Welding temperature feedback control method based on multiple sensors and related equipment

Abstract

The invention provides a multi-sensor-based welding temperature feedback control method and related equipment, and the method comprises the steps: carrying out the preprocessing of multi-dimensional temperature data collected in a welding region, and obtaining a multi-dimensional temperature feature vector; performing weighted fusion on the multi-dimensional temperature feature vectors to obtain an effective temperature feedback result of the welding area; based on the effective temperature feedback result, the environment temperature data and the welding speed, a target temperature curve is generated through a preset fuzzy rule; and iterative learning is carried out according to the target temperature curve and the effective temperature feedback result, temperature errors of all sampling moments are determined, and a welding output instruction is updated according to the temperature errors. Through the implementation of the scheme, fusion and weighting processing are performed on the multi-dimensional temperature data, and the effective temperature feedback result of the welding area can be generated, so that precise iterative adjustment of welding heat source output is realized, and the accuracy of heat source power regulation and control is effectively improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of welding, in particular to a welding temperature feedback control method based on multiple sensors and related equipment. BACKGROUND

[0002] With the development of high-end equipment manufacturing and energy industry, automatic welding is widely used in aerospace, nuclear power equipment and large ship manufacturing fields. In these high-precision welding tasks, the weld quality directly affects the structural strength and safety performance. The traditional welding process mainly relies on single sensor temperature detection, such as infrared temperature measurement or thermocouple. However, single sensor has problems such as slow response speed, limited measurement position and sensitive to environmental interference, which is difficult to completely capture the dynamic temperature distribution of the welding area. Especially under the conditions of complex weld geometry and multi-layer plate thickness material, single sensor is difficult to provide effective temperature information of the key area of welding, resulting in lag or inaccurate power regulation of heat source, thereby affecting the stability of the welding process and the quality of the weld. SUMMARY

[0003] The present application provides a welding temperature feedback control method based on multiple sensors and related equipment, which solves the problem that single sensor in related technology is difficult to provide effective temperature information of the key area of welding, resulting in lag or inaccurate power regulation of heat source.

[0004] The first aspect of the present application provides a welding temperature feedback control method based on multiple sensors, which comprises: Pretreating the infrared temperature measurement signal, thermocouple signal, two-dimensional temperature field image data and environmental temperature data collected in the welding area with instantaneous high temperature and high gradient to obtain a multi-dimensional temperature feature vector; Obtaining the effective temperature feedback result of the welding area by weighted fusion of the multi-dimensional temperature feature vector, wherein the weight of weighted fusion is dynamically allocated based on the credibility coefficient of each sensor in the welding dynamic process; Generating a target temperature curve based on the effective temperature feedback result, the environmental temperature data and the real-time welding speed through a preset fuzzy rule; Determining the temperature error sequence of each sampling time according to the target temperature curve and the effective temperature feedback result, performing sliding average and deviation accumulation analysis on the temperature error sequence, and obtaining an error trend feature set; Dynamically adjusting the learning gain parameter and the constraint coefficient in the iterative learning control law according to the error trend feature set, calculating the power correction amount of the welding heat source according to the adjusted control law, and updating the welding output instruction.

[0005] Optionally, in the first implementation of the first aspect of this application, before the step of preprocessing the infrared temperature measurement signal, thermocouple signal, two-dimensional temperature field image data, and ambient temperature data collected in the welding area with instantaneous high temperature and high gradient to obtain a multidimensional temperature feature vector, the method further includes: Based on the geometric layout of the welding area and the distribution of the heat-affected zone of the weld, the placement and sampling period of the sensors for collecting infrared temperature measurement signals, thermocouple signals, two-dimensional temperature field image data, and ambient temperature data are determined; wherein, the sampling period is dynamically configured according to the instantaneous thermal change rate of the welding process. Based on the deployment location and sampling period, the sensors in each dimension are controlled to synchronously collect temperature data to obtain the raw temperature data. The outputs of sensors in various dimensions were compared with those of a standard blackbody temperature source under standard welding test conditions. The original temperature data is cross-calibrated based on the comparison output to obtain multi-dimensional temperature data composed of the infrared temperature measurement signal, the thermocouple signal, the two-dimensional temperature field image data, and the ambient temperature data.

[0006] Optionally, in the second implementation of the first aspect of this application, the step of preprocessing the infrared temperature measurement signal, thermocouple signal, two-dimensional temperature field image data, and ambient temperature data collected in the welding area with instantaneous high temperature and high gradient to obtain a multidimensional temperature feature vector includes: Based on the instantaneous high temperature characteristics of the welding process, the infrared temperature measurement signal is dynamically range-adapted and saturation region-recovered, while the thermocouple signal is dynamically response time-compensated to generate a synchronized denoised temperature sequence. Based on the high temperature gradient distribution characteristics of the welding area, the two-dimensional temperature field image is segmented based on molten pool contour recognition, and the average temperature, temperature gradient field and high temperature area change rate of the molten pool area, heat-affected zone and base material area are calculated to generate a region feature set. The ambient temperature data is normalized, and dynamic baseline adjustment is applied to the normalization parameters to obtain the adjusted normalized environmental parameters. The denoised temperature sequence, the regional feature set, and the normalized environmental parameters are spatiotemporally aligned and correlated according to the welding time sequence to generate a multidimensional temperature feature vector.

[0007] Optionally, in the third implementation of the first aspect of this application, the step of obtaining the effective temperature feedback result of the welding area by weighted fusion of the multidimensional temperature feature vector includes: Based on the time window data of the denoised temperature sequence, the variance and drift rate of the infrared temperature measurement signal and the thermocouple signal in adjacent sampling periods are determined to obtain a set of noise evaluation indicators. The reliability coefficients of the infrared temperature measurement signal and the thermocouple signal under the current welding process stage are dynamically evaluated by combining the noise evaluation index set with the average temperature, maximum temperature gradient, and normalized environmental parameters of the regional feature set. The welding process stage includes at least the arc initiation stage, the stable welding stage, and the arc termination stage. Obtain a weight template library corresponding to the welding process stage, select a benchmark weight distribution from the weight template library according to the currently identified process stage, and adjust the benchmark weight in real time according to the dynamic confidence coefficient sequence to obtain a weight matrix; The multidimensional temperature feature vector is weighted and fused according to the weight matrix to obtain the effective temperature feedback result of the welding area.

[0008] Optionally, in the fourth implementation of the first aspect of this application, the step of generating a target temperature curve based on the effective temperature feedback result, the ambient temperature data, and the real-time welding speed using preset fuzzy rules includes: The current welding heat input is determined based on the real-time welding speed and the effective temperature feedback result, and the welding heat input is compensated for by heat dissipation based on the ambient temperature data to obtain the compensated welding heat input. Based on the material type, geometry type of the workpiece to be welded, and the compensated welding heat input, the initial process curve corresponding to the material type and geometry type is retrieved from the preset process database; The initial process curve is segmented to extract the temperature setpoints and time nodes of each stage, and the temperature difference between the temperature setpoint and the effective temperature feedback result in the same time interval is determined. The temperature difference, the compensated welding heat input, the ambient temperature data, and the workpiece heat capacity parameters are combined into a fuzzy rule input vector. The fuzzy rule input vector is input into a preset fuzzy rule library, and fuzzification processing and operations based on the inference rule table are performed in sequence to output the set of correction coefficients for the temperature setpoint and the set of scaling coefficients for the time node. The segmented temperature setpoints and time nodes are parametrically adjusted according to the set of correction coefficients and the set of scaling coefficients, and the adjusted segmented temperature setpoints are reorganized into the target temperature curve in chronological order.

[0009] Optionally, in the fifth implementation of the first aspect of this application, after the steps of determining the temperature error sequence at each sampling time based on the target temperature curve and the effective temperature feedback result, performing moving average and deviation accumulation analysis on the temperature error sequence, and obtaining the error trend feature set, the method further includes: The error trend feature set is correlated with the confidence coefficient to obtain a dynamic adjustment parameter set. The welding output command of the previous cycle is updated according to the dynamically adjusted parameter set to obtain the welding output command of the current cycle.

[0010] Optionally, in the sixth implementation of the first aspect of this application, the step of updating the welding output command of the previous cycle according to the dynamically adjusted parameter set to obtain the welding output command of the current cycle includes: By performing point-by-point correlation calculations between the temperature error sequence and the dynamic adjustment parameter set, the power correction factor sequence corresponding to each sampling time is determined. The power correction curve is generated by performing differential calculations based on the power correction factor sequence and the power output value at each moment of the welding output command in the previous cycle. The welding output command of the previous cycle is updated according to the power correction curve to obtain the welding output command of the current cycle.

[0011] A second aspect of this application provides a welding temperature feedback control device based on multiple sensors. This device is used to implement a welding temperature feedback control method based on multiple sensors. The welding temperature feedback control device based on multiple sensors includes: The processing module is used to preprocess infrared temperature measurement signals, thermocouple signals, two-dimensional temperature field image data and ambient temperature data collected in the welding area with instantaneous high temperature and high gradient to obtain multi-dimensional temperature feature vectors. The fusion module is used to obtain the effective temperature feedback result of the welding area by weighted fusion of the multi-dimensional temperature feature vector; wherein, the weight of the weighted fusion is dynamically allocated based on the credibility coefficient of each sensor in the welding dynamic process; The generation module is used to generate a target temperature curve based on the effective temperature feedback result, the ambient temperature data, and the real-time welding speed, using preset fuzzy rules. The analysis module is used to determine the temperature error sequence at each sampling time based on the target temperature curve and the effective temperature feedback result, and to perform moving average and deviation accumulation analysis on the temperature error sequence to obtain the error trend feature set. The update module is used to dynamically adjust the learning gain parameter and constraint coefficient in the iterative learning control law according to the error trend feature set, calculate the power correction amount of the welding heat source according to the adjusted control law, and update the welding output command.

[0012] A third aspect of this application provides an electronic device, including a memory and a processor, wherein the processor is configured to execute a computer program stored in the memory, and when the processor executes the computer program, it implements the steps of the multi-sensor-based welding temperature feedback control method provided in the first aspect of this application.

[0013] The fourth aspect of this application provides a computer-readable storage medium storing a computer program thereon. When the computer program is executed by a processor, it implements the steps of the multi-sensor-based welding temperature feedback control method provided in the first aspect of this application.

[0014] In summary, the welding temperature feedback control method and related equipment based on multi-sensor provided in this application preprocess the multi-dimensional temperature data collected in the welding area to obtain a multi-dimensional temperature feature vector; by weighted fusion of the multi-dimensional temperature feature vector, an effective temperature feedback result for the welding area is obtained; based on the effective temperature feedback result and the ambient temperature data of the multi-dimensional temperature data, a target temperature curve is generated through preset fuzzy rules; the temperature error sequence at each sampling time is determined according to the target temperature curve and the effective temperature feedback result, and the temperature error sequence is subjected to moving average and deviation accumulation analysis to obtain an error trend feature set; the learning gain parameter and constraint coefficient in the iterative learning control law are dynamically adjusted according to the error trend feature set, the power correction of the welding heat source is calculated according to the adjusted control law, and the welding output command is updated. Through the implementation of this application, the fusion and weighted processing of multi-dimensional temperature data can generate an effective temperature feedback result for the welding area, thereby achieving precise iterative adjustment of the welding heat source output and effectively improving the accuracy of heat source power control. Attached Figure Description

[0015] Figure 1 A schematic flowchart of a welding temperature feedback control method based on multiple sensors provided in an embodiment of this application; Figure 2 A schematic diagram of the program module of the welding temperature feedback control device based on multiple sensors provided in the embodiments of this application; Figure 3 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Detailed Implementation

[0016] To make the inventive objectives, features, and advantages of this application more apparent and understandable, the technical solutions in the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0017] To address the problem in related technologies where a single sensor cannot provide effective temperature information for critical welding zones, leading to lag or inaccuracy in heat source power control, embodiments of this application provide a welding temperature feedback control method based on multiple sensors, such as... Figure 1 This is a flowchart illustrating the welding temperature feedback control method based on multiple sensors provided in this embodiment. The welding temperature feedback control method based on multiple sensors includes the following steps: Step 110: Preprocess the infrared temperature measurement signal, thermocouple signal, two-dimensional temperature field image data and ambient temperature data collected in the welding area with instantaneous high temperature and high gradient to obtain a multidimensional temperature feature vector.

[0018] Specifically, during welding, the molten pool and its surrounding area exhibit instantaneous high temperatures and strong spatial temperature gradients, placing higher demands on the data acquisition and combined use of different types of sensors. To obtain input information that reflects the true thermal state, when preprocessing infrared thermometry signals, thermocouple signals, two-dimensional temperature field image data, and ambient temperature data, corresponding compensation mechanisms can be designed to address issues such as range jumps, local overexposure, thermocouple response delays, and image brightness saturation caused by instantaneous high temperatures. For example, by applying dynamic range adaptation to infrared signals, continuity can be maintained and measurement compression avoided during rapid temperature increases; simultaneously, dynamic response time compensation can be applied to thermocouple signals to reduce inherent hysteresis under high-temperature rapid-rise conditions. By detecting the contour morphology of the molten pool and dividing the region according to its edge gradient, representative information such as the average temperature, gradient field, and rate of change of the high-temperature area in the molten pool region, heat-affected zone, and base material region can be obtained from the two-dimensional temperature field image. After synchronization and alignment using a unified timestamp, the denoised temperature sequence, regional feature information, and ambient temperature parameters are correlated and integrated according to the welding sequence to form a multi-dimensional temperature feature vector.

[0019] In one optional embodiment of this example, before preprocessing the infrared temperature measurement signals, thermocouple signals, two-dimensional temperature field image data, and ambient temperature data collected in the welding area with instantaneous high temperature and high gradient to obtain a multi-dimensional temperature feature vector, the method further includes: determining the deployment positions and sampling periods of the sensors for collecting infrared temperature measurement signals, thermocouple signals, two-dimensional temperature field image data, and ambient temperature data according to the geometric layout of the welding area and the distribution of the weld heat-affected zone; wherein the sampling period is dynamically configured according to the instantaneous thermal change rate of the welding process; controlling the sensors of each dimension to synchronously collect temperature data according to the deployment positions and sampling periods to obtain raw temperature data; collecting the comparison outputs of the sensors of each dimension and a standard blackbody temperature source under standard welding test conditions; and cross-calibrating the raw temperature data according to the comparison outputs to obtain multi-dimensional temperature data composed of infrared temperature measurement signals, thermocouple signals, two-dimensional temperature field image data, and ambient temperature data.

[0020] In this embodiment, when arranging sensors in the welding area, the specific positions and sampling periods of sensors in each dimension need to be determined based on the geometry of the weldment and the distribution of the heat-affected zone (HAZ) of the weld. The HAZ refers to the area where the temperature around the material rises significantly under the influence of laser or other heat sources. The size and shape of this area are closely related to the thickness of the weldment, the thermal conductivity of the material, and the welding power. Therefore, the sensors should be placed in positions that can both cover the core of the weld and reflect the surrounding thermal gradient, thereby ensuring that the collected data can completely describe the thermal state during the welding process. When heated, the weld typically exhibits a thermal field pattern of "high temperature at the center and decreasing circumferentially." The extent of its HAZ is closely related to the bevel shape, plate thickness, welding speed, and welding process. The geometric arrangement directly affects the accessibility of the infrared sensor to the viewpoint above the molten pool and whether the thermocouple embedding position can capture the true internal temperature changes. Therefore, different spatial locations along the longitudinal path of the weld, covering the center of the molten pool, the boundary of the heat-affected zone, and the base material area, can be selected to deploy infrared temperature measurement units and thermocouple sensing elements. The optical axis of the two-dimensional temperature field acquisition unit should pass through the central region of the weld, and suitable viewing angles (either frontal or tilted) should be set to ensure that optical imaging covers multiple high-gradient regions near the molten pool. Furthermore, ambient temperature sensors can be placed at a certain distance from the weld to avoid the influence of radiant heat and maintain stable measurements of the ambient background temperature. After determining the spatial placement, the sampling period of various sensors can be dynamically configured according to the instantaneous thermal change rate during the welding process. Instantaneous thermal change rate can be understood as the amount of temperature change per unit time. For example, in high-energy-density welding processes, there can be drastic temperature rises on the order of milliseconds. Therefore, infrared thermometry and two-dimensional temperature field acquisition can use a higher sampling frequency to capture rapid temperature changes and molten pool morphology changes. Thermocouples, due to their embedded contact measurement response characteristics, can have their sampling period adjusted appropriately based on the material's thermal diffusion rate and embedding depth. Ambient temperature acquisition, not being directly located in a high-gradient region, can use a lower sampling rate. However, when welding speed changes or ventilation disturbances are strong, the sampling rate can be automatically increased based on the trend to improve timeliness. For example, when an increase in welding speed leads to a faster change in molten pool temperature, the rate of temperature change at the most recent sampling points can be calculated to determine if the thermal change rate has reached a threshold. If it has, the sampling period of the infrared sensor and thermal imaging sensor is automatically shortened to adapt the acquisition frequency to the new rate of change. Once the deployment location and sampling period are determined, all sensors can be controlled to synchronously acquire temperature data according to the preset synchronization clock signal. This ensures that the infrared temperature sequence, thermocouple temperature sequence, two-dimensional temperature field image sequence, and ambient temperature sequence are strictly aligned on the time axis, thereby avoiding data mismatch caused by time offset in subsequent correlation calculations.Based on the raw data obtained through synchronous acquisition, to ensure that the temperature data from different sensors have unified dimensions and comparability, the output of each sensor can be collected using a standard blackbody temperature source under standard welding test conditions. A blackbody temperature source is a radiator with a surface emissivity close to 1; its surface radiation temperature has a definite physical correspondence with the actual temperature of an object, thus serving as a unified calibration benchmark. By sequentially acquiring infrared temperature measurement signals, thermocouple outputs, electrical imaging data, and environmental sensor feedback at different constant temperature points of the blackbody temperature source, the output curves of each sensor at known true temperatures can be obtained. By fitting the output data, the response deviation, range compression region, and nonlinearity interval of each sensor can be obtained, providing basic data for cross-calibration. Cross-calibration refers to comparing the output differences of different sensors under the same true temperature conditions within a unified temperature reference framework and performing data correction to ensure that infrared data, thermocouple data, two-dimensional image temperature data, and environmental temperature data have unified temperature dimensions and offset compensation capabilities. For example, when infrared sensors are prone to saturation in high-temperature regions while thermocouples provide more stable indications at the same temperature, thermocouple outputs can be used to perform restorative correction of the infrared signal in the saturation region. Similarly, when thermal imaging systems exhibit brightness shifts in low-temperature regions, scaling can be applied based on infrared temperature measurements. This cross-calibration method ensures consistency in the resulting multidimensional temperature data when used for fusion or feature extraction, making it more accurately reflect the true thermal state of the welding area under instantaneous high temperatures and rapid gradient changes.

[0021] In one optional embodiment of this example, the step of preprocessing infrared temperature measurement signals, thermocouple signals, two-dimensional temperature field image data, and ambient temperature data collected in a welding area with instantaneous high temperature and high gradient to obtain a multidimensional temperature feature vector includes: performing dynamic range adaptation and saturation zone recovery processing on the infrared temperature measurement signals according to the instantaneous high temperature characteristics of the welding process, and simultaneously performing dynamic response time compensation on the thermocouple signals to generate a synchronous denoised temperature sequence; performing region segmentation on the two-dimensional temperature field image based on molten pool contour recognition according to the high temperature gradient distribution characteristics of the welding area, and calculating the average temperature, temperature gradient field, and high temperature area change rate of the molten pool area, heat-affected zone, and base material area to generate a region feature set; performing normalization processing on the ambient temperature data, and applying dynamic baseline adjustment to the normalization parameters to obtain adjusted normalized environmental parameters; and performing spatiotemporal alignment and correlation integration of the denoised temperature sequence, region feature set, and normalized environmental parameters according to the welding time sequence to generate a multidimensional temperature feature vector.

[0022] In this embodiment, during the preprocessing of multi-source temperature data of the welding area, the molten pool experiences a rapid temperature rise and forms a strong spatial temperature gradient within a very short time. Therefore, the infrared temperature measurement signal, thermocouple signal, and two-dimensional temperature field image data are affected by factors such as range limitations, response lag, and image brightness distortion during the initial acquisition. To dynamically adapt the infrared temperature measurement signal to its range, the signal characteristics of the infrared detector output near its maximum measurable temperature range can be continuously monitored during sampling. The range mapping relationship of the infrared signal is actively adjusted based on the signal rise rate and peak value change trend to prevent compression effects during rapid temperature increases. Furthermore, when the signal reaches the detector's saturation limit in the high-temperature region, the true temperature change direction near the saturation point can be recovered by referring to the previous unsaturated value change trend, the gradient change in the infrared signal neighborhood, and the internal temperature change provided by the thermocouple. For example, when multiple consecutive sampling points of the infrared signal at a certain location reach the saturation range, and the rate of change of brightness in the thermocouple signal and the two-dimensional temperature image in that area still shows a significant upward trend, the saturation region can be recovered based on the change ratio of the signal in the unsaturated region, allowing the mapped temperature curve to continue reflecting the continuous heating process at the center of the molten pool. Furthermore, as a contact sensor, the thermocouple's output is affected by the metal's thermal conductivity and its own heat capacity, resulting in a response delay. Therefore, its temperature change over a short period can be recorded, and the rapid temperature rise characteristics reflected in the infrared signal and the two-dimensional image can be used as a reference. By shortening the time offset between the thermocouple signal and the actual change, it can be made consistent with the infrared data on the time scale, ultimately forming a synchronized denoised temperature sequence. During the processing of the two-dimensional temperature field image, due to the significant temperature difference between the molten pool and the surrounding materials, its boundary exhibits stable brightness abrupt changes in the image. Therefore, the molten pool contour can be identified based on the brightness gradient change. During the identification process, the brightness abrupt change locations can be detected along multiple directions of the image, and a closed molten pool contour curve can be constructed by finding the boundary point where the gradient is maximum. After identifying the molten pool region, the heat-affected zone (HAZ) can be obtained by expanding outward layer by layer based on the brightness changes around the molten pool. The remaining region is then divided into the base material region, thus forming a high-temperature region structure with spatial hierarchy. For each of these regions, the average temperature, local gradient direction, gradient magnitude, and the rate of change of the high-temperature region area over time can be calculated. For example, when the welding current increases, causing the molten pool volume to expand, the area of ​​the molten pool contour will show a continuous increase in adjacent images. By detecting the relative change in the area of ​​the molten pool region, the rate of change of the high-temperature region area can be obtained. This allows the construction of a regional feature set reflecting the spatial temperature distribution and dynamic changes.To ensure dimensional consistency of ambient temperature data when used in conjunction with multi-source temperature sequences, normalization is performed. To eliminate slow drift caused by environmental disturbances during welding, dynamic baseline adjustment is incorporated into the normalization process. The ambient temperature is stabilized using the average value over a period as a baseline, and minor compensation is applied to the normalization result based on the direction of temperature change, ensuring it reflects the true trend of ambient temperature variation during welding. If changes in ventilation conditions at the welding site cause a slow rise in ambient temperature, short-term temperature increments can be recorded and mapped to the compensation term of the normalization parameters. This ensures the final normalized environmental parameters maintain stability while reflecting the trend. When aligning the denoised temperature sequence, regional feature set, and normalized environmental parameters according to the welding sequence, a unified sampling timestamp can be used as a matching benchmark to correlate and integrate data collected by different sensors at the same welding moment. Because infrared thermometry, thermocouples, and two-dimensional images have different sampling periods, various data can be mapped onto a unified time axis through interpolation or by using the time position of the most recent sampling point. This ensures that each sampling moment corresponds to a synchronized temperature sequence, regional characteristics, and environmental parameters. During the integration process, spatial characteristics can be correlated with the time series based on welding speed, weld location, and heat-affected zone size. Ultimately, a multi-dimensional temperature feature vector containing temporal temperature variations, spatial distribution characteristics, and environmental background information can be constructed, comprehensively describing the impact of instantaneous high temperatures, high gradients, and environmental changes on the welding thermal process.

[0023] Step 120: By weighted fusion of the multidimensional temperature feature vectors, the effective temperature feedback result of the welding area is obtained.

[0024] Specifically, the weights for weighted fusion are dynamically allocated based on the reliability coefficients of each sensor during the welding process. After the multidimensional temperature feature vector is constructed, the reliability coefficients of each input feature can be dynamically determined according to the differences in the response capabilities of different sensors under instantaneous high temperature and high gradient conditions during the welding process. The reliability is obtained based on the variance, drift rate, and matching degree with the regional gradient features of the sensor within the sliding time window, while also considering different stages of the welding process (such as the strong impact at arc initiation, the high-speed heating stage, the low-fluctuation range of the stable welding stage, and the temperature decay range of the arc termination stage) and selecting corresponding benchmark weights from the weight template library. By correlating and adjusting the benchmark weights with the real-time reliability coefficients, a weight matrix that can adapt to the dynamic welding process can be obtained. The weight matrix is ​​then applied to the multidimensional temperature feature vector for fusion to obtain an effective temperature feedback result that reflects the overall thermal state.

[0025] In an optional implementation of this embodiment, the step of obtaining the effective temperature feedback result of the welding area by weighted fusion of multi-dimensional temperature feature vectors includes: determining the variance and drift rate of the infrared temperature measurement signal and the thermocouple signal in adjacent sampling periods based on the time window data of the denoised temperature sequence to obtain a noise evaluation index set; dynamically evaluating the reliability coefficient of the infrared temperature measurement signal and the thermocouple signal under the current welding process stage by combining the noise evaluation index set with the average temperature, maximum temperature gradient, and normalized environmental parameters of the regional feature set, wherein the welding process stage includes at least the arc initiation stage, the stable welding stage, and the arc termination stage; obtaining a weight template library corresponding to the welding process stage, selecting a benchmark weight distribution from the weight template library according to the currently identified process stage, and adjusting the benchmark weight in real time according to the dynamic reliability coefficient sequence to obtain a weight matrix; and weighted fusion of the multi-dimensional temperature feature vectors based on the weight matrix to obtain the effective temperature feedback result of the welding area.

[0026] In this embodiment, when weighted fusion of multidimensional temperature feature vectors to obtain effective temperature feedback results for the welding area, the fluctuation of infrared temperature measurement signals and thermocouple signals in adjacent sampling periods can be analyzed based on the changes in the denoised temperature sequence within a continuous time window. To this end, several consecutive sampling points can be considered as an observation interval, within which the dispersion of temperature values ​​is calculated, allowing the magnitude of temperature fluctuations to be characterized in the form of variance. Simultaneously, by comparing the mean of adjacent windows or the overall trend of temperature changes within a window, the temporal offset rate of the temperature signal, i.e., the drift rate, can be obtained. This indicator reflects whether the sensor gradually deviates from the true temperature in a high-temperature environment. For example, when an infrared detector experiences output offset under prolonged high-energy radiation, its drift rate will continuously increase, indicating that its reliability should be reduced at this stage. Combining variance and drift rate forms a noise evaluation index set, which can simultaneously reflect the characteristics of instantaneous fluctuations and long-term offsets. After obtaining the noise evaluation index set, the reliability of infrared temperature measurement signals and thermocouple signals can be dynamically evaluated by combining the average temperature, maximum temperature gradient, and normalized environmental parameters of the regional feature set. Average temperature reflects the overall heating level of a region, while the maximum temperature gradient reflects the rate of thermal change and the drastic changes near the molten pool boundary. These spatial characteristics, along with ambient temperature, determine the applicability of the sensor at a specific moment. For example, when the maximum temperature gradient of the region is large, infrared temperature measurement signals have a faster response capability in processing spatial temperature changes above the molten pool, while thermocouples experience hysteresis due to the presence of internal conduction paths. In this case, the reliability coefficient can be biased towards the infrared signal. Conversely, when welding enters a stable stage with a lower gradient, infrared temperature measurement is more susceptible to interference from metal vapor and arc light. At this time, the internal temperature signal provided by the thermocouple has a greater advantage in stability, and its reliability coefficient can be improved. Since the welding process can be divided into the arc initiation stage, the stable welding stage, and the arc termination stage, and the thermal behavior patterns of different stages are drastically different, the reliability assessment process needs to be combined with the current welding process stage so that the reliability sequence can be automatically adjusted as the stage changes. To make the weighted fusion conform to the thermal distribution characteristics of different process stages, a weight template library corresponding to each welding process stage can be pre-constructed. A weighted template library can be understood as a priori distribution of the relationship between infrared temperature measurement signals, thermocouple signals, and regional image features. Through the induction of a large number of welding experiments or process experience, representative benchmark weights can be established for the arc initiation, stable welding, and arc termination stages. For example, in the arc initiation stage, the molten pool has not yet formed and the temperature fluctuation is relatively violent, so the weight of the infrared temperature measurement signal in the molten pool area in the weighted template can be set higher; in the stable welding stage, the molten pool volume is fixed and the heat conduction path is stable, so the weight of the thermocouple signal at the average temperature level can be relatively increased; in the arc termination stage, the heat gradually decreases and the surface radiation changes are unstable, so the weight of the regional image temperature field can be automatically increased according to the temperature gradient change.After identifying the current welding stage, the corresponding baseline weight distribution can be selected from the weight template library. Then, the baseline weights are adjusted in real-time according to the aforementioned confidence coefficient, ensuring that the final weight matrix retains the characteristics of the process stage while adapting to the signal state at the current moment. After obtaining the weight matrix, it can be applied to the pre-constructed multi-dimensional temperature feature vector, ensuring that the data from each sensor dimension has an impact level matching the current operating conditions during the fusion process. Weighted fusion can be achieved by multiplying the weight matrix element-wise and aligning the results over time, so that the final output not only reflects the instantaneous temperature value but also embodies the comprehensive relationship between the time series and spatial regional characteristics. For example, when the molten pool experiences abnormal expansion at a certain moment and the area change rate given by the two-dimensional temperature field image increases significantly, the weight matrix will automatically increase the weight of the image information at this moment based on dynamic confidence, ensuring that the effective temperature feedback can reflect this change in a timely manner. The effective temperature feedback result fused in this way can then be used for subsequent target temperature curve generation and welding control, enabling the entire control process to accurately reflect the actual welding state under instantaneous high temperature and strong gradient conditions.

[0027] Step 130: Based on the effective temperature feedback results, ambient temperature data, and real-time welding speed, generate the target temperature curve through preset fuzzy rules.

[0028] Specifically, after obtaining effective temperature feedback, the system can further combine ambient temperature, workpiece material and geometric parameters, and real-time welding speed to retrieve an initial process curve matching the current workpiece from the process database. Based on the weld's behavior at different thermal stages, the initial curve is segmented and compared with the effective temperature feedback within the corresponding time intervals to obtain temperature difference information for each stage. Using the temperature difference, ambient temperature, and workpiece heat capacity parameters as inputs, a fuzzy rule base is used to calculate membership degrees and perform inference operations, thereby generating corresponding correction coefficients. By applying these correction coefficients to the temperature setpoints and time nodes of each segment, a target temperature curve reflecting the current actual thermal environment and dynamic welding speed characteristics can be constructed.

[0029] In one optional implementation of this embodiment, the step of generating a target temperature curve based on effective temperature feedback results, ambient temperature data, and real-time welding speed using preset fuzzy rules includes: determining the current welding heat input based on the real-time welding speed and effective temperature feedback results, and performing heat dissipation compensation on the welding heat input based on ambient temperature data to obtain a compensated welding heat input; retrieving an initial process curve corresponding to the material type and geometry type of the welding workpiece from a preset process database based on the material type, geometry type, and compensated welding heat input; extracting the temperature setpoints and time nodes of each stage of the initial process curve in segments, and determining the temperature difference between the temperature setpoint and the effective temperature feedback result in the same time interval; combining the temperature difference, compensated welding heat input, ambient temperature data, and workpiece heat capacity parameters into a fuzzy rule input vector; inputting the fuzzy rule input vector into a preset fuzzy rule library, sequentially performing fuzzification processing and calculations based on the inference rule table, and outputting a set of correction coefficients for the temperature setpoints and a set of scaling coefficients for the time nodes; parametrically adjusting the segmented temperature setpoints and time nodes based on the set of correction coefficients and the set of scaling coefficients, and recombining the adjusted segmented temperature setpoints into a target temperature curve in chronological order.

[0030] In this embodiment, under conditions of instantaneous high temperature and high temperature gradient in the welding area, to obtain a process control benchmark that reflects the actual thermal state, the current welding heat input can be determined based on the real-time welding speed and effective temperature feedback results. Welding heat input can be understood as the energy absorbed per unit length of weld seam, and the welding speed directly determines the duration of energy action on the weld seam. Therefore, as the welding speed increases, the actual heat input decreases, causing the molten pool temperature to show a contraction trend; conversely, as the speed decreases, the molten pool temperature and the heat-affected zone expand. To ensure that the heat input calculation process reflects the dynamics of the welding process, the effective temperature feedback curve can be correlated with the real-time welding speed. The effective distribution of heat input can be determined by observing the temperature change trend and temperature gradient change rate in the molten pool area. For example, when the temperature feedback results show that the temperature in the central region of the molten pool decreases and the maximum temperature gradient increases, it can be inferred that the heat input is insufficient relative to the current welding speed, thereby increasing the heat input estimate in the calculation. Furthermore, by combining ambient temperature data with heat dissipation compensation for the heat input, the heat input not only reflects the welding power output but also the influence of external environmental heat dissipation conditions. When the ambient temperature is low, heat dissipates from the workpiece surface to the outside more quickly, and the compensated heat input should be increased to offset the dissipation. When the ambient temperature rises or there are radiant heat sources in the surrounding environment, the compensation amount can be appropriately reduced to make the compensated welding heat input closer to the actual thermal behavior. After obtaining the compensated welding heat input, the initial process curve matching the material type and geometry of the workpiece can be retrieved from the preset process database. The process database can store heating, holding, and cooling curves for various materials and structural forms. For example, steel plates have high thermal conductivity and a high melting point, so their initial process curves will show a steeper heating section. Aluminum, due to its high thermal conductivity, heats up and cools down more quickly, so its curve shape will be different. Different geometries also affect the heat diffusion pattern. For example, heat diffusion is slower when welding thick plates, while heat diffuses more easily along the plane when welding thin plates. The retrieval process selects the initial curve that best reflects the thermal behavior based on the material type and geometric information. After obtaining the initial process curve, the curve can be segmented according to the heating, steady-state, and cooling stages of the welding process. Corresponding temperature setpoints and time nodes can be extracted from each segment to ensure the curve structure corresponds to the actual welding process. By comparing the temperature difference between the temperature setpoint and the effective temperature feedback within the same time interval, the deviation between the ideal temperature curve and the actual feedback can be obtained. To ensure that the subsequently generated target temperature curve reflects material properties, heat input status, and environmental conditions, the temperature difference, compensated welding heat input, ambient temperature data, and workpiece heat capacity parameters can be combined into a fuzzy rule input vector.The heat capacity parameter describes the relationship between the heat absorbed by the workpiece and the temperature change. For example, workpieces with large volume and slow heat diffusion have a lower temperature rise rate. Therefore, adding heat capacity information to the input vector allows fuzzy rules to identify the differences in thermal inertia of different workpieces during the heating and cooling stages. The constructed input vector can be input into a preset fuzzy rule library, and each input value is mapped to multiple fuzzy sets through fuzzification processing. Then, inference operations are performed according to a preset inference rule table to finally obtain a set of correction coefficients representing the adjustment range of the temperature setpoint and a set of scaling coefficients representing the adjustment range of the time node. For example, when the temperature difference is consistently high and the heat capacity parameter is low, the inference rule will output a large time scaling coefficient, extending the time of that cooling stage. Conversely, when the ambient temperature is high, leading to reduced heat dissipation, the inference result will reduce the correction coefficient of the heating stage, thereby preventing excessive temperature. After obtaining the set of correction coefficients and the set of scaling coefficients, they can be applied to the segmented temperature setpoints and time nodes respectively. By adjusting the temperature setpoints up or down by a predetermined ratio and expanding or compressing them on the time axis according to the scaling coefficients, a target temperature curve that fits the current welding state can be formed. Finally, by connecting the adjusted temperature setpoints in chronological order, a target temperature curve can be obtained that reflects the combined effects of changes in weld thermal behavior, material properties, and environmental heat dissipation conditions, providing a clear temperature tracking benchmark for subsequent iterative control.

[0031] Step 140: Determine the temperature error sequence at each sampling time based on the target temperature curve and the effective temperature feedback result, and perform moving average and deviation accumulation analysis on the temperature error sequence to obtain the error trend feature set.

[0032] Specifically, in implementing iterative learning control, the set temperature value of the target temperature curve at each sampling time and the feedback temperature value of the effective temperature feedback result at the corresponding sampling time are differentially calculated to generate a temperature error sequence. The temperature error sequence reflects the magnitude and trend of the deviation between the actual weld point temperature and the desired target temperature, and is a key quantity for determining the need to adjust welding power. For example, in laser welding, if the core area temperature is lower than the set target, the error is positive, indicating that the heating power needs to be increased; if the temperature is higher than the target, the error is negative, requiring a reduction in power, thus ensuring that the weld point temperature accurately tracks the target curve. Performing moving average and deviation accumulation analysis on the temperature error sequence can further extract error trend characteristics. Moving average, by applying a time window to average continuous error values ​​on the error sequence, can smooth transient spikes and high-frequency noise, making the trend characteristics more stable; deviation accumulation analysis is used to identify the continuous shift of error over time, reflecting power response lag or heat accumulation effects during welding. For example, when the error shows positive accumulation over several consecutive sampling periods, it indicates insufficient heating and a gradually decreasing weld point temperature, requiring increased control gain or extended heating time to avoid excessive temperature deviation accumulation. The error trend feature set formed by the moving average and deviation accumulation not only reflects the current deviation, but also contains dynamic information on temperature changes, providing a more reliable basis for power adjustment.

[0033] Step 150: Dynamically adjust the learning gain parameter and constraint coefficient in the iterative learning control law according to the error trend feature set, calculate the power correction of the welding heat source according to the adjusted control law, and update the welding output command.

[0034] Specifically, when dynamically adjusting the learning gain parameter and constraint coefficients in the iterative learning control law based on the temperature error trend feature set, statistical analysis of the time variation trend of the temperature error sequence can be performed to extract information on the direction, amplitude, and duration of error changes. The average slope and fluctuation frequency of error changes can be calculated using the sliding time window method to reflect the system's response characteristics at different stages. The adjustment of the learning gain parameter is based on the sign and magnitude of the error change rate. A proportional coefficient controls the learning rate, allowing the control law to appropriately correct rapidly fluctuating errors while maintaining a relatively gentle adjustment intensity in the steady-state range. The constraint coefficients are set according to the deviation between the current power output and the temperature feedback to limit the power adjustment amplitude and prevent sudden changes in heat source output. This parameter adjustment method enables the control law to have flexible adjustment capabilities at different operating stages, providing a reliable parameter basis for subsequent power correction calculations. When calculating the power correction amount for the welding heat source based on the adjusted control law, the updated learning gain parameter and constraint coefficients can be introduced into the control law model to perform weighted calculations on the temperature error signal of the current sampling period. The required correction magnitude for the heat source power is obtained by combining the difference between the power output of the previous cycle and the current temperature response. This calculation process can be implemented using matrix operations, ensuring that the power correction remains continuous in time and corresponds to key temperature points in the weld area in spatial distribution. By superimposing the power correction onto the output curve of the previous cycle, a new power command is formed, thereby updating the heat source output.

[0035] In one optional implementation of this embodiment, after determining the temperature error sequence at each sampling time based on the target temperature curve and the effective temperature feedback result, and performing moving average and deviation accumulation analysis on the temperature error sequence to obtain the error trend feature set, the method further includes: performing correlation calculation between the error trend feature set and the confidence coefficient to obtain a dynamic adjustment parameter set; and updating the welding output command of the previous cycle based on the dynamic adjustment parameter set to obtain the welding output command of the current cycle.

[0036] In this embodiment, a dynamic adjustment parameter set is generated by correlating the error trend feature set with the reliability coefficients of sensors in each dimension. The reliability coefficient represents the reliability of each sensor within the current sampling period, and combined with the error trend features, the true source of the deviation can be determined. For example, if an infrared thermometer produces a spike error under smoke obstruction, its reliability coefficient is low, and the weight of this error in dynamic adjustment is reduced to avoid excessive influence of outliers on the control output; conversely, high-reliability thermocouple and thermal imaging information contribute more to parameter adjustment. Through correlating the parameters, the dynamic adjustment parameter set can include the power correction factor and gain adjustment coefficient at each sampling moment, enabling the controller to perform real-time optimization while retaining historical error experience. The welding output command of the previous period is updated according to the dynamic adjustment parameter set, i.e., by calculating the power correction factor and the output command of the previous period point by point to generate the welding output command of the current period. The update process can be a weighted correction of the power curve or a direct increase or decrease in the output value, so that the laser power adapts to temperature error changes at each sampling moment. For example, when the accumulated error during the heat preservation stage is too high, the power output will be appropriately increased to accelerate the temperature rise; when the temperature in the heating stage is close to the target, the power output will be reduced to avoid overheating. The final generated welding output command not only considers the deviation between the target temperature curve and the actual temperature, but also integrates the reliability of the sensor and the dynamic characteristics of the error, to achieve precise power adjustment of the welding process, thereby ensuring that the weld point temperature tracks the target curve efficiently and stably, improving welding quality and control accuracy.

[0037] In one optional implementation of this embodiment, the step of updating the welding output command of the previous cycle according to the dynamic adjustment parameter set to obtain the welding output command of the current cycle includes: determining the power correction factor sequence corresponding to each sampling time by performing point-by-point correlation calculation between the temperature error sequence and the dynamic adjustment parameter set; generating a power correction curve by performing differential calculation between the power correction factor sequence and the power output value of the welding output command of the previous cycle at each time; and updating the welding output command of the previous cycle according to the power correction curve to obtain the welding output command of the current cycle.

[0038] In this embodiment, when combining the temperature error sequence with the dynamic adjustment parameter set, a power correction factor sequence corresponding to each sampling moment can be obtained through point-by-point correlation calculation. The power correction factor is a coefficient that adjusts the welding laser output power in real time, and its value directly reflects the power adjustment requirement of the temperature error at the current sampling moment. The principle of point-by-point correlation calculation is to multiply or weight the temperature error with the corresponding dynamic adjustment parameter at each sampling point, so that the correction factor is more significant when the error is larger or the sensor reliability is higher, while the correction factor is weaker when the error is smaller or the reliability is lower, thus forming a power correction factor sequence that corresponds one-to-one with the time sequence. For example, if the temperature is lower than the target curve and the thermocouple reliability is high during the heating stage, the power correction factor at that moment is positive and large to increase the laser output; if the temperature is close to the target curve during the cooling stage and there is noise in the infrared temperature measurement, the power correction factor may be zero or negative to prevent over-adjustment of power. By performing differential calculation based on the power correction factor sequence and the power values ​​at each moment of the welding output command in the previous cycle, a power correction curve can be generated. Differential calculation refers to performing arithmetic operations on the output power value of the previous cycle and the correction factor at the corresponding moment to obtain the adjusted power increase / decrease curve. The power correction curve reflects the power adjustment magnitude and direction at each sampling point throughout the welding process, providing the controller with a continuous adjustment strategy. For example, during the welding heat preservation stage, if the temperature error shows a cumulative downward trend, the power correction curve will remain positive during the heat preservation period, indicating an increase in output power to accelerate the temperature rise; in the initial heating stage, if the temperature is slightly higher than the target curve, the correction curve will show a negative value, suggesting an appropriate reduction in laser power to avoid overheating. Through differential calculation, the error and adjustment parameters are mapped to specific power adjustment amounts, realizing a closed-loop conversion from temperature deviation to power control. The welding output command of the previous cycle is updated based on the power correction curve to obtain the welding output command of the current cycle. The update process involves adding or multiplying the power correction curve on the time axis with the corresponding point of the power output of the previous cycle to generate a new power curve. This curve not only retains the smoothness and temporal structure of historical outputs but also integrates temperature error feedback and dynamic adjustment information, enabling the laser power to be automatically optimized for each sampling moment. For example, during continuous welding, when the core temperature gradually decreases while the ambient temperature is higher, the updated output command increases power in the core area while maintaining or slightly decreasing power in the outer area, achieving fine local control. Through this power update mechanism, the welding system can dynamically respond to temperature errors in each sampling cycle, keeping the weld point temperature stably tracking the target curve, thereby improving welding quality, reducing weld point defects, and enhancing the adaptive capability of the entire control system.

[0039] It should be noted that by dynamically adjusting the learning gain parameters and constraint coefficients in the iterative learning control law based on the error trend feature set, the welding heat source power correction is achieved with the self-learning mechanism of the control law as the core. This method is mainly suitable for scenarios where the welding process has significant dynamic changes or nonlinear thermal response characteristics. By analyzing the time variation trend of the temperature error signal, features such as transient error, steady-state error, and error change rate are extracted, and the learning gain parameters can be updated with weights. The constraint coefficients are then corrected according to the current temperature deviation range. The adjusted control law recalculates the temperature error to generate a new power correction amount, thereby updating the welding output command. This method has good control flexibility in multi-layer welding, dissimilar material welding, or conditions with significant temperature response delay. It can dynamically compensate the control parameters through continuous iteration to reflect the real-time thermal behavior of the welding process. By correlating the error trend feature set with the reliability coefficient, a dynamically adjusted parameter set is obtained. The welding output command of the previous cycle is updated based on the parameter set to obtain the welding output command of the current cycle. This method is mainly suitable for situations where the welding process is relatively stable, the operating conditions are highly repeatable, and the requirements for real-time response are high. By combining the reliability coefficient calculated from multi-sensor feedback data with the error trend feature set, a rapid response correction basis can be formed without recalculating the internal parameters of the control law. The system directly corrects the power output based on the dynamically adjusted parameter set, thereby maintaining stable control accuracy while shortening data computation and execution delays and improving output response speed. In summary, the two methods mentioned above differ significantly at the system control level. The former operates at the parameter update layer of the control law, achieving internal adjustment of the control model through dynamic adjustment of learning gain and constraint coefficients, suitable for the system in the learning and parameter adaptation stage; the latter operates at the power output command layer, achieving rapid external correction through real-time calculation of error trends and sensor reliability, suitable for the stage where the control system has entered a steady state and only requires real-time fine-tuning of the power output. The two control strategies can be selected or used alternately according to welding process requirements, temperature fluctuation characteristics, and control system response performance to ensure the continuity and stability of temperature control in the welding area.

[0040] According to the welding temperature feedback control method based on multiple sensors provided in this application, multi-dimensional temperature data collected in the welding area is preprocessed to obtain multi-dimensional temperature feature vectors. By weighted fusion of these multi-dimensional temperature feature vectors, an effective temperature feedback result for the welding area is obtained. Based on the effective temperature feedback result and ambient temperature data from the multi-dimensional temperature data, a target temperature curve is generated using preset fuzzy rules. The temperature error sequence at each sampling time is determined based on the target temperature curve and the effective temperature feedback result. Moving average and deviation accumulation analysis are performed on the temperature error sequence to obtain an error trend feature set. The learning gain parameter and constraint coefficient in the iterative learning control law are dynamically adjusted based on the error trend feature set. The power correction amount of the welding heat source is calculated based on the adjusted control law, and the welding output command is updated. Through the implementation of this application, the fusion and weighted processing of multi-dimensional temperature data can generate an effective temperature feedback result for the welding area, thereby achieving precise iterative adjustment of the welding heat source output and effectively improving the accuracy of heat source power control.

[0041] Figure 2 This application provides a multi-sensor-based welding temperature feedback control device, which can be used to implement the multi-sensor-based welding temperature feedback control method described in the foregoing embodiments. Figure 2 As shown, the welding temperature feedback control device based on multiple sensors mainly includes: The processing module 10 is used to preprocess the infrared temperature measurement signal, thermocouple signal, two-dimensional temperature field image data and ambient temperature data collected in the welding area with instantaneous high temperature and high gradient to obtain a multi-dimensional temperature feature vector. The fusion module 20 is used to obtain the effective temperature feedback result of the welding area by weighted fusion of multi-dimensional temperature feature vectors; wherein, the weight of the weighted fusion is dynamically allocated based on the credibility coefficient of each sensor in the welding dynamic process. The generation module 30 is used to generate a target temperature curve based on the effective temperature feedback results, ambient temperature data and real-time welding speed, using preset fuzzy rules. Analysis module 40 is used to determine the temperature error sequence at each sampling time based on the target temperature curve and the effective temperature feedback result, and to perform moving average and deviation accumulation analysis on the temperature error sequence to obtain the error trend feature set; The update module 50 is used to dynamically adjust the learning gain parameter and constraint coefficient in the iterative learning control law according to the error trend feature set, calculate the power correction amount of the welding heat source according to the adjusted control law, and update the welding output command.

[0042] In an optional embodiment of this example, the processing module is further configured to: determine the placement and sampling period of sensors for collecting infrared temperature measurement signals, thermocouple signals, two-dimensional temperature field image data, and ambient temperature data based on the geometric layout of the welding area and the distribution of the weld heat-affected zone; wherein the sampling period is dynamically configured according to the instantaneous thermal change rate of the welding process; control the sensors for synchronous temperature acquisition based on the placement and sampling period to obtain raw temperature data; collect the comparison outputs of the sensors for each dimension and a standard blackbody temperature source under standard welding test conditions; and cross-calibrate the raw temperature data based on the comparison outputs to obtain multi-dimensional temperature data composed of infrared temperature measurement signals, thermocouple signals, two-dimensional temperature field image data, and ambient temperature data.

[0043] In one optional implementation of this embodiment, the processing module is specifically used to: perform dynamic range adaptation and saturation zone recovery processing on the infrared temperature measurement signal according to the instantaneous high temperature characteristics of the welding process, and simultaneously perform dynamic response time compensation on the thermocouple signal to generate a synchronous denoised temperature sequence; perform region segmentation on the two-dimensional temperature field image based on molten pool contour recognition according to the high temperature gradient distribution characteristics of the welding area, and calculate the average temperature, temperature gradient field, and high temperature area change rate of the molten pool area, heat-affected zone, and base material area to generate a region feature set; perform normalization processing on the ambient temperature data, and apply dynamic baseline adjustment to the normalization parameters to obtain the adjusted normalized environmental parameters; and perform spatiotemporal alignment and correlation integration of the denoised temperature sequence, region feature set, and normalized environmental parameters according to the welding time sequence to generate a multidimensional temperature feature vector.

[0044] In an optional implementation of this embodiment, the fusion module is specifically used to: determine the variance and drift rate of the infrared temperature measurement signal and the thermocouple signal in adjacent sampling periods based on the time window data of the denoised temperature sequence, and obtain a noise evaluation index set; dynamically evaluate the reliability coefficient of the infrared temperature measurement signal and the thermocouple signal under the current welding process stage by combining the noise evaluation index set with the average temperature, maximum temperature gradient and normalized environmental parameters of the regional feature set, wherein the welding process stage includes at least the arc initiation stage, the stable welding stage and the arc termination stage; obtain a weight template library corresponding to the welding process stage, select a benchmark weight distribution from the weight template library according to the currently identified process stage, and adjust the benchmark weight in real time according to the dynamic reliability coefficient sequence to obtain a weight matrix; and perform weighted fusion of the multi-dimensional temperature feature vector according to the weight matrix to obtain the effective temperature feedback result of the welding area.

[0045] In one optional implementation of this embodiment, the generation module is specifically used for: determining the current welding heat input based on the real-time welding speed and effective temperature feedback results, and performing heat dissipation compensation on the welding heat input based on the ambient temperature data to obtain the compensated welding heat input; retrieving the initial process curve corresponding to the material type and geometry type from the preset process database based on the material type, geometry type of the welding workpiece, and the compensated welding heat input; extracting the temperature set points and time nodes of each stage of the initial process curve in segments, and determining the temperature difference between the temperature set point and the effective temperature feedback result in the same time segment; combining the temperature difference, the compensated welding heat input, the ambient temperature data, and the workpiece heat capacity parameters into a fuzzy rule input vector; inputting the fuzzy rule input vector into the preset fuzzy rule library, sequentially performing fuzzification processing and calculations based on the inference rule table, and outputting the correction coefficient set of the temperature set point and the scaling coefficient set of the time node; parametrically adjusting the segmented temperature set points and time nodes according to the correction coefficient set and the scaling coefficient set, and reorganizing the adjusted segmented temperature set points into the target temperature curve in chronological order.

[0046] In one optional implementation of this embodiment, the update module is specifically used to: perform correlation calculation between the error trend feature set and the confidence coefficient to obtain a dynamic adjustment parameter set; and update the welding output command of the previous cycle according to the dynamic adjustment parameter set to obtain the welding output command of the current cycle.

[0047] In an optional implementation of this embodiment, the update module is further configured to: determine the power correction factor sequence corresponding to each sampling moment by performing point-by-point correlation calculation between the temperature error sequence and the dynamic adjustment parameter set; generate a power correction curve by performing differential calculation between the power correction factor sequence and the power output value of the welding output command at each moment of the previous cycle; and update the welding output command of the previous cycle according to the power correction curve to obtain the welding output command of the current cycle.

[0048] According to the welding temperature feedback control device based on multiple sensors provided in this application, multi-dimensional temperature data collected in the welding area is preprocessed to obtain multi-dimensional temperature feature vectors. By weighted fusion of the multi-dimensional temperature feature vectors, an effective temperature feedback result for the welding area is obtained. Based on the effective temperature feedback result and ambient temperature data from the multi-dimensional temperature data, a target temperature curve is generated using preset fuzzy rules. The temperature error sequence at each sampling time is determined based on the target temperature curve and the effective temperature feedback result. Moving average and deviation accumulation analysis are performed on the temperature error sequence to obtain an error trend feature set. The learning gain parameter and constraint coefficient in the iterative learning control law are dynamically adjusted based on the error trend feature set. The power correction amount of the welding heat source is calculated based on the adjusted control law, and the welding output command is updated. Through the implementation of this application, the fusion and weighted processing of multi-dimensional temperature data can generate an effective temperature feedback result for the welding area, thereby achieving precise iterative adjustment of the welding heat source output and effectively improving the accuracy of heat source power control.

[0049] According to the scheme provided in this application Figure 3 An electronic device is provided as an embodiment of this application. This electronic device can be used to implement the multi-sensor-based welding temperature feedback control method described in the foregoing embodiments, and mainly includes: The system includes a memory 301, a processor 302, and a computer program 303 stored in the memory 301 and executable on the processor 302. The memory 301 and the processor 302 are connected via communication. When the processor 302 executes the computer program 303, it implements the multi-sensor-based welding temperature feedback control method described in the foregoing embodiments. The number of processors can be one or more.

[0050] The memory 301 can be a high-speed random access memory (RAM) or a non-volatile memory, such as a disk storage device. The memory 301 is used to store executable program code, and the processor 302 is coupled to the memory 301.

[0051] Furthermore, embodiments of this application also provide a computer-readable storage medium, which may be disposed in the electronic device described in the above embodiments, and the computer-readable storage medium may be as described above. Figure 3 The memory in the illustrated embodiment.

[0052] The computer-readable storage medium stores a computer program that, when executed by a processor, implements the multi-sensor-based welding temperature feedback control method described in the foregoing embodiments. Furthermore, the computer-readable storage medium can also be a USB flash drive, external hard drive, read-only memory (ROM), RAM, magnetic disk, or optical disk, or any other medium capable of storing program code.

[0053] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0054] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0055] The above-described embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application.

Claims

1. A welding temperature feedback control method based on multiple sensors, characterized in that, include: Infrared temperature measurement signals, thermocouple signals, two-dimensional temperature field image data, and ambient temperature data collected in welding areas with instantaneous high temperature and high gradient are preprocessed to obtain multidimensional temperature feature vectors. The effective temperature feedback result of the welding area is obtained by weighted fusion of the multidimensional temperature feature vector; wherein the weight of the weighted fusion is dynamically allocated based on the reliability coefficient of each sensor in the welding dynamic process. Based on the effective temperature feedback result, the ambient temperature data, and the real-time welding speed, a target temperature curve is generated using preset fuzzy rules. Based on the target temperature curve and the effective temperature feedback result, the temperature error sequence at each sampling time is determined, and the temperature error sequence is subjected to moving average and deviation accumulation analysis to obtain the error trend feature set. The learning gain parameter and constraint coefficient in the iterative learning control law are dynamically adjusted according to the error trend feature set. The power correction of the welding heat source is calculated according to the adjusted control law, and the welding output command is updated.

2. The welding temperature feedback control method based on multiple sensors according to claim 1, characterized in that, Before the step of preprocessing the infrared temperature measurement signals, thermocouple signals, two-dimensional temperature field image data, and ambient temperature data collected in the welding area with instantaneous high temperature and high gradient to obtain a multi-dimensional temperature feature vector, the method further includes: Based on the geometric layout of the welding area and the distribution of the heat-affected zone of the weld, the placement and sampling period of the sensors for collecting infrared temperature measurement signals, thermocouple signals, two-dimensional temperature field image data, and ambient temperature data are determined; wherein, the sampling period is dynamically configured according to the instantaneous thermal change rate of the welding process. Based on the deployment location and sampling period, the sensors in each dimension are controlled to synchronously collect temperature data to obtain the raw temperature data. The outputs of sensors in various dimensions were compared with those of a standard blackbody temperature source under standard welding test conditions. The original temperature data is cross-calibrated based on the comparison output to obtain multi-dimensional temperature data composed of the infrared temperature measurement signal, the thermocouple signal, the two-dimensional temperature field image data, and the ambient temperature data.

3. The welding temperature feedback control method based on multiple sensors according to claim 1, characterized in that, The step of preprocessing the infrared temperature measurement signals, thermocouple signals, two-dimensional temperature field image data, and ambient temperature data collected in the welding area with instantaneous high temperature and high gradient to obtain a multidimensional temperature feature vector includes: Based on the instantaneous high temperature characteristics of the welding process, the infrared temperature measurement signal is dynamically range-adapted and saturation region-recovered, while the thermocouple signal is dynamically response time-compensated to generate a synchronized denoised temperature sequence. Based on the high temperature gradient distribution characteristics of the welding area, the two-dimensional temperature field image is segmented based on molten pool contour recognition, and the average temperature, temperature gradient field and high temperature area change rate of the molten pool area, heat-affected zone and base material area are calculated to generate a region feature set. The ambient temperature data is normalized, and dynamic baseline adjustment is applied to the normalization parameters to obtain the adjusted normalized environmental parameters. The denoised temperature sequence, the regional feature set, and the normalized environmental parameters are spatiotemporally aligned and correlated according to the welding time sequence to generate a multidimensional temperature feature vector.

4. The welding temperature feedback control method based on multiple sensors according to claim 3, characterized in that, The step of obtaining the effective temperature feedback result of the welding area by weighted fusion of the multidimensional temperature feature vector includes: Based on the time window data of the denoised temperature sequence, the variance and drift rate of the infrared temperature measurement signal and the thermocouple signal in adjacent sampling periods are determined to obtain a set of noise evaluation indicators. The reliability coefficients of the infrared temperature measurement signal and the thermocouple signal under the current welding process stage are dynamically evaluated by combining the noise evaluation index set with the average temperature, maximum temperature gradient, and normalized environmental parameters of the regional feature set. The welding process stage includes at least the arc initiation stage, the stable welding stage, and the arc termination stage. Obtain a weight template library corresponding to the welding process stage, select a benchmark weight distribution from the weight template library according to the currently identified process stage, and adjust the benchmark weight in real time according to the dynamic confidence coefficient sequence to obtain a weight matrix; The multidimensional temperature feature vector is weighted and fused according to the weight matrix to obtain the effective temperature feedback result of the welding area.

5. The welding temperature feedback control method based on multiple sensors according to claim 4, characterized in that, The step of generating a target temperature curve based on the effective temperature feedback result, the ambient temperature data, and the real-time welding speed using preset fuzzy rules includes: The current welding heat input is determined based on the real-time welding speed and the effective temperature feedback result, and the welding heat input is compensated for by heat dissipation based on the ambient temperature data to obtain the compensated welding heat input. Based on the material type, geometry type of the workpiece to be welded, and the compensated welding heat input, the initial process curve corresponding to the material type and geometry type is retrieved from the preset process database; The initial process curve is segmented to extract the temperature setpoints and time nodes of each stage, and the temperature difference between the temperature setpoint and the effective temperature feedback result in the same time period is determined. The temperature difference, the compensated welding heat input, the ambient temperature data, and the workpiece heat capacity parameters are combined into a fuzzy rule input vector. The fuzzy rule input vector is input into a preset fuzzy rule library, and fuzzification processing and operations based on the inference rule table are performed in sequence to output the set of correction coefficients for the temperature setpoint and the set of scaling coefficients for the time node. The segmented temperature setpoints and time nodes are parametrically adjusted according to the set of correction coefficients and the set of scaling coefficients, and the adjusted segmented temperature setpoints are reorganized into the target temperature curve in chronological order.

6. The welding temperature feedback control method based on multiple sensors according to claim 5, characterized in that, After the steps of determining the temperature error sequence at each sampling time based on the target temperature curve and the effective temperature feedback result, performing moving average and deviation accumulation analysis on the temperature error sequence, and obtaining the error trend feature set, the method further includes: The error trend feature set is correlated with the confidence coefficient to obtain a dynamic adjustment parameter set. The welding output command of the previous cycle is updated according to the dynamically adjusted parameter set to obtain the welding output command of the current cycle.

7. The welding temperature feedback control method based on multiple sensors according to claim 6, characterized in that, The step of updating the welding output command of the previous cycle according to the dynamically adjusted parameter set to obtain the welding output command of the current cycle includes: By performing point-by-point correlation calculations between the temperature error sequence and the dynamic adjustment parameter set, the power correction factor sequence corresponding to each sampling time is determined. The power correction curve is generated by performing differential calculations based on the power correction factor sequence and the power output value at each moment of the welding output command in the previous cycle. The welding output command of the previous cycle is updated according to the power correction curve to obtain the welding output command of the current cycle.

8. A welding temperature feedback control device based on multiple sensors, characterized in that, The multi-sensor-based welding temperature feedback control device is used to implement the multi-sensor-based welding temperature feedback control method of claim 1, wherein the multi-sensor-based welding temperature feedback control device comprises: The processing module is used to preprocess infrared temperature measurement signals, thermocouple signals, two-dimensional temperature field image data and ambient temperature data collected in the welding area with instantaneous high temperature and high gradient to obtain multi-dimensional temperature feature vectors. The fusion module is used to obtain the effective temperature feedback result of the welding area by weighted fusion of the multi-dimensional temperature feature vector; wherein, the weight of the weighted fusion is dynamically allocated based on the credibility coefficient of each sensor in the welding dynamic process; The generation module is used to generate a target temperature curve based on the effective temperature feedback result, the ambient temperature data, and the real-time welding speed, using preset fuzzy rules. The analysis module is used to determine the temperature error sequence at each sampling time based on the target temperature curve and the effective temperature feedback result, and to perform moving average and deviation accumulation analysis on the temperature error sequence to obtain the error trend feature set. The update module is used to dynamically adjust the learning gain parameter and constraint coefficient in the iterative learning control law according to the error trend feature set, calculate the power correction amount of the welding heat source according to the adjusted control law, and update the welding output command.

9. An electronic device, characterized in that, Includes memory and processor, of which: The processor is used to execute computer programs stored in the memory; When the processor executes the computer program, it implements the steps in the multi-sensor-based welding temperature feedback control method according to any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the steps in the multi-sensor-based welding temperature feedback control method according to any one of claims 1 to 7.

Images