Electric welding operation safety authorization management and control method based on Internet of Things

Abstract

The invention relates to the technical field of safety management and control, and discloses an electric welding operation safety authorization management and control method based on the Internet of Things, and the method comprises the steps: collecting the operation parameters of electric welding equipment through an Internet of Things sensor network, and carrying out the abnormality recognition; performing propagation path reasoning along a physical connection and function dependency relationship edge in a map by using a pre-constructed equipment fault knowledge map and taking a component health state corresponding to the abnormal parameter as a starting point, and accumulating time delay on a path so as to deduce a potential fault chain with a clear timeline; and finally, according to the risk level and the predicted occurrence time of each node in the fault chain, the operation authority level of the operator is dynamically adjusted, quantitative prediction of the electric welding operation risk and differential authority control of the operator are realized, and the intrinsic safety level of the electric welding operation is improved from the source.

Description

Technical Field

[0001] This invention relates to the field of safety management and control technology, and in particular to a method for safety authorization and management of electric welding operations based on the Internet of Things. Background Technology

[0002] Electric welding operations, due to their involvement in high-temperature, electric arc, and flammable environments, are a key area for risk management in industrial safety production. Currently, monitoring the status of welding equipment using IoT technology has become a common direction for industry upgrades. The mainstream technical solution involves deploying a sensor network on the welding machine to collect operating parameters such as welding current, voltage, and welding torch temperature in real time, and monitoring and alarming based on preset fixed thresholds. This achieves a shift from manual experience-based judgment to data threshold-based alarms. When any parameter exceeds its independently set safety limit, an alarm is triggered, or even an emergency shutdown is executed.

[0003] However, existing technologies based on independent parameters and fixed thresholds reveal a fundamental technical limitation when faced with complex equipment failures with cascading characteristics: they can only passively respond to existing, explicit, and isolated abnormal states, and cannot predict or proactively intervene in potential, implicit, and systemic risks with propagation and evolution characteristics. Summary of the Invention

[0004] Therefore, the purpose of this invention is to overcome the limitations of existing technologies that can only passively respond to isolated, already occurring welding equipment anomalies. It provides an IoT-based method for the safety authorization and control of welding operations. This method uses a knowledge graph of equipment faults to infer the propagation path of abnormal parameters, obtains potential fault chains with time predictions, and dynamically adjusts operating permissions accordingly. This enables quantitative prediction of welding operation risks and differentiated permission control for operators, thereby improving the inherent safety level of welding operations from the source.

[0005] To address the aforementioned technical problems, this invention provides a method for safe authorization and control of welding operations based on the Internet of Things, comprising: The operating parameters of the welding equipment are collected through an Internet of Things (IoT) sensor network deployed on the welding equipment; anomalies are identified based on the operating parameters to obtain abnormal parameters. Based on the equipment fault knowledge graph, the propagation path of the abnormal parameters is inferred to obtain the potential fault chain; wherein, the equipment fault knowledge graph uses the health status of the welding equipment components as nodes and the physical connection relationship and functional dependency relationship between the welding equipment components as edges, and each edge stores the probability weight and time delay of fault propagation. The propagation path reasoning includes: taking the health status of the welding equipment component corresponding to the abnormal parameter as the starting node, traversing downstream fault nodes according to the edge relationships and fault propagation probability weights in the equipment fault knowledge graph, and accumulating the time delay of each edge on the traversal path to obtain the potential fault chain. Based on the risk level of the downstream fault node in the potential fault chain and the corresponding time delay, the current operating permissions of the welding equipment operator are adjusted.

[0006] Preferably, the operating parameters include electrical parameters, temperature parameters, and spatial environment parameters; obtaining the abnormal parameters through anomaly identification includes: determining the risk category of the current work scenario based on the spatial environment parameters; determining the electrical weight of the electrical parameters and the temperature weight of the temperature parameters in the safety assessment based on the risk category; weighting the electrical parameters based on the electrical weights to obtain weighted electrical parameters; weighting the temperature parameters based on the temperature weights to obtain weighted temperature parameters; and identifying parameters in the weighted electrical parameters and the weighted temperature parameters that exceed the corresponding safety thresholds as the abnormal parameters.

[0007] Preferably, the spatial environment parameters include combustible gas concentration, ambient temperature, and equipment vibration amplitude. Determining the risk category of the current work scenario based on these spatial environment parameters includes: comparing the measured values ​​of the combustible gas concentration, ambient temperature, and equipment vibration amplitude with the judgment conditions corresponding to each of multiple predefined risk categories; determining whether the current measured value fully meets all the judgment conditions of any predefined risk category based on the comparison results; if fully met, then the predefined risk category is determined as the risk category of the current work scenario; if not fully met, then a risk proximity analysis is performed, and the risk category of the current work scenario is determined based on the degree of closeness between the current measured value and the judgment conditions of each predefined risk category.

[0008] Preferably, the risk proximity analysis includes: determining at least one potential risk category based on the comparison results; wherein, the potential risk category is defined as one whose judgment condition is partially met; for each potential risk category, calculating the difference between its unmet judgment condition and the corresponding measured value, and assessing the degree of non-metdness of the judgment condition based on the difference; calculating the risk proximity score of the potential risk category by comprehensively considering the degree of non-metdness corresponding to all unmet judgment conditions of each potential risk category; selecting the category with the highest risk proximity score from all potential risk categories as a candidate risk category; obtaining the current operation type of the welding equipment, and determining whether there is a correlation between the risk represented by the candidate risk category and the operation type; if there is, determining the candidate risk category as the risk category of the current operation scenario; if not, determining the baseline risk category as the risk category of the current operation scenario.

[0009] Preferably, adjusting the current operating permissions of the welding equipment operator based on the risk level of the downstream fault node in the potential fault chain and the corresponding time delay includes: identifying the downstream fault node with the shortest time delay from the potential fault chain; comparing the risk level of the node with a risk level threshold; if the risk level of the node is higher than the risk level threshold, adjusting the operator's current operating permissions to a restriction level corresponding to the risk level of the node; if the risk level of the node is equal to or lower than the risk level threshold, maintaining the current operating permissions unchanged.

[0010] Preferably, constructing the equipment fault knowledge graph includes: acquiring historical fault data of each component of the welding equipment and analyzing the causal relationship between faults; defining components and their health status as graph nodes, and defining the physical connection relationship and functional dependency relationship between components as graph edges; assigning an initial fault propagation probability weight and time delay to each edge based on historical fault statistics; and establishing a graph update mechanism to dynamically adjust the probability weight and time delay according to actual fault propagation cases that occur during system operation.

[0011] Preferably, the propagation path reasoning step further includes correcting the fault propagation probability weight, which includes: during the traversal process, obtaining the operating load rate of the welding equipment; based on the operating load rate, finding the corresponding correction coefficient from a preset load-probability correction curve; multiplying the static fault propagation probability weight stored in the equipment fault knowledge graph with the correction coefficient to obtain the corrected fault propagation probability weight; and using the corrected fault propagation probability weight to traverse downstream fault nodes; wherein, the load-probability correction curve satisfies the following condition: the higher the operating load rate, the larger the correction coefficient.

[0012] Preferably, constructing the load-probability correction curve includes: conducting welding equipment fault simulation experiments under different operating load rates, and statistically analyzing the actual probability of fault propagation along a specific path; plotting a scatter plot with the operating load rate on the horizontal axis and the ratio of the actual propagation probability to the static probability on the vertical axis; and performing curve fitting on the scatter plot to obtain the load-probability correction curve.

[0013] Preferably, the cumulative time delay of each edge on the traversal path includes: starting from the starting node, recording each edge traversed along the traversal path and its corresponding time delay; and summing the time delays of all edges on the path sequentially to obtain the cumulative time delay from the starting node to the current traversal node.

[0014] Preferably, when multiple paths lead to the same node, the path with the shortest cumulative time delay is selected as the main propagation path of that node, and its cumulative time delay is used as the effective time delay of that node.

[0015] Compared with the prior art, the above-described technical solution of the present invention has the following advantages: The IoT-based safety authorization and control method for electric welding operations described in this invention constructs a knowledge graph of equipment faults based on the physical connections and functional dependencies of components such as welding torches, cables, power modules, and cooling systems in welding equipment. It maps detected welding anomalies, such as abnormal current, unstable arc, or localized overheating, to the initial fault states of specific components. Based on the fault propagation time delay calibrated according to welding process parameters in the graph, it deduces a temporal propagation chain from the current anomaly to subsequent components. Finally, based on the risk level and expected occurrence time of each node in this propagation chain, it forms differentiated permission adjustments adapted to the welding operation process, achieving quantitative prediction of welding operation risks and differentiated permission control for operators, thereby improving the inherent safety level of electric welding operations from the source. Attached Figure Description

[0016] To make the content of this invention easier to understand, the invention will be further described in detail below with reference to specific embodiments and accompanying drawings, wherein: Figure 1 This is a flowchart of the IoT-based safety authorization and control method for electric welding operations in a preferred embodiment of the present invention; Figure 2 This is a flowchart for determining risk categories in a preferred embodiment of the present invention; Figure 3 This is a flowchart illustrating the adjustment of the current operating permissions of the welding equipment in a preferred embodiment of the present invention. Detailed Implementation

[0017] The present invention will be further described below with reference to the accompanying drawings and specific embodiments, so that those skilled in the art can better understand and implement the present invention. However, the embodiments described are not intended to limit the present invention.

[0018] To facilitate understanding, the following will first explain the various components of welding equipment, as well as the nodes and edges of the equipment fault knowledge graph: Taking a common inverter-type DC arc welding machine as an example, welding equipment is a device that integrates power electronics, mechanics, control, and cooling systems. Its main components can be classified as follows: I. Power conversion and output components, including: (1) Input filtering and rectification unit, which includes input EMI filter and rectifier bridge; the healthy state is characterized by the decay of the filter capacitor value and the deterioration of the rectifier diode performance.

[0019] (2) Inverter unit, whose core components include insulated gate bipolar transistor (IGBT) or field-effect transistor (MOSFET); the health status is characterized by: abnormal junction temperature of the switching transistor, distortion of the driving waveform, and increased saturation voltage drop.

[0020] (3) Medium frequency transformer: performs voltage transformation and isolation; health status is characterized by abnormal winding temperature, decreased insulation resistance, and loose magnetic core with abnormal noise.

[0021] (4) Output rectification and filtering unit: including fast recovery diode and DC reactor; the health status is characterized by: increased reverse leakage current of diode and change of inductance of reactor.

[0022] (5) Output terminals and welding gun cable: directly connected to the load; health status is characterized by: increased terminal contact resistance, aging or damage to the cable insulation layer.

[0023] II. Control and auxiliary components, including: (1) Main control circuit board: generates PWM drive signal to realize constant current / constant voltage and other external characteristic control; the health status is manifested as: reference voltage drift, AD sampling error increase, and program crash.

[0024] (2) Driver circuit board: amplifies the control signal to drive the power switching transistor; the health status is characterized by insufficient driving capability and output pulse distortion.

[0025] (3) Human-machine interface: including buttons, digital tubes or screens; health status may be manifested as: button malfunction or display abnormality.

[0026] (4) Current / voltage / temperature sensor: used for real-time sampling feedback; health status is manifested by: sampling value drift, accuracy exceeding tolerance, and signal disconnection.

[0027] III. Cooling and thermal management components, including: (1) Heat sink: to dissipate heat from power devices; healthy condition is characterized by: dust accumulation on the surface, dried thermal grease, and deformed fins.

[0028] (2) Cooling fan: forced air cooling; health status is characterized by: abnormal noise from bearing wear, speed decrease, and stoppage.

[0029] (3) Temperature switch or thermistor: overheat protection sensor; health status is manifested as: action point drift, failure (normally open or normally closed).

[0030] IV. Power supply and peripheral components, including: (1) Switching power supply: provides low-voltage DC power to control circuits, drive circuits, fans, etc.; its healthy state is characterized by unstable output voltage and reduced load capacity.

[0031] (2) Housing and structural components: provide physical protection and support; health status is manifested by: deformation and poor grounding.

[0032] Example of component health status as a node in a device fault knowledge graph: In a device fault knowledge graph, nodes are not component names, but rather represent a specific, observable, or inferable health status of that component at a given moment. These states serve as the starting point, intermediate point, or endpoint for fault propagation reasoning. The following example illustrates how to map the aforementioned components to ordinary nodes: (1) Example of a health status node for an IGBT (inverter unit): IGBT - Drive is normal - Junction temperature is normal; IGBT - Drive undervoltage - Performance limited; IGBT - Junction temperature too high - Derating operation; IGBT - Junction temperature exceeded limit - Thermal protection triggered; IGBT-CE breakdown-short circuit fault.

[0033] (2) Example of a health state node for an intermediate frequency transformer: Transformer winding temperature rise is normal; Transformer - Localized overheating of windings; Transformer - Insulation slightly damp (insulation resistance decreased); Transformer - minor inter-turn short circuit (increased harmonics); Transformer - winding burnout - open circuit fault.

[0034] (3) Example of a healthy state node for the output rectifier diode: The rectifier diode's on-state voltage drop is normal. The rectifier diode has an excessive reverse leakage current. Rectifier diode-PN junction thermal fatigue; Rectifier diode - breakdown and short circuit.

[0035] A physical component corresponds to multiple different health status nodes, and each status node should be associated with one or more monitorable parameters or phenomena; the fault propagation edge links a poor health status from one component to a poor health status of another component, and is accompanied by propagation probability and time delay.

[0036] Reference Figure 1 As shown in the figure, this invention discloses a method for safety authorization and control of electric welding operations based on the Internet of Things, including: S100: Collect the operating parameters of the welding equipment through an Internet of Things (IoT) sensor network deployed on the welding equipment; S200: Identify anomalies based on operating parameters and obtain abnormal parameters; S300. Based on the equipment fault knowledge graph, perform propagation path reasoning on abnormal parameters to obtain potential fault chains. The equipment fault knowledge graph uses the health status of welding equipment components as nodes and the physical connection and functional dependency relationships between welding equipment components as edges. Each edge stores the probability weight and time delay of fault propagation. Propagation path reasoning includes: taking the health status of the welding equipment component corresponding to the abnormal parameter as the starting node, traversing downstream fault nodes according to the edge relationships and fault propagation probability weights in the equipment fault knowledge graph, and accumulating the time delay of each edge on the traversal path to obtain potential fault chains. S400. Adjust the current operating permissions of the welding equipment operator based on the risk level and corresponding time delay of the downstream fault node in the potential fault chain.

[0037] In step S100, in a specific implementation scenario, multiple types of IoT sensors are deployed at key locations of the target welding equipment to form a heterogeneous sensor network. This includes deploying Hall current sensors and voltage sensors at the input and output ends to collect input current, output voltage, and welding current in real time; deploying digital temperature sensors or thermocouples at hot spots such as the surface of the IGBT module heat sink, the main transformer housing, and the output terminals to collect the temperature of key components; and deploying combustible gas concentration sensors (such as MQ-5), ambient temperature and humidity sensors, and vibration sensors (such as MPU6050) in the welding machine operating area to monitor the operating environment.

[0038] Each sensor samples synchronously or quasi-synchronously at a preset sampling frequency. The acquired analog signals are converted from analog to digital and then preliminarily processed and packaged by an embedded microcontroller. The data packets are uploaded in real time to the monitoring platform's data receiving service deployed on a cloud server or edge server via a 4G DTU module or industrial Wi-Fi module, through the mobile internet or local area network. The monitoring platform performs preprocessing on the received raw data, including timestamp alignment, missing value imputation, and preliminary filtering, to form a structured, multi-dimensional real-time data stream.

[0039] In a specific implementation scenario, step S200 determines whether the currently monitored operating parameters are abnormal based on a threshold comparison method. In this embodiment of the invention, the operating parameters include electrical parameters, temperature parameters, and environmental parameters, with electrical parameters including current and voltage. Considering the need for anomaly assessment based on multiple parameters, and given that the potential safety risks of welding operations are highly dependent on their operating environment, traditional monitoring systems use uniform monitoring weights and safety thresholds for parameters in all environments, leading to insensitive risk perception or the generation of numerous invalid alarms. To address this problem, this embodiment of the invention provides an anomaly identification process that can adapt to dynamically changing operating environments, achieving intelligent focusing and dynamic weight allocation of monitoring priorities under different risk scenarios, thereby improving the accuracy of risk identification, the effectiveness of alarms, and the utilization efficiency of system resources. The specific solution includes the following steps: S210. Based on spatial environment parameters, determine the risk category of the current work scenario; S220. Determine the electrical weights of electrical parameters and the temperature weights of temperature parameters in the safety assessment based on the risk category; weight the electrical parameters based on the electrical weights to obtain weighted electrical parameters; weight the temperature parameters based on the temperature weights to obtain weighted temperature parameters. S230. Identify parameters in the weighted electrical parameters and weighted temperature parameters that exceed the corresponding safety thresholds as abnormal parameters.

[0040] In a specific implementation scenario, step S220 involves a pre-defined risk category-weight mapping table. This table defines the basic monitoring weights that electrical and temperature parameters should be assigned under different risk categories, as shown in the example below: Based on the risk category determined in step S210, the mapping table is queried to directly obtain the corresponding electrical weight and temperature weight, which satisfy the condition that the sum of these two weights is 1.

[0041] In a specific implementation scenario, step S230 first normalizes the collected raw electrical and temperature parameters. The normalized parameter values ​​are then multiplied by their corresponding weights to obtain weighted parameter values. The system sets an anomaly trigger threshold for the weighted parameters, for example, 0.8, indicating that reaching 80% of the safety limit requires attention. Since the parameters are weighted and normalized, this threshold is universal.

[0042] The weighted electrical parameters are compared with the electrical safety threshold. If the weighted electrical parameters are greater than the electrical safety threshold, an electrical parameter anomaly is determined, and the anomaly type, associated components, anomaly intensity, and timestamp are recorded and encapsulated into an anomaly parameter object. If the weighted temperature parameters are greater than the temperature safety threshold, a temperature parameter anomaly is determined, and an anomaly parameter object is also generated. The list of generated anomaly parameter objects is passed as the output of step S200 to the subsequent fault propagation path reasoning module.

[0043] Upon receiving abnormal parameters, instead of performing a simple threshold comparison, the parameters are mapped to a starting state node in the equipment fault knowledge graph. This activates graph-based traversal and time accumulation. Along the edge relationship network defined by the graph, the propagation of fault states is simulated. By accumulating the time delay of each edge on the path, the estimated time to reach each downstream fault node is dynamically calculated, thereby obtaining a structured and time-sequential potential fault chain.

[0044] In specific implementation scenarios, the equipment fault knowledge graph is a directed weighted graph, where nodes represent the specific health status of a particular component of the welding equipment. For example, for an IGBT module, its nodes include IGBT - normal junction temperature, IGBT - high junction temperature - no performance degradation, IGBT - excessively high junction temperature - requires derating, IGBT - thermal breakdown - short circuit, etc. Each node is associated with monitoring parameter conditions that can trigger that state.

[0045] Edges represent the propagation of fault states between components; edges are directional, pointing from the cause to the effect. Each edge contains two key attributes: probability weight and time delay. The probability weight represents the likelihood that a fault will propagate along this edge to the effect state in the cause state; the time delay represents the time required from the occurrence of the cause state to the occurrence of the effect state.

[0046] The types of edge relationships include physical connection relationships and functional dependency relationships. Physical connection relationships are based on the direct transfer of current, heat, and mechanical force; for example, cooling fan stops → heat accumulates → radiator temperature spikes. Functional dependency relationships are based on logical associations such as control and protection; for example, abnormal drive circuit output → control failure → IGBT switching malfunction.

[0047] The abnormal parameters obtained in step S200 are mapped to corresponding nodes in the device fault knowledge graph. This mapping is completed by precise or fuzzy matching of the abnormal type and component identifier in the graph node attribute library. Starting with the mapped node, a graph traversal algorithm is initiated. The algorithm explores along all outgoing edges of the node, visiting downstream faulty nodes. During the traversal, the core operation is to accumulate the time delay. For example, if the edge delay from the starting node A to node B is t1, and the edge delay from B to C is t2, then the cumulative time delay from A to C is t1 + t2. The system records the shortest cumulative time delay to reach each downstream node and its propagation path.

[0048] The traversal stops after reaching the preset depth or the cumulative delay exceeds the maximum warning time. The system integrates the traversal results into one or more potential fault chains. Each potential fault chain is an ordered sequence of nodes, for example: [IGBT - Junction temperature too high - Derating required] — (5 min) → [Output capacitor - High temperature aging] — (10 min) → [Main control board - Power supply abnormality]. Each potential fault chain contains the possible development path of the fault and the time of each stage.

[0049] In step S400, in a specific implementation scenario, each downstream fault node in the potential fault chain is evaluated. Each node has a predefined risk level, and the urgency of each downstream fault node is assessed based on its cumulative time delay. For example, a high-risk node has an extremely high urgency if the delay is only 2 minutes; if the delay is 30 minutes, the urgency is moderate. A permission adjustment policy table is maintained, defining the operational permission levels to be implemented under different urgency levels. Permission levels can be divided from high to low as follows: full-function operation, supervised operation (safety officer required), restricted operation (e.g., limiting maximum current, prohibiting continuous welding), maintenance-only mode, and complete lockout.

[0050] This invention's solution uses equipment failure knowledge graph reasoning to link single-point anomalies with potential cascading failures, issuing early warnings before serious failures actually occur, thus shifting the safety control checkpoint forward. By accumulating time delays, a quantified time window is provided for each predicted risk, making early warnings and interventions more timely and targeted, avoiding blind panic or delayed responses. Based on risk level and time urgency, differentiated and tiered dynamic adjustments to permissions are implemented, changing the one-size-fits-all shutdown model and minimizing unnecessary interruptions to normal production while ensuring basic safety.

[0051] In the aforementioned embodiments, determining the risk category of the current work scenario based on spatial environment parameters presents a challenge in practical implementation. Predefined risk categories rely on clear parameter boundary conditions, while real work environments are dynamic, continuous, and parameters are often in a critical state. To address this issue, the embodiments of this invention provide a risk category determination logic capable of handling both critical and complex environments. This logic not only identifies typical and clearly defined high-risk scenarios but also effectively identifies and quantifies atypical risk states that are in the process of formation, ensuring the continuous and stable operation of the safety decision chain even in complex environments. (Refer to...) Figure 2 As shown, the specific solution is as follows: Preferably, the spatial environment parameters include combustible gas concentration C, ambient temperature T, and equipment vibration amplitude V. Determining the risk category of the current work scenario based on these spatial environment parameters includes: comparing the measured values ​​of combustible gas concentration, ambient temperature, and equipment vibration amplitude with the judgment conditions corresponding to each of multiple predefined risk categories; determining whether the current measured value fully meets all the judgment conditions of any predefined risk category based on the comparison results; if fully met, then the predefined risk category is determined as the risk category of the current work scenario; if not fully met, then a risk proximity analysis is performed, and the risk category of the current work scenario is determined based on the degree of closeness between the current measured value and the judgment conditions of each predefined risk category.

[0052] In a specific implementation scenario, the system has a pre-built structured risk category library. Each predefined risk category is explicitly associated with a set of judgment conditions, which are thresholds or ranges set for each spatial environment parameter. Subsequently, the current spatial environment parameter vector [C, T, V] is logically compared with each non-benchmark risk category. The comparison is a "yes / no" judgment for each condition item in the set of judgment conditions.

[0053] For each predefined risk category, check whether its measured value satisfies all the conditions of a set of judgment criteria. This is equivalent to performing a logical AND operation on the judgment results of all the conditions in a set of judgment criteria. If there is a predefined risk category whose measured value satisfies all the judgment criteria of that risk category, it is judged as fully satisfied, the predefined risk category is determined as the final risk category, and the subsequent risk approximation analysis is skipped.

[0054] If, for all non-benchmark risk categories, the measured values ​​fail to meet all the judgment conditions of each predefined risk category (i.e., the logical AND operation results are all false), then the risk proximity analysis process begins. This indicates that the current environment does not belong to any typical "high-risk" or "medium-risk" scenario, but may be in the formation or complex state of some kind of risk. At this point, all non-benchmark risk categories are traversed to find all predefined risk categories where at least one condition is met. These predefined risk categories are called potential risk categories. For each identified potential risk category, its risk proximity score is calculated. One feasible calculation method is to take the average or weighted sum of the scores of all condition items, with a value between [0, 1]. The higher the value, the closer the current environment is to fully satisfying the potential risk category. If a condition item is met, the score for that item is 1. If a condition item is not met, a decay score is calculated using a decay function based on relative distance.

[0055] Finally, from all potential risk categories, the risk category with the highest approximation score is selected as the risk category for the current work scenario.

[0056] In implementing the risk approximation analysis scheme, scoring and selection are based solely on the mathematical approximation of spatial environmental parameters and predefined risk category judgment conditions. In actual implementation, this can lead to pseudo-high-risk judgments that are irrelevant to the actual welding operation risk. Specifically, this manifests as follows: Assuming the current environmental parameters are: high vibration amplitude (due to the operation of nearby heavy equipment), but normal combustible gas concentration and ambient temperature; the system, based on approximation calculations, might classify medium-risk – mechanical instability – as the highest-scoring candidate category. However, if the welding equipment is currently performing ground-based, low-current repair welding, the operation itself is extremely insensitive to mechanical vibration, and the risk of welding quality degradation or safety accidents due to vibration is very small; in this case, if the system still classifies this as a medium-risk scenario and adjusts the weights, the monitoring of electrical / temperature parameters will be inappropriately weakened, while the actual potential risks of overcurrent or overheating will be ignored.

[0057] To address this issue, this invention further provides a method that, based on the identification of potential environmental risks using proximity scores, introduces operational context information for correlation verification. This filters out risk scenarios that are mathematically close but operationally irrelevant, ensuring that the final determined risk category is highly correlated with the actual safety requirements of the current welding operation, thus improving the accuracy of risk assessment. The specific solution is as follows: Risk proximity analysis includes: determining at least one potential risk category based on comparison results; wherein, a potential risk category is defined as one whose judgment condition is partially met; for each potential risk category, calculating the difference between its unmet judgment condition and the corresponding measured value, and assessing the degree of non-metdency of the judgment condition based on the difference; calculating the risk proximity score of the potential risk category by comprehensively considering the degree of non-metdency of all unmet judgment conditions for each potential risk category; selecting the category with the highest risk proximity score from all potential risk categories as a candidate risk category; obtaining the current operation type of the welding equipment and determining whether there is a correlation between the risk represented by the candidate risk category and the operation type; if there is, determining the candidate risk category as the risk category of the current operation scenario; if not, determining the baseline risk category as the risk category of the current operation scenario.

[0058] This approach is initiated after determining that no risk category can be fully satisfied. Its goal is to obtain a final risk category that reflects both the trends of environmental parameters and is relevant to the current operations.

[0059] In a specific implementation scenario, all predefined non-benchmark risk categories are traversed. For each non-benchmark risk category, its measured value is checked to see if it meets at least one of its judgment conditions. If it does, the non-benchmark risk category is marked as a potential risk category, and it is recorded which conditions are met and which are not.

[0060] For each potential risk category, identify all conditions in its set of criteria that are not met by the current measurement value. For each unmet condition, calculate the standardized difference between the measurement value and the corresponding safety threshold; a larger value indicates a greater distance from meeting the condition. Map the standardized difference to a dissatisfaction score; a linear or non-linear decay function can be used. A smaller standardized difference results in a higher dissatisfaction score, indicating a closer proximity to meeting the condition, i.e., a milder degree of dissatisfaction. The risk proximity score of a potential risk category is derived from the combined contributions of all its conditions (including those met and unmet). Compare the risk proximity scores of all potential risk categories and select the potential risk category with the highest score as the candidate risk category.

[0061] Finally, correlation verification is performed based on the job type to verify whether candidate risks are related to the current job. Specifically, the job type of the current welding equipment is determined through one or more of the following methods: From process parameters: parsing the parameters set by the welding power source, such as current magnitude, wire diameter, and preset program. High current, continuous welding usually corresponds to high-strength welding; low current, spot welding corresponds to repair welding. From the work order system: interfacing with the manufacturing execution system to obtain the process card information of the current welding task. Operator selection: on the welding machine's human-machine interface or the accompanying APP, the welder selects or confirms the job type before starting the operation.

[0062] A risk-job correlation database is pre-defined, identifying which job types are strongly or weakly correlated with each risk category. For example, the risk of mechanical instability (vibration causing weld detachment, equipment displacement) is strongly correlated with job types requiring high stability, such as high-altitude welding and vertical welding of large components; however, it is weakly or uncorrelated with ground platform repair welding. The risk of flammability and explosiveness is strongly correlated with all welding operations in potentially flammable environments. The risk of high-temperature heat sources is strongly correlated with welding in confined spaces and prolonged continuous welding in summer.

[0063] The system queries the correlation database to determine whether there is a known strong correlation between the risk represented by the candidate risk category and the currently acquired job type. If a strong correlation exists, the risk is determined to be real and requires attention under the current job, and the candidate risk category is determined as the final risk category for the current job scenario. If no strong correlation exists (i.e., weak correlation or no correlation), it is determined that although environmental parameters indicate a risk, it does not pose a significant threat to the specific job currently in progress, and the predefined baseline risk category is determined as the risk category for the current job scenario.

[0064] Furthermore, referring to Figure 3 As shown, the present invention addresses the adjustment of the current operating permissions of welding equipment operators by providing a solution that focuses on the most urgent node to simplify decision-making logic and achieve a rapid and clear safety response. The specific solution includes the following steps: identifying the downstream fault node with the shortest time delay from a potential fault chain; comparing the risk level of this node with a risk level threshold; if the risk level of this node is higher than the risk level threshold, adjusting the operator's current operating permissions to the restriction level corresponding to the risk level of this node; if the risk level of this node is equal to or lower than the risk level threshold, maintaining the current operating permissions unchanged.

[0065] In a specific implementation scenario, the system receives potential fault chain data from step S300. This data is a list, where each element represents a downstream fault node and contains at least three fields: node ID, risk level, and cumulative time delay. The system initializes a temporary variable, setting its cumulative time delay to a maximum value (e.g., 9999 minutes), and then iterates through all nodes in the potential fault chain. For each node encountered, its cumulative time delay is compared with the cumulative time delay in the current temporary variable. If the current node's delay is shorter, the temporary variable is updated to reflect the current node's delay. After the iteration is complete, the temporary variable stores the node with the shortest time delay. This node represents the most likely fault risk to occur first in the entire fault chain.

[0066] The system pre-defines a risk level threshold, which is a clearly defined standard, for example, set to medium (M) risk. This means that the system considers only faults of medium risk or higher to require immediate permission adjustments; while low (L) risk can be monitored but does not require immediate operation restrictions. The risk level threshold is set based on the security policy and can be configured during deployment. For ease of comparison, the risk level is quantified; for example, defining L=1, M=2, H=3; the risk level threshold is also quantified to 2 (corresponding to medium risk). The quantified risk level value R of the temporary variable is read and compared with the risk level threshold.

[0067] Branch 1: If R > risk level threshold, meaning the risk level of the shortest delay node is higher than medium risk (i.e., high risk), then it is determined that a high-risk fault is about to occur in a short time and immediate intervention is necessary. The system maintains a risk level-permission level mapping table; for example: H (high risk) → permission level restrictive operation; under this level, it may be necessary to forcibly reduce the maximum welding current, prohibit continuous welding, or require a safety officer to be present.

[0068] Branch 2: If R ≤ risk level threshold, that is, the risk level of the shortest delay node is equal to or lower than medium risk, i.e., medium or low risk, then it is determined that: the risk level of the first failure does not reach the threshold that requires immediate forced adjustment of permissions, and the current operator's permissions remain unchanged.

[0069] The solutions proposed in this invention focus on the shortest time node to ensure the system's responsiveness to the most pressing threats. All operations are based on traversal, comparison, and table lookup operations, resulting in low computational overhead, high real-time performance, and ease of stable implementation in embedded systems or server-side environments.

[0070] In the above embodiment, the propagation path reasoning of abnormal parameters based on the equipment fault knowledge graph and the construction of the equipment fault knowledge graph include: obtaining historical fault data of each component of the welding equipment and analyzing the causal relationship between faults; defining the components and their health status as graph nodes and the physical connection relationship and functional dependency relationship between components as graph edges; assigning an initial fault propagation probability weight and time delay to each edge based on historical fault statistics; and establishing a graph update mechanism to dynamically adjust the probability weight and time delay according to the actual fault propagation cases that occur during system operation.

[0071] The solution of this invention learns from the historical and real-time data of the target device, and the graph parameters continuously converge towards the true characteristics of the device, making the inference and prediction increasingly accurate. The update mechanism enables the graph to track characteristic changes caused by device aging and component replacement, maintaining the current effectiveness of the model.

[0072] In the above embodiments, the fault propagation probability weights stored in the equipment fault knowledge graph are static values ​​derived from historical fault data statistics. While reflecting long-term average trends, they fail to capture the dynamic changes in fault propagation probability due to real-time equipment operating load. Therefore, when inferring the propagation path of abnormal parameters based on these weights, risk prediction becomes distorted. Specifically, during welding operations, equipment operates under high load rates, subjecting its internal components to greater electrical, thermal, and mechanical stresses. In this situation, the likelihood of a minor initial anomaly leading to a serious fault increases significantly, and the propagation speed accelerates. If static probability weights based on average statistics across all operating conditions are used for inference, the system will underestimate the actual risk under high load, resulting in untimely or insufficient early warnings.

[0073] To address this issue, the embodiment of the present invention further includes correcting the fault propagation probability weight in the propagation path reasoning step during specific implementation. This includes: obtaining the operating load rate of the welding equipment during the traversal process; finding the corresponding correction coefficient from the preset load-probability correction curve based on the operating load rate; multiplying the static fault propagation probability weight stored in the equipment fault knowledge graph with the correction coefficient to obtain the corrected fault propagation probability weight; and using the corrected fault propagation probability weight to traverse downstream fault nodes. The load-probability correction curve satisfies the following condition: the higher the operating load rate, the larger the correction coefficient.

[0074] In specific implementation scenarios, the operating load rate is calculated as the ratio of the current output current of the welding equipment to its rated output current. Before system deployment, based on engineering experience or experimental data, a load rate-probability correction coefficient curve or mapping table is predefined. Specifically, constructing the load-probability correction curve includes: conducting welding equipment fault simulation experiments under different operating load rates, and statistically analyzing the actual probability of fault propagation along a specific path; plotting a scatter plot with the operating load rate on the x-axis and the ratio of the actual propagation probability to the static probability on the y-axis; and performing curve fitting on the scatter plot to obtain the load-probability correction curve.

[0075] The load rate-probability correction coefficient curve exhibits a monotonically increasing relationship where the correction coefficient increases with the operating load rate. A typical implementation is a piecewise function or lookup table. During inference, the system uses the operating load rate as input and quickly obtains the corresponding real-time correction coefficient by looking up a table or calculating a piecewise function. When the traversal algorithm moves along the edges of the knowledge graph, preparing to explore the next downstream node from the current node, it reads the static fault propagation probability weight stored on the edge connecting the two nodes. This fault propagation probability weight is a baseline value based on historical data statistics during graph construction. The static weight is multiplied by the real-time correction coefficient. In subsequent traversal decision logic, the original fault propagation probability weight is no longer used; instead, the corrected fault propagation probability weight is used.

[0076] The effects on the inference results are as follows: Under high load, the adjusted fault propagation probability weight increases, causing fault propagation paths related to the current load to be given higher priority during traversal or more likely to reach the threshold for triggering an early warning. This leads to the system identifying risk chains faster or issuing early warnings for the same risk earlier under high load. Under low load, the adjusted fault propagation probability weight decreases, suppressing some non-urgent propagation paths. The risk chains inferred by the system may be shorter or have a lower probability, thus reducing disruptive early warnings under safe operating conditions.

[0077] Based on the above embodiments, the cumulative time delay of each edge on the traversal path includes: starting from the starting node, recording each edge traversed along the traversal path and its corresponding time delay; adding the time delays of all edges on the path in sequence to obtain the cumulative time delay from the starting node to the current traversal node; when multiple paths reach the same node, selecting the path with the shortest cumulative time delay as the main propagation path of that node, and using its cumulative time delay as the effective time delay of that node.

[0078] The solution of this invention selects the shortest time path to determine the effective time delay. Essentially, it is based on worst-case analysis and uses the earliest possible time point of failure as a benchmark for early warning, thereby reserving the maximum time margin for safe response.

[0079] Obviously, the above embodiments are merely illustrative examples for clear explanation and are not intended to limit the implementation. Those skilled in the art will recognize that other variations or modifications can be made based on the above description. It is neither necessary nor possible to exhaustively list all possible implementations here. However, obvious variations or modifications derived therefrom are still within the scope of protection of this invention.

Claims

1. A method for safety authorization and control of electric welding operations based on the Internet of Things, characterized in that, include: The operating parameters of the welding equipment are collected through an Internet of Things (IoT) sensor network deployed on the welding equipment. Anomalies are identified based on the operating parameters to obtain abnormal parameters. Based on the equipment fault knowledge graph, the propagation path of the abnormal parameters is inferred to obtain the potential fault chain; wherein, the equipment fault knowledge graph uses the health status of the welding equipment components as nodes and the physical connection relationship and functional dependency relationship between the welding equipment components as edges, and each edge stores the probability weight and time delay of fault propagation. The propagation path reasoning includes: taking the health status of the welding equipment component corresponding to the abnormal parameter as the starting node, traversing downstream fault nodes according to the edge relationships and fault propagation probability weights in the equipment fault knowledge graph, and accumulating the time delay of each edge on the traversal path to obtain the potential fault chain. Based on the risk level of the downstream fault node in the potential fault chain and the corresponding time delay, the current operating permissions of the welding equipment operator are adjusted.

2. The IoT-based safety authorization and control method for electric welding operations according to claim 1, characterized in that, The operating parameters include electrical parameters, temperature parameters, and space environment parameters; Anomaly identification obtains the anomaly parameters including: Based on the aforementioned spatial environment parameters, the risk category of the current work scenario is determined; The electrical weight of the electrical parameter in the safety assessment and the temperature weight of the temperature parameter in the safety assessment are determined based on the risk category. The electrical parameters are weighted based on the electrical weights to obtain weighted electrical parameters; the temperature parameters are weighted based on the temperature weights to obtain weighted temperature parameters. The parameters that exceed the corresponding safety thresholds among the weighted electrical parameters and the weighted temperature parameters are identified as the abnormal parameters.

3. The IoT-based safety authorization and control method for electric welding operations according to claim 2, characterized in that, The space environment parameters include combustible gas concentration, ambient temperature, and equipment vibration amplitude. Based on the aforementioned spatial environment parameters, the risk category of the current work scenario is determined as follows: The measured values ​​of combustible gas concentration, ambient temperature, and equipment vibration amplitude are compared with the judgment conditions corresponding to each of the multiple predefined risk categories. Based on the comparison results, determine whether the current measurement value fully meets all the judgment conditions of any predefined risk category; If all conditions are met, then the predefined risk category is determined as the risk category for the current work scenario; If not all the judgment conditions for any predefined risk category are fully met, a risk proximity analysis is performed to determine the risk category of the current work scenario based on the degree of closeness between the current measured value and the judgment conditions of each predefined risk category.

4. The IoT-based safety authorization and control method for electric welding operations according to claim 3, characterized in that, The risk proximity analysis includes: Based on the comparison results, at least one potential risk category is determined; wherein, the potential risk category is defined as one whose determination criteria are partially met; For each of the potential risk categories, calculate the difference between the unmet judgment condition and the corresponding measured value, and assess the degree of non-metdness of the judgment condition based on the difference; By combining the degree of non-compliance of all unmet judgment conditions for each potential risk category, the risk proximity score of the potential risk category is calculated; Select the category with the highest risk approximation score from all the potential risk categories as the candidate risk category; Obtain the current operation type of the welding equipment and determine whether there is a correlation between the risk represented by the candidate risk category and the operation type; If such a risk category exists, it will be determined as the risk category for the current work scenario. If it does not exist, the baseline risk category will be determined as the risk category of the current work scenario.

5. The IoT-based safety authorization and control method for electric welding operations according to claim 1, characterized in that, Based on the risk level of the downstream failure node in the potential failure chain and the corresponding time delay, adjust the current operating permissions of the welding equipment operator, including: Identify the downstream fault node with the shortest time delay from the potential fault chain; Compare the risk level of this node with the risk level threshold; If the risk level of the node is higher than the risk level threshold, the operator's current operating permissions will be adjusted to the restriction level corresponding to the risk level of the node. If the risk level of the node is equal to or lower than the risk level threshold, the current operating permissions will remain unchanged.

6. The IoT-based safety authorization and control method for electric welding operations according to claim 1, characterized in that, Constructing the equipment fault knowledge graph includes: Acquire historical fault data of each component of the welding equipment and analyze the causal relationships between the faults; Components and their health status are defined as graph nodes, and the physical connections and functional dependencies between components are defined as graph edges; Based on historical fault statistics, each edge is assigned an initial fault propagation probability weight and time delay. A graph update mechanism is established to dynamically adjust the probability weights and time delays based on actual fault propagation cases that occur during system operation.

7. The IoT-based safety authorization and control method for electric welding operations according to claim 1 or 6, characterized in that, The propagation path reasoning step also includes correcting the fault propagation probability weights, which includes: During the traversal, the operating load rate of the welding equipment is obtained, and the corresponding correction coefficient is found from the preset load-probability correction curve based on the operating load rate. The corrected fault propagation probability weight is obtained by multiplying the static fault propagation probability weight stored in the equipment fault knowledge graph with the correction coefficient. Use the corrected fault propagation probability weights to traverse downstream fault nodes; The load-probability correction curve satisfies the following condition: the higher the operating load rate, the larger the correction coefficient.

8. The IoT-based safety authorization and control method for electric welding operations according to claim 7, characterized in that, Constructing the load-probability correction curve includes: Welding equipment failure simulation experiments were conducted under different operating load rates to statistically analyze the actual probability of failure propagation along a specific path. Plot a scatter plot with the operating load rate on the x-axis and the ratio of the actual propagation probability to the static probability on the y-axis. Curve fitting is performed on the scattered points to obtain the load-probability correction curve.

9. The IoT-based safety authorization and control method for electric welding operations according to claim 1, characterized in that, The time delay for traversing each edge of the path includes: Starting from the starting node, record each edge traversed along the traversal path and its corresponding time delay; The time delays of all edges on the path are added together to obtain the cumulative time delay from the starting node to the currently traversed node.

10. The IoT-based safety authorization and control method for electric welding operations according to claim 9, characterized in that, When multiple paths lead to the same node, the path with the shortest cumulative time delay is selected as the main propagation path of that node, and its cumulative time delay is taken as the effective time delay of that node.

Images