical operating data of the welding machine to help analyze the long-term operating trends and failure modes of the equipment, thereby providing data support for subsequent equipment maintenance and optimization.

[0090] By constructing a digital twin, the operating status of the welding machine can be accurately displayed on a remote platform with extremely high real-time performance and accuracy. The digital model can map various working states, fault manifestations, and parameter changes one by one, providing intuitive dynamic feedback and significantly improving the remote monitoring capabilities of the welding machine. Through precise real-time data updates and status displays, operators can promptly identify potential faults or performance anomalies, reducing equipment downtime and improving production efficiency. Meanwhile, the visualized historical data query interface provides strong support for equipment optimization and fault analysis, promoting the intelligent and precise management of equipment.

[0091] In one possible implementation, at the remote monitoring end, adjustment commands are first generated based on real-time diagnostic results. These diagnostic results are typically derived by analyzing the welding machine's operating status data (such as welding current, voltage, wire feed speed, etc.). If the diagnostic results indicate abnormal fluctuations or deviations from optimal operating conditions in the welding machine's parameters, the remote monitoring end will generate control commands, specifically adjusting parameters such as welding voltage, welding current, or wire feed speed. The purpose of these adjustments is to ensure the welding machine operates at its optimal state, improving welding quality and work efficiency.

[0092] Once a control command is generated, the system transmits it to the local control system of the welding machine via a wireless communication network. This network can employ wireless communication technologies such as Wi-Fi, Bluetooth, and 5G to ensure real-time data exchange and command transmission between the remote monitoring terminal and the welding machine. The advantage of wireless transmission lies in breaking spatial limitations, enabling real-time control of remote equipment, and greatly improving management flexibility and emergency response speed.

[0093] After receiving a control command, the local control system of the welding machine first analyzes it to understand the specific requirements for adjusting parameters such as welding voltage, current, or wire feed speed. The control system then converts these into corresponding drive signals using its internal control algorithm. These signals activate the welding machine's actuators, such as the current regulation module and wire feed motor, thereby adjusting the welding machine's operating parameters. Through this automated adjustment, the welding machine can quickly respond to commands from the remote monitoring terminal to adapt to different working environments and welding needs.

[0094] After parameter adjustments are completed, the welding machine will re-collect operating data, including new welding voltage, current, and wire feed speed. The welding machine then transmits this data back to the remote monitoring terminal via a communication network. The monitoring terminal uses this real-time data to verify the adjustment effect, ensuring the welding machine's operating status is optimized and remains stable in actual production. If the adjustment effect does not meet expectations, the system can continue to issue new adjustment commands, forming a closed-loop control process to achieve optimal operating conditions.

[0095] This adjustment mechanism based on remote monitoring and control effectively improves the automation and intelligence levels of welding machines. First, remote adjustment avoids manual intervention, improving operational efficiency and accuracy, especially suitable for unattended or complex environments. Second, adjustments based on diagnostic results make the welding machine's operation more precise, enabling timely responses to problems and preventing equipment damage or production quality issues caused by improper parameters. Furthermore, real-time data transmission and effect verification ensure the welding machine is always operating at its optimal state, thereby improving production efficiency, extending equipment lifespan, and reducing downtime. This method achieves remote intelligent equipment management, significantly enhancing the operational reliability and safety of welding machines.

[0096] In one possible implementation, during the operation of the welding machine, when the system detects a fault or abnormality, the remote monitoring terminal will make a fault judgment based on an existing fault diagnosis model. If there is a difference between the model prediction and the actual fault handling result, or if the model's confidence level is low (i.e., the model's diagnosis of a certain fault mode is uncertain or inaccurate), the system will record the difference data. This data includes the specific context in which the fault occurred, a comparison between the model's diagnostic results and the actual handling results, and the corresponding electrical, mechanical, or environmental parameter characteristics.

[0097] When the fault diagnosis model's diagnostic results do not match the actual processing results, or when the diagnostic confidence is low, the system collects the correct fault state label for the event, along with detailed data features related to the fault (such as parameters like temperature, pressure, and welding current). This data is considered the "true label" and added to the training dataset along with the corresponding features. This ensures that the model can make more accurate predictions when facing similar faults or anomalies in the future.

[0098] To adapt to the constantly changing operating conditions of welding machines and the emergence of new fault modes, the fault diagnosis model needs to be updated regularly. An incremental learning algorithm is employed, allowing each new training data point to optimize the existing model. This incremental learning method can gradually improve the model's accuracy and adaptability through new data without completely retraining the model. Incremental learning allows the model to maintain its ability to adapt quickly when dealing with new fault types, thereby improving the accuracy of diagnostic results.

[0099] After acquiring new training data, the system periodically updates the parameters of the fault diagnosis model using this data. By updating the model's weights or key parameters in the algorithm, the model can better reflect the behavior of the welding machine under different operating conditions and respond to new fault modes. This process typically involves fine-tuning or retraining the model parameters to ensure that the model can provide accurate diagnosis under new operating environments and technical conditions.

[0100] The online optimization process of the fault diagnosis model can significantly improve the accuracy and adaptability of welding machine fault prediction. First, by collecting real-time data on the differences between actual fault handling results and model predictions, the model can continuously self-calibrate, avoiding long-term accumulated errors. Second, the use of incremental learning allows the fault diagnosis model to quickly respond to new fault modes or operating conditions without needing to start training from scratch, greatly improving efficiency and flexibility. Most importantly, with the continuous increase in training data and continuous model optimization, the system will be able to more accurately diagnose the operating status of the welding machine in various environments, reducing equipment failures and improving production efficiency and safety. Furthermore, this online optimization method also reduces the need for manual intervention, making equipment management more intelligent and automated. Through continuous improvement, the fault diagnosis system can gradually become the intelligent brain of the welding machine, providing data support for fault early warning and maintenance decisions.

[0101] In one possible implementation, the welding angle is a key parameter affecting welding quality, significantly influencing the weld shape, strength, and overall welding effect. To accurately capture the angle changes of the welding torch during welding, a tilt sensor mounted on the torch is used for real-time monitoring. This sensor measures the tilt angle of the welding torch in multiple axial directions, typically including the horizontal axis, vertical axis, and the torch's swing angle. By accurately measuring these tilt angles, the welding torch's posture changes during welding can be tracked in real time, ensuring that the welding angle always conforms to preset standards and avoiding welding defects (such as porosity and uneven weld) caused by inappropriate welding angles.

[0102] The angle data measured by the sensor will be transmitted to the remote monitoring system in real time, helping operators to understand the current working status of the welding torch and adjust the welding process as needed, thereby further improving welding accuracy and quality.

[0103] In electric welding, carbon dioxide concentration is a crucial environmental parameter, affecting not only the atmosphere and stability of the welding process but also potentially the properties of the welding materials. To monitor gas concentration changes in real time during welding, a gas sensor based on the principle of infrared absorption is used to measure carbon dioxide concentration. This sensor emits infrared light of a specific wavelength and measures the absorption intensity of this light as it passes through carbon dioxide gas, thereby calculating the gas concentration. The absorption characteristics of carbon dioxide gas at specific wavelengths enable this sensor to accurately measure carbon dioxide concentration in the welding environment.

[0104] The gas sensor transmits concentration data to the remote monitoring system in real time. By analyzing this data, the monitoring terminal can determine whether the gas environment is normal during the welding process and issue an alarm when the gas concentration exceeds the safe range, so as to avoid the welding quality from deteriorating or the health of the operators due to abnormal gas concentration.

[0105] Simultaneous acquisition of two key parameters—welding angle and carbon dioxide concentration—can significantly improve the accuracy of monitoring the operating status of welding machines. Real-time monitoring of the welding angle ensures the stability of the welding process, avoiding welding defects caused by improper angles, thereby improving welding quality and production efficiency. Meanwhile, monitoring of carbon dioxide concentration provides technical support for ensuring the safety of the welding environment, preventing abnormal gas concentrations from affecting welding results or the health of operators.

[0106] Furthermore, by collecting these multi-dimensional parameters, the remote monitoring system can obtain more comprehensive and accurate welding machine operation data, providing more reliable data support for fault diagnosis, parameter optimization, and intelligent decision-making. This real-time collection of multi-dimensional data not only enhances the intelligent and automated management capabilities of the welding machine but also effectively prevents potential faults, improving equipment reliability and production efficiency.

[0107] In one possible implementation, a multi-class support vector machine (SVM) is a powerful machine learning algorithm suitable for handling classification problems involving multiple classes. In remote monitoring and operational status diagnosis of welding machines, welding machines may experience various types of faults (such as overheating, overload, current fluctuations, etc.), thus requiring the use of a multi-class SVM for accurate classification. SVM improves classification accuracy by finding the optimal separating hyperplane to maximize the margin between classes.

[0108] One of the core advantages of SVM is its ability to map data to a higher-dimensional space using kernel functions, thus effectively handling nonlinear classification problems. In the fault diagnosis of welding machines, the operating status of the equipment often exhibits complex nonlinear relationships, making the selection of an appropriate kernel function crucial. Common kernel functions include linear kernels, radial basis function (RBF) kernels, and polynomial kernels.

[0109] To ensure model performance, the selection of the kernel function and its parameters (such as the width parameter of the RBF kernel) need to be optimized. During the optimization process, the system searches within a predefined parameter range and selects the most suitable kernel function and its parameter combination, enabling the fault diagnosis model to accurately classify under different operating conditions.

[0110] The penalty factor (C) is an important hyperparameter in SVM, controlling the model's tolerance for classification errors. In fault diagnosis, an appropriate penalty factor helps the model better balance overfitting and underfitting. Setting the penalty factor too high may lead to overfitting, while setting it too low may lead to underfitting. Optimizing the penalty factor can improve the model's ability to identify different types of faults and avoid classification errors.

[0111] Grid search is a systematic hyperparameter optimization method that finds the optimal parameter combination by comprehensively searching every possible combination of parameters in a predefined parameter space. Specifically, grid search iterates through multiple hyperparameters such as kernel function parameters and penalty factors, evaluates the performance of each parameter set, and verifies the effectiveness of each set using cross-validation, selecting the parameter combination with the best cross-validation results. This process ensures that the SVM model achieves an optimal performance balance between the training and test sets, improving the accuracy of fault diagnosis.

[0112] By employing a multi-class support vector machine (SVM) and utilizing a grid search strategy to optimize the kernel function and penalty factor, the performance of the welding machine fault diagnosis model can be significantly improved. First, the SVM algorithm effectively handles complex nonlinear classification problems, ensuring the model can accurately identify different types of faults. Second, by optimizing the kernel function and penalty factor, overfitting or underfitting can be avoided, thereby improving the stability and reliability of fault diagnosis. The grid search strategy ensures optimal parameter selection, maximizing model performance and improving fault diagnosis accuracy.

[0113] The embodiments of the present invention not only improve the accuracy of remote monitoring and operation status diagnosis of welding machines, but also enhance the robustness of the system in the face of multiple fault modes, enabling the welding machine to maintain high diagnostic efficiency under complex working conditions, thereby reducing equipment failure rate and improving production efficiency and safety.

[0114] Accordingly, embodiments of the present invention provide a remote monitoring and operation status diagnosis system for an electric welding machine, including a memory configured to store instructions, a processor configured to call the instructions from the memory, and capable of implementing any of the remote monitoring and operation status diagnosis methods for an electric welding machine described in any of the embodiments of the present invention when executing the instructions.

[0115] This invention encompasses any substitutions, modifications, equivalent methods, and solutions made within the spirit and scope of this invention. To provide the public with a thorough understanding of this invention, specific details are described in detail in the following preferred embodiments; however, those skilled in the art will fully understand the invention even without these details. Furthermore, to avoid unnecessary misunderstanding of the essence of this invention, well-known methods, processes, procedures, components, and circuits are not described in detail.

[0116] The above description is only a preferred embodiment of the present invention. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principle of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A method for remote monitoring and operational status diagnosis of an electric welding machine, characterized in that, Includes the following steps: The system synchronously collects multi-dimensional parameters of the welding machine during operation, including electrical parameters, mechanical parameters, and environmental parameters. The collected multi-dimensional parameters are fused, including time synchronization of data from different sensors using a unified timestamp, filtering of the synchronized data to eliminate noise, and then extracting time-domain features, frequency-domain features, and correlation features reflecting the relationship between parameters from the preprocessed data. The extracted features are input into a fault diagnosis model trained based on machine learning algorithms for operational status diagnosis. The fault diagnosis model is trained using historical data of the welding machine under different known working conditions and can output diagnostic results representing the current health status of the welding machine. The diagnostic results, real-time collected data, and a digital twin reflecting the real-time status of the welding machine, constructed based on these data, are transmitted to a remote monitoring terminal for display and monitoring. When the diagnostic results indicate an abnormality or fault, parameter adjustment instructions are received from the remote monitoring terminal to adjust the operating parameters of the welding machine accordingly.

2. The method for remote monitoring and operational status diagnosis of a welding machine according to claim 1, characterized in that, The specific process of synchronously acquiring multi-dimensional parameters includes: acquiring welding voltage and welding current through voltage and current sensors, wherein the voltage signal is rectified and filtered, and the current signal is acquired and converted into a voltage signal through non-contact sensing. The speed signals of the wire feeding motor and the traveling mechanism are collected by encoders installed on the wire feeding motor and the traveling mechanism, and the actual wire feeding speed and welding speed are calculated based on the gear radius of the motor. The welding angle between the welding torch and the workpiece is collected using an angle sensor. The ambient temperature, humidity, and carbon dioxide concentration data of the welding machine's working area are collected using temperature sensors, humidity sensors, and gas sensors, respectively. All sensor data acquisition is synchronized through a unified timing control mechanism to ensure data consistency over time.

3. The method for remote monitoring and operational status diagnosis of an electric welding machine according to claim 1, characterized in that, The data fusion processing includes filtering the data as follows: for welding voltage and welding current signals, a filtering algorithm based on recursive estimation is used for smoothing to suppress random noise; For mechanical parameter data, a filtering algorithm based on median calculation is used to process the data to eliminate sudden interference. For environmental parameter data, a filtering algorithm based on linear averaging is used to process the data in order to maintain its stability. The feature extraction process includes: calculating the arithmetic mean, variance, peak value, and fluctuation coefficient of the welding voltage and welding current signals as time-domain features; Spectral transformation is performed on welding voltage and welding current signals to extract the main oscillation frequency components and frequency band energy distribution as frequency domain features; The synergistic relationship between welding current and wire feed speed, and the mutual influence between welding voltage and ambient temperature were analyzed, and their statistical correlation strength was calculated as a multi-parameter correlation feature.

4. The method for remote monitoring and operational status diagnosis of a welding machine according to claim 1, characterized in that, The training process of the fault diagnosis model includes: collecting historical multi-dimensional parameter data of the welding machine under normal conditions, multiple known abnormal conditions, and fault conditions to form a training sample set; Feature extraction is performed on the data in the training sample set to form a corresponding set of feature vectors; The model is trained using a machine learning algorithm capable of handling multi-classification problems. The internal parameters of the model are adjusted by optimizing the algorithm so that the model can accurately distinguish different working states. The model's performance is evaluated using cross-validation, and the model's structural parameters are adjusted based on the evaluation results until the model reaches the predetermined performance indicators. During real-time diagnosis, the fault diagnosis model calculates the confidence level of its diagnosis results. When the confidence level is lower than the threshold determined by statistical analysis of historical operation data, a collaborative diagnosis mechanism based on a combination of multiple machine learning algorithms is activated to verify and synthesize the diagnosis results.

5. The method for remote monitoring and operational status diagnosis of a welding machine according to claim 1, characterized in that, The process of constructing the digital twin includes: Establish a three-dimensional geometric model of the welding machine and a digital representation of its electrical and control systems; The real-time collected electrical, mechanical, and environmental parameters are mapped onto the digital model, driving the corresponding state variables in the model to be updated, so that the digital twin can reflect the actual working state of the welding machine in real time, including the motion state of each component, the changing trend of electrical parameters, and the influence of environmental parameters. The remote monitoring terminal displays real-time operating parameter curves of the welding machine, a visual display of operating status diagnostic results, a list of early warning information prompts, and a historical data query interface.

6. The method for remote monitoring and operational status diagnosis of a welding machine according to claim 1, characterized in that, The process of receiving parameter adjustment instructions and adjusting the operating parameters of the welding machine includes: the remote monitoring terminal generating control instructions to adjust the welding voltage, welding current or wire feed speed based on the diagnostic results; Control commands are transmitted to the local control system of the welding machine via a wireless communication network; The local control system parses the instructions and converts them into drive signals, which drive the actuators to adjust the working parameters of the welding machine. After the parameters are adjusted, the working data of the welding machine is collected again and transmitted to the remote monitoring terminal to verify the adjustment effect.

7. The method for remote monitoring and operational status diagnosis of a welding machine according to claim 1, characterized in that, The method also includes an online optimization process for the fault diagnosis model: Collect data on the differences between actual fault handling results and model diagnostic results; When a model diagnosis error or low confidence is found, the correct state label and corresponding data feature are added to the training dataset. An algorithm capable of incremental learning is adopted to periodically update the parameters of the fault diagnosis model using newly added training data, enabling the model to adapt to changes in the working state of the welding machine and new fault modes.

8. The method for remote monitoring and operational status diagnosis of a welding machine according to claim 2, characterized in that, During the process of synchronously acquiring multi-dimensional parameters, the welding angle is specifically acquired through an angle sensor installed on the welding torch, which measures the tilt angle of the welding torch in multiple axial directions. Carbon dioxide concentration is collected using a gas sensor based on the principle of infrared absorption, which calculates the gas concentration by measuring the absorption intensity at a specific infrared wavelength.

9. The method for remote monitoring and operational status diagnosis of a welding machine according to claim 4, characterized in that, The machine learning algorithm used in the training of the fault diagnosis model is a multi-class support vector machine. Its kernel function parameters and penalty factors are optimized through a grid search strategy, that is, a system search is performed within a predefined range of parameter values, and the optimal parameter combination is selected based on the cross-validation results.

10. A remote monitoring and operation status diagnosis system for an electric welding machine, characterized in that, The method includes a memory configured to store instructions, a processor configured to retrieve the instructions from the memory, and, when executing the instructions, to implement a remote monitoring and operation status diagnosis method for a welding machine as described in any one of claims 1-9.

Images