CN117182370A

CN117182370A - Intelligent welding optimization and error source analysis method

Info

Publication number: CN117182370A
Application number: CN202311128253.2A
Authority: CN
Inventors: 黎鸿宾; 李艺晗; 李光金; 许立昂; 邱光堂; 邓城承; 王璐烽
Original assignee: Chongqing Xinwu Technology Co ltd
Current assignee: Chongqing Industry Polytechnic College
Priority date: 2023-09-01
Filing date: 2023-09-01
Publication date: 2023-12-08

Abstract

The invention relates to the technical field of intelligent manufacturing, in particular to an intelligent welding optimization and error source analysis method, which comprises the following steps: collecting welding parameters in the welding process, and preprocessing; calculating to obtain an overall performance change value, inputting welding parameters and the overall performance change value into a pre-trained deep learning model, obtaining the sensitivity of each geometric error source, and identifying key geometric error sources; the welding parameters are dynamically adjusted in real time based on the adjustment strategy, the adjusted performance indexes are evaluated, and the deep learning model and the adjustment strategy are optimized based on the evaluation result, so that the optimal performance can be ensured under various production environments and conditions; adapting to the continuously changing production requirements; the production cost can be effectively reduced, the production efficiency can be improved, the quality of the final product can be improved, and the method has obvious economic benefit.

Description

Intelligent welding optimization and error source analysis method

Technical Field

The invention relates to the technical field of intelligent manufacturing, in particular to an intelligent welding optimization and error source analysis method.

Background

Welding is a critical assembly process in the manufacturing industry and has a direct impact on product quality, production efficiency, and overall manufacturing costs. With the continuous development of advanced manufacturing technology, requirements on precision and efficiency of the welding process are also increasing.

The welding process involves a number of parameters (e.g., weld geometry, weld speed, pressure, etc.) that interact with each other, resulting in optimizing one parameter that may affect the performance of the other; traditional welding technology is usually carried out based on preset parameters, and is difficult to dynamically adjust according to real-time conditions (such as material properties, equipment states and the like); typically, the assessment of the welding process is more focused on a single or a few performance indicators, such as weld quality, while other factors, such as production speed and equipment utilization, are ignored.

Existing welding techniques and methods are often limited to fixed parameter settings, lack of adaptive capabilities, which results in the need for manual intervention when faced with different production environments and requirements, increasing production costs and time; the existing method usually only pays attention to a few performance indexes, such as weld quality, but pays less attention to other possibly equal or more important indexes, such as production efficiency and equipment utilization rate; current welding systems often lack real-time data analysis and feedback mechanisms, making timely adjustments difficult even when deviations or problems occur.

Disclosure of Invention

Aiming at the problems existing in the prior art, the invention provides an intelligent welding optimization and error source analysis method, firstly, the invention can accurately identify and quantify various key geometric error sources by applying a deep learning model, and improve welding quality and accuracy; secondly, the invention supports real-time adjustment of welding parameters and equipment scheduling so as to cope with uncertainty and change in the production environment, thereby improving the equipment utilization rate and the production efficiency; finally, the invention has self-optimizing capability, can continuously adjust the deep learning model and the scheduling strategy according to the real-time evaluation result, and ensures continuous optimization and adaptability; the invention has high reliability and efficiency in industrial welding applications.

The intelligent welding optimization and error source analysis method comprises the following steps:

collecting welding parameters in the welding process, and preprocessing;

calculating to obtain an overall performance change value, inputting welding parameters and the overall performance change value into a pre-trained deep learning model, obtaining the sensitivity of each geometric error source, and identifying key geometric error sources;

and dynamically adjusting welding parameters in real time based on an adjustment strategy, evaluating the adjusted performance indexes, and optimizing the deep learning model and the adjustment strategy based on an evaluation result.

Preferably, the welding parameters include: weld geometry, weld speed, weld pressure, material properties, and equipment status, wherein,

the weld geometry parameters include: width, depth, and angle of the weld;

the material properties include: material type, thickness, and temperature;

the device state includes: current, voltage and operating temperature of the device.

Preferably, the pretreatment includes: and (5) data cleaning and normalization processing.

Preferably, the calculation expression of the overall performance change value is:

wherein ΔE is the overall performance change value, w _i Is the weight of the ith geometric error source, Δe _i Is the variation value of the ith geometric error source.

Preferably, the deep learning model structure is a convolutional neural network, and the deep learning model is pre-trained by taking the welding parameters in the history welding record as input and the overall performance change value as output, so as to predict and analyze the influence of each geometric error source on the overall performance.

Preferably, the welding parameters and the overall performance variation value are input into a pre-trained deep learning model, and the calculation expression for obtaining the sensitivity of each geometric error source is as follows:

wherein S is _i Is the sensitivity value of the ith geometric error source.

Preferably, a geometric error source with sensitivity greater than a first threshold and an overall performance change value greater than a second threshold is used as the key geometric error source.

Preferably, the adjustment strategy comprises:

when the sensitivity of a certain key geometric error source exceeds a third threshold, the welding parameters are adjusted in real time, and the third threshold is determined based on the sensitivity and the overall performance change value; if a certain key geometric error source affects the welding speed, adjusting the numerical value according to the sensitivity by reducing or increasing the welding speed; if a certain key geometric error source influences the welding pressure, the welding pressure is adjusted according to the numerical value of the sensitivity; if a certain key geometric error source affects a plurality of parameters, adjusting each parameter one by one according to the weight of the sensitivity;

When the sensitivity of at least two key geometric error sources exceeds a third threshold, calculating a comprehensive adjustment coefficient according to the sensitivity of each key geometric error source; according to the comprehensive adjustment coefficient, adjusting welding parameters;

and when the total sensitivity of the specific type of key geometric error source exceeds a fourth threshold, comprehensively evaluating all equipment parameters influenced by the key geometric error source, and adjusting welding parameters according to an evaluation result.

Preferably, the performance index includes: and (3) comparing the performance index with the overall performance change value to finish the evaluation of the performance index, wherein the welding precision, the welding speed, the equipment utilization rate and the weld quality are the same.

Preferably, the optimizing the deep learning model and the adjustment strategy based on the evaluation result includes:

when the estimation result is that the welding precision is poor, fine tuning parameters related to a key geometric error source in the deep learning model; in the regulation strategy, the welding speed is reduced, and the welding pressure is increased;

when the evaluation result is that the welding speed is low, retraining the deep learning model, and adjusting the key geometric error sources related to the speed; in the regulation strategy, the welding speed is increased, and the welding pressure is reduced;

When the evaluation result is that the equipment utilization rate is low, analyzing related key geometric error sources and other additional factors influencing the equipment utilization rate, and adjusting the deep learning model based on the information; and in the regulation strategy, regulating the equipment scheduling strategy.

Compared with the prior art, the invention has the advantages that:

(1) The method and the device identify and adjust key geometric error sources in real time through the deep learning model, and can provide highly personalized welding parameter settings under various production environments and conditions, thereby ensuring optimal performance;

(2) By integrating a plurality of performance indexes and overall performance change values, the invention not only can comprehensively evaluate welding quality, speed and equipment utilization rate, but also can dynamically adjust a model and strategy according to real-time feedback so as to adapt to the continuously-changing production requirements;

(3) Because the key geometric error sources are accurately identified and optimized, the invention can effectively reduce the production cost and improve the production efficiency, and simultaneously can improve the quality of the final product, thereby having remarkable economic benefit.

Drawings

FIG. 1 is a schematic flow chart of the method of the present invention;

FIG. 2 is a schematic diagram of the content of the adjustment strategy according to the present invention;

FIG. 3 is a schematic block diagram of a neutral indicator structure according to the present invention;

FIG. 4 is a schematic block diagram of the evaluation result structure in the present invention.

Detailed Description

Hereinafter, embodiments of the present disclosure will be described with reference to the accompanying drawings. It should be understood that the description is only exemplary and is not intended to limit the scope of the present disclosure. In the following detailed description, for purposes of explanation, numerous specific details are set forth in order to provide a thorough understanding of the embodiments of the present disclosure. It may be evident, however, that one or more embodiments may be practiced without these specific details. In addition, in the following description, descriptions of well-known structures and techniques are omitted so as not to unnecessarily obscure the concepts of the present disclosure.

The terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of the disclosure. The terms "comprises," "comprising," and/or the like, as used herein, specify the presence of stated features, steps, operations, and/or components, but do not preclude the presence or addition of one or more other features, steps, operations, or components.

All terms (including technical and scientific terms) used herein have the same meaning as commonly understood by one of ordinary skill in the art unless otherwise defined. It should be noted that the terms used herein should be construed to have meanings consistent with the context of the present specification and should not be construed in an idealized or overly formal manner.

Where expressions like at least one of "A, B and C, etc. are used, the expressions should generally be interpreted in accordance with the meaning as commonly understood by those skilled in the art (e.g.," a system having at least one of A, B and C "shall include, but not be limited to, a system having a alone, B alone, C alone, a and B together, a and C together, B and C together, and/or A, B, C together, etc.). Where a formulation similar to at least one of "A, B or C, etc." is used, in general such a formulation should be interpreted in accordance with the ordinary understanding of one skilled in the art (e.g. "a system with at least one of A, B or C" would include but not be limited to systems with a alone, B alone, C alone, a and B together, a and C together, B and C together, and/or A, B, C together, etc.).

Some of the block diagrams and/or flowchart illustrations are shown in the figures. It will be understood that some blocks of the block diagrams and/or flowchart illustrations, or combinations of blocks in the block diagrams and/or flowchart illustrations, can be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general purpose computer, special purpose computer, or other programmable data processing apparatus, such that the instructions, when executed by the processor, create means for implementing the functions/acts specified in the block diagrams and/or flowchart. The techniques of this disclosure may be implemented in hardware and/or software (including firmware, microcode, etc.). Additionally, the techniques of this disclosure may take the form of a computer program product on a computer-readable storage medium having instructions stored thereon, the computer program product being for use by or in connection with an instruction execution system.

As shown in fig. 1, the intelligent welding optimization and error source analysis method comprises the following steps:

collecting welding parameters in the welding process, and preprocessing;

according to the invention, the welding parameters are dynamically adjusted in real time based on the adjustment strategy, the performance index after adjustment is evaluated, and the deep learning model and the adjustment strategy are optimized based on the evaluation result. Due to real-time monitoring and adjustment, the scheme of the invention is expected to obviously improve welding precision. Through deep learning model and sensitivity analysis, real-time adjustment can make the welding process smoother, thereby improving efficiency. The scheme of the invention can carry out self-adjustment and optimization according to the real-time data and the evaluation result, and has high self-adaptability. The invention solves the problem of precision control in the complex welding process through a data driving and intelligent adjusting strategy, and has high theoretical and practical values.

The weld geometry parameters include: width, depth, and angle of the weld; these parameters (width, depth, angle of weld) directly affect the mechanical properties and appearance quality of the welded joint. Improper geometry results in weld defects such as slag inclusions, holes, etc.; if a tank is to be welded, the width and depth of the weld must be large enough to withstand the internal pressure, while the angle needs to be set properly to ensure that there are no stress concentrations.

Welding speed, welding pressure, these parameters relate to heat input and metal flow, affecting the geometry and microstructure of the weld; different weld speeds and pressures can affect the size and shape of the heat affected zone.

The material properties include: material type, thickness, and temperature; the type, thickness and temperature of the material have great influence on the stability of the welding process; different materials have different melting points, thermal conductivities and coefficients of thermal expansion, all of which need to be considered during the welding process; if stainless steel and carbon steel are welded, it is necessary to adjust the current and voltage in consideration of their different coefficients of thermal expansion and melting points.

The device state includes: current, voltage and operating temperature of the device; including current, voltage and operating temperature of the device; these parameters are related to the stability of the power supply and the operating conditions of the gun, thereby affecting the controllability and accuracy of the overall welding process, as the operating temperature of the equipment increases during long continuous welding, when appropriate current and voltage reductions, or shutdown cooling, are required.

These parameters are all interrelated and influencing. For example, increasing the current requires increasing the voltage, but this increases the heat input, which in turn affects the microstructure and mechanical properties of the material.

Because the parameters interact, a comprehensive and comprehensive method is needed to optimize the parameters to achieve the desired welding effect. This is why deep learning models are required to understand these complex correlations.

The welding quality can be obviously improved and defects and unqualified products can be reduced by carefully selecting and optimizing the welding parameters; the proper parameter setting can reduce the downtime and repair time in the welding process, thereby improving the production efficiency; optimizing parameters can reduce material waste and energy consumption, thereby reducing production cost.

The data cleaning comprises the following steps:

data integrity: any missing or incomplete data needs to be identified and processed; data loss due to sensor failure, data transmission errors, or other non-standard operations;

outlier processing: during welding, outliers or outliers are encountered; these data are generated due to equipment failure or operator error and need to be identified and rejected;

Consistency: ensuring that all data follow the same unit and scale; for example, if one sensor reports current units of amperes and the other is milliamperes, they need to be consolidated to one standard unit.

In one embodiment, it is assumed that during the welding process, the current reading suddenly jumps to an abnormally high value, such as from 20A to 200A. Such outliers may be identified and deleted by the data cleansing process.

The normalization process comprises the following steps:

range scaling: the data range of all features is scaled to the same interval (typically [0,1] or [ -1,1 ]). This helps the machine learning model converge faster.

Mean and variance adjustment: another common normalization technique is Z-Score normalization, i.e., subtracting the mean and dividing by the standard deviation, to yield data with a mean of 0 and a variance of 1.

And (5) weight adjustment: in some cases, certain features are more important than others. Normalization may also be used to adjust according to these weights.

In one embodiment, welding is performed in multiple production lines, with different current and voltage ranges for each line having different equipment conditions. Through normalization processing, all data can be ensured to be on the same scale, so that the deep learning model can accurately predict.

Assuming that the material type has a greater impact on the weld quality, relative to other parameters such as current and voltage. In this case, the material type may be given a higher weight during the normalization process.

In the invention, the preprocessed data is easier to be processed by the deep learning model or other machine learning models, thereby improving the accuracy and stability of the model; because the data range and distribution have been standardized, the model is easier to quickly converge during the training process; data cleansing can eliminate errors and anomalies, thereby improving reliability of model predictions.

Weight w of ith geometric error source _i This is a coefficient representing the extent to which the ith geometric error source affects the overall performance change. The weights may be empirically based, based on a physical model, or derived by an optimization algorithm. The weights range from 0 to 1 and the sum of all weights is typically 1 to represent a normalized ratio.

Variation value deltae of ith geometric error source _i Representing the amount of variation of the ith geometric error source (e.g., weld width, depth, angle, etc.).These variations may be actual measurements or differences with respect to some reference value, such as an ideal weld geometry.

The minimum value of ΔE represents optimal welding performance because all geometric error sources are minimized or weighted to optimal conditions.

The invention provides a flexible way to quantify the comprehensive influence of multiple geometric error sources on the whole performance because the weight and the error value are variables; weight w _i The model can be optimized according to different applications or different welding tasks, so that the model can adapt to different scenes; the expression provides a quantitative way to evaluate the performance of the welding process, which is very useful for further analysis and optimization.

In one embodiment, three geometric error sources are assumed: width of weld (Δe) ₁ ) Depth (Δe) ₂ ) And angle (Δe) ₃ ) Their weights are 0.4, 0.3 and 0.3, respectively.

In an ideal state: Δe ₁ ＝Δe ₂ ＝Δe ₃ =0, where Δe=0, indicates that there is no geometric error and that the performance is optimal.

In actual operation: suppose Δe is measured ₁ ＝0.1，Δe ₂ ＝0.2，Δe ₃ =0.1, the overall performance change value Δe=0.4×0.1+0.3×0.2+0.3×0.1=0.16.

The deep learning model selected in the present invention is a convolutional neural network (Convolutional Neural Network, CNN). CNN is mainly composed of a convolutional layer, an activation function, a pooling layer, and a fully-connected layer. CNN is typically used to process data (e.g., images) having a grid structure, but here it is applied to process welding parameters and overall performance variation values.

Input: welding parameters (e.g., geometry of the weld, welding speed, welding pressure, material properties, and equipment status) in the historical welding record.

And (3) outputting: overall performance change value (Δe).

The model is trained beforehand using a large number of historical welding records. This training process involves minimization of a loss function, which is typically the difference between the actual output (i.e., true ΔE) and the model predicted output.

Once the model is pre-trained, it can be used to predict and analyze the impact of each geometrical error source (e.g., weld width, depth, angle, etc.) on the overall performance (Δe) in real time.

Convolutional neural networks have a high degree of accuracy in complex pattern recognition, which helps to accurately predict the impact of each geometric error source on overall performance; because the model is pre-trained based on data, it automatically learns how to trade-off different geometric error sources; once trained, the model can be used in a real-time environment to monitor and optimize the welding process.

In one embodiment, assuming a pre-trained CNN model, inputs include width, depth and angle of weld, welding speed, welding pressure, material type, thickness, temperature, current and voltage.

During training, models have learned that, for example, the depth and width of the weld is the most important source of geometric error affecting overall performance (ΔE).

During actual welding, if the input parameters indicate a small weld depth, the model predicts that ΔE will increase (performance degradation).

Based on this prediction, the welding parameters (e.g., increasing welding pressure or slowing welding speed) may be adjusted in real time to improve weld depth, thereby optimizing overall performance.

Therefore, the pre-trained CNN model not only can predict the influence of each geometric error source on the whole performance, but also can provide basis for real-time optimization decision.

wherein S is _i Is the sensitivity value of the ith geometric error source.

Sensitivity (S) in the present invention _i ) Is a key index for quantifying the extent of influence of the ith geometric error source (e.g., weld width, depth, etc.) on the overall performance change value (Δe), and is intended to explain the dependency between the model output and input.

By calculating the sensitivity of each geometrical error source, the invention is able to quantify the impact of these error sources on overall performance; high sensitivity value (S) _i ) Meaning that the corresponding geometric error source has a greater influence and can be preferentially adjusted; based on the sensitivity information, the welding process can be optimized and adjusted in real time.

In one embodiment, three geometric error sources are assumed: weld Width (Width), weld Depth (Depth), and weld Angle (Angle).

Calculating sensitivity: using a pre-trained deep learning model, we find:

sensitivity of width (S ₁ )＝0.5

Sensitivity of depth (S ₂ )＝0.3

Sensitivity of angle (S ₃ )＝0.2

According to the sensitivity analysis, the width has the greatest effect on overall performance. Therefore, the optimization strategy should first focus on controlling the weld width.

If the real-time monitoring shows that the width of the welding line deviates, the welding line is immediately adjusted to optimize the overall performance.

In this way, the sensitivity analysis not only provides a degree of influence of each geometrical error source on the overall performance, but also provides a basis for real-time optimization and adjustment.

The first threshold and the second threshold are set as a basis for evaluating the importance of the geometric error source. The first threshold is for sensitivity (S _i ) And the second threshold is for an overall performance change value (deltae). The threshold is typically determined by prior empirical, analog or statistical analysis, or may be specifically set according to actual requirements.

By considering both sensitivity and overall performance variation values, the invention can more fully evaluate the impact of each geometric error source, ensuring that only truly important factors are selected as the key geometric error source.

The invention can more effectively identify the geometric error sources with the greatest influence on welding quality and performance by using two thresholds; by focusing on the key geometric error sources, the complexity and risk of over-optimizing non-key parameters can be reduced; according to the key geometric error sources, real-time dynamic adjustment can be carried out more specifically, so that the overall performance is optimized.

In one embodiment, it is assumed that in one weld project there are several geometric error sources of sensitivity and overall performance variation values:

width of weld (S) ₁ ＝0.6,ΔE＝0.7)

Depth of weld (S) ₂ ＝0.2,ΔE＝0.1)

Weld angle (S) ₃ ＝0.4,ΔE＝0.2)

Assuming that the first threshold is 0.5, the second threshold is 0.6.

In this case, only the "bead width" satisfies the condition that the sensitivity is greater than 0.5 (first threshold value) and the overall performance variation value is greater than 0.6 (second threshold value). Thus, weld width is considered a critical source of geometric errors.

Engineers or operators may focus on optimizing and adjusting the weld width to achieve the greatest improvement in overall performance. This also reduces excessive concerns and resource waste for less important sources of geometric errors.

Preferably, as shown in fig. 2, the adjustment strategy includes:

The third threshold is determined based on the sensitivity and the overall performance change value. This means that when the impact of one error source exceeds this value, its impact on overall performance will be considered significant, requiring immediate real-time adjustment.

Depending on the sensitivity of the geometrical error source, the adjustment strategy for the welding parameters can be decided.

If the sensitivity is high, the amplitude of the adjustment will be larger. Conversely, if the sensitivity is lower, the amplitude of the adjustment will be smaller. The direction of adjustment (e.g., increasing or decreasing the welding speed) is dependent on the positive or negative value of the sensitivity or other performance related factors.

If a particular source of geometric errors is known to primarily affect a certain parameter (e.g., welding speed or welding pressure), adjustments may be made specifically, rather than blindly adjusting all parameters.

According to the invention, the welding process can be ensured to be always carried out under the optimal condition by setting the third threshold value and carrying out real-time monitoring and adjustment; the key geometric error sources are adjusted in a targeted manner, so that unnecessary resource waste can be avoided; with real-time optimization of welding parameters, improvements in weld quality and stability may be expected.

In one embodiment, assuming a critical geometrical error source, the depth of the weld, the sensitivity is 0.9, well above the third threshold of 0.7. Since this error source is known to primarily affect the welding speed and welding pressure, adjustments need to be made:

If the depth of the weld increases to cause the welding speed to be too high, the welding speed needs to be reduced. The magnitude of the specific decrease will be determined based on a sensitivity value of 0.9; assuming that an increase in the depth of the weld also results in a too high welding pressure, it is also necessary to reduce the welding pressure. The size of the adjustment will also be determined based on the sensitivity value.

If the depth of the weld affects both the welding speed, the welding pressure and other parameters, it is also necessary to consider how the sensitivity of 0.9 is assigned to each parameter, ensuring that the adjustment of each parameter achieves the best results.

when the sensitivity of at least two critical geometrical error sources exceeds a third threshold, a single adjustment is insufficient to solve the problem. At this time, a more complex, diversified solution is required. The integrated adjustment factor is an integrated value representing the sensitivity of the plurality of error sources, and is a weighted sum or other composite value thereof, used to integrate the impact of the plurality of critical geometric error sources on the welding parameters.

According to the calculated comprehensive adjustment coefficient, the real-time dynamic adjustment of the welding parameters can be performed more accurately. Because this factor integrates the effects of multiple error sources, it provides a multi-dimensional view to make adjustments to the welding parameters, making the adjustments more comprehensive and accurate.

According to the invention, by considering the influence of a plurality of key geometric error sources, a more accurate adjustment basis is provided by the comprehensive adjustment coefficient, so that the welding process can be more accurately optimized; under the complex condition that a plurality of key geometric error sources need to be adjusted, the overall performance can be ensured to be more stable by using the comprehensive adjustment coefficient; through accurate adjustment, limited resources can be effectively utilized, and waste is avoided.

In one embodiment, assuming two sources of critical geometric errors-weld depth and weld width, their sensitivities are 0.9 and 0.8, respectively, both exceeding a third threshold of 0.7.

Calculating a comprehensive adjustment coefficient: this coefficient can be calculated in a number of ways, such as simply taking the average of the two sensitivities, or using a more complex weighted average. Assuming that a weighted average is used, weights of 0.6 and 0.4, respectively, then the overall adjustment coefficient=0.9×0.6+0.8×0.4=0.86

Since the overall adjustment factor is 0.86, this is a fairly high value, indicating that the welding parameters need to be adjusted significantly. For example, if the weld depth affects mainly the welding speed and the weld width affects mainly the welding pressure, both parameters need to be adjusted by a correspondingly large amount according to the overall adjustment coefficient.

Through such multi-dimensional adjustment, not only can single parameters be more precisely controlled, but also the welding process can be more comprehensively optimized.

"Total sensitivity of a particular type of critical geometric error source" generally means that such error sources (e.g., weld geometry, material properties, etc.) have an accumulated or integrated effect, as derived by weighted averaging or other mathematical means. The total sensitivity is actually a value reflecting the overall impact of these error sources.

After the total sensitivity exceeds the fourth threshold, a comprehensive assessment of all equipment parameters affected by these error sources (e.g., welding speed, welding pressure, current, voltage, etc.) is required. This evaluation would use a multi-index evaluation method, considering not only the geometric error sources themselves, but also their interactions with other parameters.

The evaluation result is used as a basis for adjusting welding parameters. The goal of this stage is to find the optimal or sub-optimal welding parameter settings based on the evaluation results in order to minimize resource usage and costs while maintaining or improving the weld quality.

In the present invention, when the total sensitivity of the multiple key geometric error sources exceeds a threshold, this means that a more global, higher level of optimization is required, not just the adjustment of a single parameter; by comprehensively evaluating and optimizing a plurality of parameters, resources can be utilized more effectively; and comprehensively evaluating and optimizing to find out parameter settings capable of stably improving welding quality.

In one embodiment, it is assumed that there is a particular type of critical geometric error source, such as "weld geometry," which includes the width and depth of the weld. Assuming sensitivities of 0.8 and 0.9, respectively, the total sensitivity is 0.8×0.5+0.9×0.5=0.85, exceeding the fourth threshold value of 0.8.

All equipment parameters affected by these geometric parameters, such as welding speed and welding pressure, are evaluated. Assuming that the evaluation result shows that the welding speed needs to be reduced and the welding pressure needs to be increased; you will reduce the welding speed and increase the welding pressure according to the evaluation result. For example, if the current welding speed is 4mm/s, you would decrease it to 3.8mm/s; if the current welding pressure is 150MPa, you will increase it to 155MPa.

Thus, when the total sensitivity of a particular type of critical geometric error source exceeds a fourth threshold, the problem can be more fully solved by performing a comprehensive evaluation and parameter adjustment.

Preferably, as shown in fig. 3, the performance indexes include: and (3) comparing the performance index with the overall performance change value to finish the evaluation of the performance index, wherein the welding precision, the welding speed, the equipment utilization rate and the weld quality are the same.

Performance metrics are key elements of the quantitative task objective. These indicators reflect aspects of the welding process, such as operating efficiency, economy of equipment use, weld quality, etc., either directly or indirectly.

The overall performance change is a more comprehensive quantitative representation that is derived by integrating the change values and weights of multiple geometric error sources. This index not only quantifies the performance of a single aspect, but also fuses aspects, giving a more global view.

By comparing the performance indicators with the overall performance change values, a more comprehensive and comprehensive evaluation result can be obtained. This helps to identify which performance index or indices are most consistent with the overall performance change value, and thus can be more specifically optimized.

Before comparison, each performance index and overall performance change value need to be normalized to be compared under the same dimension. Furthermore, the individual performance indicators have different weights, which also need to be taken into account when calculating the overall performance change value.

This step allows the system to self-evaluate and adjust in a more scientific and accurate manner. Optimizing a performance index is more effective in improving overall performance if the index is highly consistent with the overall performance variation. Otherwise, if a certain index is inconsistent with the overall performance change value, the weight of the index needs to be considered again or the calculation mode of the overall performance change value needs to be modified.

In one embodiment, it is assumed that the performance index of the welding speed and the equipment utilization is highly consistent with the overall performance variation value in one welding task, while the welding accuracy is relatively low. This means that optimizing the welding speed and the equipment utilization improves the overall performance even more. Thus, in the subsequent optimization process, special attention may be paid to both of these metrics, for example, to increase the equipment utilization by adjusting the workflow or optimizing the equipment schedule, or to increase the welding speed by changing the welding parameters.

Preferably, as shown in fig. 4, the optimizing the deep learning model and the adjustment strategy based on the evaluation result includes:

when the estimation result is that the welding precision is poor, fine tuning parameters related to a key geometric error source in the deep learning model; in the regulation strategy, the welding speed is reduced, and the welding pressure is increased; this means that model weights, activation functions, etc. are optimized to increase the sensitivity of the model to geometric errors; at the same time, the welding speed is reduced and the welding pressure is increased, which is based on physical principles, since slowing down and increasing the pressure generally improves the weld quality.

When the evaluation result is that the welding speed is low, retraining the deep learning model, and adjusting the key geometric error sources related to the speed; in the regulation strategy, the welding speed is increased, and the welding pressure is reduced; the deep learning model needs to be retrained here, with emphasis on adjusting to the key geometric error sources associated with the welding speed, increasing the welding speed and reducing the welding pressure can improve the overall production efficiency, but also needs to ensure that the welding quality is not sacrificed.

When the evaluation result is that the equipment utilization rate is low, analyzing related key geometric error sources and other additional factors influencing the equipment utilization rate, and adjusting the deep learning model based on the information; in the adjustment strategy, the device scheduling strategy is adjusted, and in this case, in addition to the geometric error source, other factors affecting the device utilization rate are considered. Such as equipment maintenance time, raw material supply, etc. And properly adjusting the deep learning model and simultaneously optimizing the equipment scheduling strategy.

The device scheduling policy is initially set to an "initial state", which typically will be an empirical or simple rule-based policy for managing the allocation and use of devices under normal operating conditions. When the evaluation result shows that the device utilization is lower than the expected value, the device scheduling policy in the initial state will be reviewed and adjusted again. The following are possible optimization directions:

dynamic priority reassignment: and re-evaluating the priority of each welding task, and carrying out equipment allocation according to the new priority.

Flexible time window scheduling: the original time window is adjusted to more flexibly cope with changing workload or other unpredictable factors.

Load rebalancing: based on the real-time device operating status, the upcoming tasks are reassigned to the lower load devices.

Predictive maintenance adjustment: if the device is about to require maintenance, the high priority tasks are prioritized and the low priority tasks are rearranged to other devices.

Multiplexing compatibility reevaluation: in view of the multitasking capabilities of the device, it is reevaluated which tasks may be performed in parallel to increase device utilization.

Energy consumption optimization considers: in the process of strategy adjustment, the energy efficiency of the equipment is also considered, and tasks are allocated to the equipment with higher energy efficiency as much as possible.

These adjustments are made based on real-time assessment and the output of the deep learning model with the aim of improving the actual utilization of the equipment while maintaining or improving the weld quality and efficiency. The dynamic adjustment mechanism enables the equipment to be more flexible to schedule and quick to respond, and can better adapt to the continuously changing requirements and conditions in the production environment.

The overall optimization process is dynamic and adaptive, and model and strategy optimization is based on real-time or near real-time assessment results. This allows the overall system to accommodate changing production conditions and requirements.

The optimized model and strategy of the invention should show significant improvements in terms of improving welding accuracy, increasing welding speed and improving equipment utilization. This will ultimately lead to reduced costs, improved production efficiency and improved product quality.

In one embodiment, the welding accuracy is poor: if the evaluation result shows that the welding precision is low, a new welding task shows higher welding precision under the same condition after the fine adjustment of the model and the parameter adjustment.

The welding speed is slow: assuming that 10 welding tasks are originally completed per hour, 12 welding tasks can be completed per hour through optimization of a model and a strategy, and meanwhile, the welding quality is maintained or improved.

The equipment utilization rate is low: the original equipment utilization rate is 60%, and the utilization rate is improved to 75% after optimization, which means less idle time and higher production efficiency.

These examples illustrate that by optimizing the deep learning model and the adjustment strategy based on the evaluation result, various performance indexes of the production process can be significantly improved.

The scheme is based on the ideas of deep learning and dynamic optimization, in particular to a Convolutional Neural Network (CNN) so as to solve the geometric error problem and performance optimization in the welding industry. The scheme covers the following key links:

prior to model training, weld data is cleaned and normalized to reduce noise and facilitate modelTraining of the model; the overall performance variation is quantified using a mathematical expression that accounts for the variation values (deltae _i ) And corresponding weights (w _i ) The method comprises the steps of carrying out a first treatment on the surface of the Pre-training the CNN by utilizing historical welding data to predict the influence of each geometrical error source on the whole performance; quantifying the influence or 'sensitivity' of each geometrical error source by a pre-trained model; setting a threshold value, and identifying which geometric error sources are key factors through sensitivity and overall performance change values; if the sensitivity of the key geometric error source exceeds a preset threshold, the welding parameters are adjusted in real time according to the sensitivity; if a plurality of key error sources exist, calculating a comprehensive adjustment coefficient, and adjusting welding parameters according to the coefficient; dynamically fine-tuning model parameters and adjustment strategies according to evaluation results (such as welding precision, welding speed, equipment utilization rate and the like); and finally, comparing various performance indexes with the calculated overall performance change value to complete comprehensive evaluation of the performance.

According to the application, by adjusting welding parameters in real time, various geometric errors can be effectively reduced by the scheme, so that the welding precision is improved; by optimizing the welding speed and the equipment utilization rate, the scheme is beneficial to improving the overall production efficiency; the scheme can obviously reduce the production cost by reducing geometric errors and improving the utilization rate of equipment; fine welding parameter control and real-time optimization are helpful for improving the quality of the final product; the scheme of the application has very high self-adaptability and flexibility, and can dynamically optimize according to the actual production environment and requirements; the scheme provides a comprehensive and quantitative performance assessment by comparison with predetermined performance metrics.

In summary, the scheme of the application provides a comprehensive, self-adaptive and highly optimized solution, which can remarkably improve the performance of the welding industry in terms of geometric error control and performance optimization.

It will be appreciated by those skilled in the art that embodiments of the present application may be provided as a method, system, or computer program product. Accordingly, the present application may take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects. Furthermore, the present application may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, and the like) having computer-usable program code embodied therein.

The present application is described with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems) and computer program products according to embodiments of the application. It will be understood that each flowchart and/or block of the flowchart illustrations and/or block diagrams, and combinations of flowcharts and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general purpose computer, special purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.

These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instruction means which implement the function specified in the flowchart flow or flows and/or block diagram block or blocks.

These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.

In one typical configuration, a computing device includes one or more processors (CPUs), input/output interfaces, network interfaces, and memory.

The memory may include volatile memory in a computer-readable medium, random Access Memory (RAM) and/or nonvolatile memory, etc., such as Read Only Memory (ROM) or flash RAM. Memory is an example of a computer-readable medium.

Computer readable media, including both non-transitory and non-transitory, removable and non-removable media, may implement information storage by any method or technology. The information may be computer readable instructions, data structures, modules of a program, or other data. Examples of storage media for a computer include, but are not limited to, phase change memory (PRAM), static Random Access Memory (SRAM), dynamic Random Access Memory (DRAM), other types of Random Access Memory (RAM), read Only Memory (ROM), electrically Erasable Programmable Read Only Memory (EEPROM), flash memory or other memory technology, compact disc read only memory (CD-ROM), digital Versatile Discs (DVD) or other optical storage, magnetic cassettes, magnetic tape disk storage or other magnetic storage devices, or any other non-transmission medium, which can be used to store information that can be accessed by a computing device. Computer-readable media, as defined herein, does not include transitory computer-readable media (transmission media), such as modulated data signals and carrier waves.

It should also be noted that the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one … …" does not exclude the presence of other like elements in a process, method, article or apparatus that comprises an element.

The foregoing is merely exemplary of the present application and is not intended to limit the present application. Various modifications and variations of the present application will be apparent to those skilled in the art. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application should be included in the scope of the claims of the present application.

Claims

1. The intelligent welding optimization and error source analysis method is characterized by comprising the following steps of:

collecting welding parameters in the welding process, and preprocessing;

2. The intelligent welding optimization and error source analysis method of claim 1, wherein the welding parameters include: weld geometry, weld speed, weld pressure, material properties, and equipment status, wherein,

the weld geometry parameters include: width, depth, and angle of the weld;

the material properties include: material type, thickness, and temperature;

3. The intelligent welding optimization and error source analysis method of claim 1, wherein the preprocessing comprises: and (5) data cleaning and normalization processing.

4. The intelligent welding optimization and error source analysis method according to claim 1, wherein the calculation expression of the overall performance change value is:

wherein ΔE is the overall performance change value, w _i Is the weight of the ith geometric error source,Δe _i is the variation value of the ith geometric error source.

5. The intelligent welding optimization and error source analysis method according to claim 1, wherein the deep learning model structure is a convolutional neural network, and the deep learning model is pre-trained by taking the welding parameters in the history welding record as input and the overall performance change value as output, so as to predict and analyze the influence of each geometrical error source on the overall performance.

6. The intelligent welding optimization and error source analysis method according to claim 1, wherein the input of the welding parameters and the overall performance variation values into the pre-trained deep learning model obtains a calculation expression of the sensitivity of each geometrical error source as follows:

wherein S is _i Is the sensitivity value of the ith geometric error source.

7. The intelligent welding optimization and error source analysis method according to claim 1, wherein a geometric error source with sensitivity greater than a first threshold and an overall performance variation value greater than a second threshold is used as the key geometric error source.

8. The intelligent welding optimization and error source analysis method of claim 1, wherein the tuning strategy comprises:

when the sensitivity of a certain key geometric error source exceeds a third threshold, the welding parameters are adjusted, and the third threshold is determined based on the sensitivity and the overall performance change value; if a certain key geometric error source affects the welding speed, adjusting the numerical value according to the sensitivity by reducing or increasing the welding speed; if a certain key geometric error source influences the welding pressure, the welding pressure is adjusted according to the numerical value of the sensitivity; if a certain key geometric error source affects a plurality of parameters, adjusting each parameter one by one according to the weight of the sensitivity;

9. The intelligent welding optimization and error source analysis method of claim 1, wherein the performance metrics comprise: and (3) comparing the performance index with the overall performance change value to finish the evaluation of the performance index, wherein the welding precision, the welding speed, the equipment utilization rate and the weld quality are the same.

10. The intelligent welding optimization and error source analysis method of claim 1, wherein optimizing the deep learning model and the adjustment strategy based on the evaluation result comprises: