A welding system including a control module for correcting the alignment of the workpiece to be welded.

Abstract

The present invention relates to a welding system that improves the accuracy and safety of welding operations by recognizing the position and status of a workpiece in real time and controlling a welding torch and a moving guide unit. In particular, it relates to a technology that utilizes sensor data and artificial intelligence algorithms to automate the alignment of workpieces, prevent safety accidents that may occur during operation, and enhance work quality and efficiency.

Description

Technology Field

[0001] The present invention relates to a welding system that improves the accuracy and safety of welding operations by recognizing the position and status of a workpiece in real time and controlling a welding torch and a moving guide unit. In particular, it relates to a technology that utilizes sensor data and artificial intelligence algorithms to automate the alignment of workpieces, prevent safety accidents that may occur during operation, and enhance work quality and efficiency. Background Technology

[0002] In traditional welding systems, the position and orientation of the workpiece were manually adjusted by the operator, or the work was performed using simple mechanical fixtures.

[0003] However, this method had the following problems:

[0004] If the workpiece is not properly secured, the welding quality deteriorates, the defect rate increases, and additional rework becomes necessary.

[0005] There is a possibility that safety accidents may occur during welding operations due to unexpected situations such as unauthorized objects (outsiders, equipment, etc.) or falling objects.

[0006] Methods relying on manual alignment and worker intuition are slow, and quality varies significantly depending on skill level.

[0007] Although some automation systems have been introduced recently, the utilization of sensor data and video analysis technology is limited, and technology for detecting dangerous situations in real time and responding immediately remains insufficient. Prior art literature

[0008] Korean Registered Patent Publication No. 10-2075258 The problem to be solved

[0009] The present invention is proposed to solve the aforementioned problems and aims to solve the following technical problems:

[0010] It analyzes the positional and rotational angle errors of the workpiece in real time and automatically performs alignment correction via guide rails to standardize welding quality.

[0011] It detects unauthorized objects, moving objects, or unexpected situations within the workplace, and based on this, stops the operation of the welding torch or generates a warning to prevent safety accidents.

[0012] The welding path and work conditions are optimized through a mathematical model (Equation 1, Equation 2), and the work speed is improved by minimizing worker intervention.

[0013] Data collected during the welding process is accumulated and utilized for quality analysis and process optimization, and changes in the future work environment are prepared through deep learning-based prediction models.

[0014] The problems of the present invention are not limited to those mentioned above, and other problems not mentioned will be clearly understood by those skilled in the art from the description below. means of solving the problem

[0016] A welding system according to an embodiment of the present invention for solving the above problem may include: a support frame portion for stably fixing a workpiece; a welding torch that generates a high-temperature arc to perform a welding operation on the workpiece; a movement guide portion for moving the welding torch along a designated welding path; and a control module that analyzes sensor data and image data and controls the welding torch and the movement guide portion.

[0017] In one embodiment, to ensure welding quality in real time, the control module obtains a first error group including x-axis position error and rotation angle error of the workpiece from a sensor, and a second error group including deviations of each of a plurality of pre-set critical points on the workpiece; determines a final alignment index by combining a first intermediate evaluation value, calculated by combining the x-axis position error and the rotation angle error in a mutual multiplication relationship rather than a simple error summation method so that the influence of the other error is amplified when one error increases, and a second intermediate evaluation value, calculated by applying an exponential relationship where the risk increases exponentially as the deviations of the critical points accumulate; accurately predicts potential defect factors even in complex error occurrence scenarios; and guarantees uniform and high-quality welding results by immediately driving the moving guide unit to automatically correct the alignment of the workpiece when the final alignment index exceeds a pre-set threshold.

[0018] In one embodiment, the control module quantitatively determines the risk level within the workspace to preemptively prevent safety accidents that may occur during operation, calculates a path deviation which is the difference between the number of pre-set allowable movement paths from the camera and the number of paths of the actual detected moving object, calculates a unique risk score for each object based on the type, size, or movement speed of each object identified through image analysis, performs logic in which the risk level is non-linearly amplified as the path deviation increases and is combined with the accumulated risk of the objects in a mutually multiplicative relationship to calculate a final emergency stop index, accurately identifies a complex risk situation with the highest probability of accident occurrence where unpredictable movements and high-risk objects appear simultaneously, and generates an emergency stop signal to stop all operations of the welding device as soon as the final emergency stop index exceeds a safety threshold, thereby preventing serious accidents at the source.

[0019] In one embodiment, the system may further include a server and a user terminal linked to the control module. The server receives work commands from the user terminal, generates a work plan including welding paths, quality standards, and safety standards, and transmits it to the control module. The control module reports critical situational data, such as alignment correction events or emergency stop events occurring during the welding process, to the server in real time for storage. Additionally, in the event of an emergency stop, the server immediately transmits a warning message and relevant site information to the user terminal, allowing the operator to recognize the situation via the user terminal and issue a command to resume work after the risk is cleared. Through this organic linkage, the autonomous quality and safety management functions of the site and the process management and operator interface functions of the central server are combined to maximize the level of automation and management efficiency of the entire welding process. Effects of the invention

[0021] The welding system according to the present invention provides the following effects:

[0022] By detecting and automatically correcting factors such as workpiece position error, rotation angle error, and gap spacing in real time, welding quality can be standardized and rework reduced.

[0023] It detects unauthorized objects and unexpected situations in real time through cameras and sensors, and maximizes safety by immediately stopping the operation of the welding torch or generating a warning.

[0024] It can automatically set the work path and conditions, and perform fast and accurate welding operations by linking the moving guide unit and the welding torch.

[0025] Based on accumulated data, it can be used to train deep learning models for quality analysis, process optimization, and safety accident prevention, thereby continuously improving the performance of the system over the long term.

[0026] It is applicable to workpieces of various sizes and shapes, and the system components are modularized to allow for easy integration into existing process lines.

[0027] The effects according to the present invention are not limited to those exemplified above, and a wider variety of effects are included within the present invention. Brief explanation of the drawing

[0029] FIGS. 1 to 3 illustrate an embodiment of the present invention. Figure 4 schematically illustrates the arrangement relationship of the present invention. Figure 5 schematically illustrates the relationship diagram of the present invention. Specific details for implementing the invention

[0030] Hereinafter, various embodiments are described in more detail with reference to the attached drawings. The embodiments described in this specification may be modified in various ways. Specific embodiments may be depicted in the drawings and described in detail in the detailed description. However, specific embodiments disclosed in the attached drawings are intended only to facilitate understanding of various embodiments. Accordingly, the technical concept is not limited by specific embodiments disclosed in the attached drawings, and it should be understood that it includes all equivalents or substitutions that fall within the spirit and scope of the invention.

[0031] Terms including ordinal numbers, such as first, second, etc., may be used to describe various components, but these components are not limited by the aforementioned terms. The aforementioned terms are used solely for the purpose of distinguishing one component from another.

[0032] Functions related to artificial intelligence according to the present disclosure are operated through a processor and memory. The processor may be composed of one or more processors. In this case, the one or more processors may be general-purpose processors such as CPUs, APs, and DSPs (Digital Signal Processors), graphics-dedicated processors such as GPUs and VPUs (Vision Processing Units), or artificial intelligence-dedicated processors such as NPUs. The one or more processors control the processing of input data according to predefined operation rules or artificial intelligence models stored in memory. Alternatively, if the one or more processors are artificial intelligence-dedicated processors, the artificial intelligence-dedicated processors may be designed with a hardware structure specialized for processing a specific artificial intelligence model.

[0033] The predefined rules of operation or artificial intelligence models are characterized by being created through learning. Here, being created through learning means that a basic artificial intelligence model is trained using a number of training data by a learning algorithm, thereby creating predefined rules of operation or artificial intelligence models configured to perform desired characteristics (or objectives). Such learning may be performed on the device itself where the artificial intelligence according to the present disclosure is executed, or it may be performed through a separate server and / or system. Examples of learning algorithms include supervised learning, unsupervised learning, semi-supervised learning, or reinforcement learning, but are not limited to the examples described above.

[0034] An artificial intelligence model can be composed of multiple neural network layers. Each of the multiple neural network layers has multiple nodes and weight values, and performs neural network operations through calculations between the results of previous layers and the multiple weights. The multiple weights possessed by the multiple neural network layers can be optimized based on the learning results of the artificial intelligence model. For example, multiple weights can be updated so that the loss value or cost value obtained by the artificial intelligence model during the learning process is reduced or minimized. Additionally, to minimize the loss value or cost value, multiple weights can be updated in a direction that minimizes the gradient associated with the loss value or cost value. Artificial neural networks may include deep neural networks (DNNs), such as Convolutional Neural Networks (CNNs), Deep Neural Networks (DNNs), Recurrent Neural Networks (RNNs), Restricted Boltzmann Machines (RBMs), Deep Belief Networks (DBNs), Bidirectional Recurrent Deep Neural Networks (BRDNNs), or Deep Q-Networks, but are not limited to the examples mentioned above.

[0035] A network is a network that serves as a transmission path for web pages; it may be a closed network such as a LAN (Local Area Network) or WAN (Wide Area Network), but it is desirable for it to be an open network such as the Internet. The Internet refers to a global open computer network structure that provides the TCP / IP protocol and various services existing at its upper layers, namely HTTP (HyperText Transfer Protocol), Telnet, FTP (File Transfer Protocol), DNS (Domain Name System), SMTP (Simple Mail Transfer Protocol), SNMP (Simple Network Management Protocol), NFS (Network File Service), and NIS (Network Information Service).

[0036] Terminals can be implemented in various forms. For example, the terminals described in this specification may include mobile terminals such as smartphones, tablet PCs, PDAs, portable multimedia players, and MP3 players, as well as fixed terminals such as smart TVs and desktop computers.

[0037] In this specification, terms such as “comprising” or “having” are intended to specify the existence of the features, numbers, steps, actions, components, parts, or combinations thereof described in the specification, and should be understood as not precluding the existence or addition of one or more other features, numbers, steps, actions, components, parts, or combinations thereof. When a component is described as being “connected” or “connected” to another component, it should be understood that it may be directly connected to or connected to that other component, or that there may be other components in between. On the other hand, when a component is described as being “directly connected” or “directly connected” to another component, it should be understood that there are no other components in between.

[0038] Meanwhile, a "module" or "part" for a component as used in this specification performs at least one function or operation. Furthermore, a "module" or "part" may perform a function or operation by hardware, software, or a combination of hardware and software. Additionally, a plurality of "modules" or a plurality of "parts," excluding a "module" or "part" that must be performed on specific hardware or on at least one processor, may be integrated into at least one module. A singular expression includes a plural expression unless the context clearly indicates otherwise.

[0039] In addition, power, power transmission, and control therefor for the following assembly configurations and embodiments, including "by control," follow conventional technology including terminals, applications, hardware control modules, etc., so they are omitted to avoid redundancy.

[0040] In addition, the operation embodiments and configurations described in a general manner without being explained in detail below follow the prior art and are omitted in order to focus on describing the purpose of the present invention and the resulting effects.

[0041] Furthermore, in describing the present invention, if it is determined that a detailed description of related known functions or configurations may unnecessarily obscure the essence of the invention, such detailed description is abbreviated or omitted.

[0042] The welding system may include a support frame (10), a welding torch (20), a moving guide (30), a control module (40), a server (100), and a worker terminal (200).

[0043] The support frame part (10) is a structure for securely and stably fixing the workpiece, and is a modular frame designed to support workpieces of various sizes and shapes.

[0044] In addition, the support frame part (10) can firmly fix the workpiece through a hydraulic or electronic clamp and can support various welding angles by adjusting the position and angle of the frame.

[0045] In one embodiment, the support frame (10) may be installed vertically at both ends of the moving guide (30) installed on the ground to support the workpiece placed on the moving guide (30) from the side as the rail of the moving guide (30) moves.

[0046] In addition, it can be designed to be positioned close to the welding torch (20) so that welding can be performed efficiently.

[0047] In addition, it is connected to a moving guide part (30) so that stable support is possible even while moving.

[0048] A welding torch (20) is a device that generates a high-temperature arc to perform welding work on a workpiece, and is equipped to support automation and precision control.

[0049] In addition, it generates high-temperature heat through the arc discharge principle and can bond the filler material and the workpiece.

[0050] In addition, the operation can be stopped immediately upon receiving a signal from the control module (40) when a dangerous situation occurs.

[0051] In addition, welding quality can be improved by dynamically adjusting the temperature according to the welding environment.

[0052] In addition, the welding position can be detected in real time, allowing for welding of the precise part of the workpiece.

[0053] In addition, work can be performed on a fixed workpiece installed on the top of the support frame part (10).

[0054] In addition, it can be connected to a moving guide part (30) and move along the welding position.

[0055] The moving guide section (30) is a structure that allows the welding torch (20) to move along a designated welding path of the workpiece and can be provided with rails, etc.

[0056] In addition, the position of the welding torch (20) can be controlled based on path data received from the control module (40).

[0057] In addition, the position can be adjusted in real time to correct errors in the workpiece.

[0058] In addition, it performs the role of moving the welding torch (20), is connected to the support frame part (10), and can execute operations by receiving commands from the control module (40).

[0059] The control module (40) can manage and control the moving guide part (30), welding torch (20), sensor, etc. in an integrated manner.

[0060] The judgment criteria used by the control module of the present invention to determine the alignment status of the workpiece consist of a basic alignment error group, a critical point deviation, and an additional quality factor group; each criterion is selected and managed considering its impact and characteristics on welding quality. The basic alignment error group is composed of an x-axis position error, which is the deviation of the actual measured x-axis coordinate relative to the reference x-axis coordinate of the workpiece, and a rotation angle error, which is the deviation of the actual tilted angle relative to the reference angle of the workpiece. The x-axis position error functions as the most fundamental measure of whether the weld line follows the correct position; since a large value of the rotation angle error causes critical defects such as uneven weld bead formation or misalignment of the weld area, it is designated as a critical management target. Additionally, the critical point deviation is included as a judgment criterion, which is the distance error between the reference coordinate and the actual measured coordinate of specific critical points defined on the CAD drawing of the workpiece, such as the center of a hole, a corner endpoint, or the center of a slot. Even if the overall alignment is correct, local deformation or errors can cause serious post-processing problems such as poor fastening; therefore, these are utilized as essential criteria for individually tracking and managing them. Furthermore, additional quality factors, which are errors in prerequisite conditions that must be secured before welding, are included as judgment criteria. In particular, the present invention manages the gap spacing, which is the deviation of the actual measured gap from the reference gap value between two base materials, using it as a major factor. Since a gap spacing that is too wide or too narrow compared to the reference value is a direct cause of poor penetration or excessive thermal deformation, it is precisely controlled.

[0061] The present invention transmits a command to determine whether final correction is required based on the four judgment criteria through a complex logical structure that reflects the physical characteristics of each error and the risk level affecting the welding process. Basic alignment errors, namely x-axis position error and rotation angle error, are processed by applying a 'logarithmic relationship.' The 'logarithmic relationship' refers to an implementation where the rate of change of the output value is greater as the input value is smaller and becomes smoother as the input value increases, by adding 1 to the input error value and taking the natural logarithm (ln), or by using a logarithmic function provided by a conventional programming language. This provides a technical basis for simultaneously ensuring system stability and precision by sensitively detecting changes even when the error is a very small value close to '0' to correct minute errors early on, and by gradually reflecting the influence of the change when the error becomes very large to prevent excessive correction operations. Additionally, the x-axis position error and rotation angle error are processed by combining them through a 'mutual multiplication relationship.' In the welding process, the two errors are not independent; while the occurrence of a single error results in a defect, the simultaneous occurrence of two errors leads to an exponential increase in the defect rate. Therefore, a multiplicative relationship is applied to accurately reflect these complex risk situations in the final judgment. Deviations at critical points are handled by applying an 'exponential relationship.' The aforementioned 'exponential relationship' refers to implementing a system where an exponential function with the natural constant e as the base is applied to the input deviation value, so that the output value increases exponentially as the input value increases. Since errors at critical points, such as holes or slots, can lead to fatal defects that render the part non-functional once they exceed a certain level, the system is designed to respond immediately and powerfully by amplifying the risk exponentially rather than proportionally as these deviations accumulate.Multiple critical point deviations or additional quality factors are integrated using a 'logic similar to vector summation,' that is, by squaring each error, summing them, and taking the square root. This effectively aggregates the overall deviation magnitude of errors occurring at various points into a single indicator and controls the system to prevent the excessive reflection of errors from specific points.

[0062] The control module acquires judgment reference data in real time from multiple sensors installed within the workplace. The x-axis position error is acquired by a 2D / 3D vision camera or a laser distance measuring sensor capturing and measuring a specific reference point of the workpiece; for example, if the laser sensor measures 1502.5mm while the reference X-coordinate is set to 1500.0mm, the x-axis position error is calculated as 2.5mm and transmitted. The rotation angle error is acquired by a vision camera recognizing the angle of the reference line of the workpiece or by an IMU sensor attached to the workpiece transmitting a tilt value; for example, if the camera recognizes the tilt of the workpiece as -3.5 degrees relative to a horizontal reference of 0.0 degrees, the rotation angle error is calculated as 3.5 degrees and managed. The critical point deviation is determined by a high-resolution camera matching CAD drawing information to locate the center point of a hole, or by a marking sensor directly measuring the coordinates of a specific point. For example, if the reference center coordinates of the hole are (x=300.0, y=500.0) and the actual measured coordinates are (x=301.2, y=499.5), the critical point deviation is calculated as approximately 1.3 mm based on the Euclidean distance. The gap spacing is measured by a laser displacement sensor scanning the gap between two base materials. For example, if the reference gap spacing is set to 1.0 mm but the actual measured average gap is 2.8 mm, the gap spacing, which is an additional quality factor, is calculated as 1.8 mm and transmitted to the control module.

[0063] The 'final alignment index' comprehensively calculated through the above logical processing procedure is compared with a pre-set threshold value and utilized to perform specific correction actions. If the final alignment index is less than threshold A, e.g., 8.0, it is determined to be in a normal state, and the welding process is allowed to proceed without separate correction. If it is greater than or equal to threshold A and less than threshold B, e.g., 15.0, it is determined to be in a 'warning' state; a warning notification is displayed as a pop-up on the operator terminal, and the system sends a command to prepare for correction. If the final alignment index is greater than or equal to threshold B, it is determined to be in an 'emergency correction' state; the operation of the welding torch is immediately paused, and an automatic alignment correction command is sent to the movement guide unit to immediately correct the position and angle of the workpiece to within the reference value range. Once the correction is complete, the index is measured again, and the process is automatically controlled to resume only after it is confirmed to be in a normal state.

[0064] In a scenario assuming that a large steel plate is mounted for welding but one end is slightly twisted and the hole position is slightly misaligned due to local thermal deformation, the control module receives data from the sensor including an x-axis position error of 1.5 mm, a rotation angle error of 2.0 degrees, a critical point deviation of 1.8 mm for hole 1, a critical point deviation of 2.1 mm for hole 2, and a gap error of 0.5 mm. The control module calculates a first intermediate evaluation value by applying mutual multiplication and a logarithmic relationship to the received x-axis position error and rotation angle error; since both errors are small but occurred simultaneously, a higher evaluation value is derived than with a simple summation method. Next, the second intermediate evaluation value is calculated by integrating the two hole deviations using a vector summation method and applying an exponential relationship; in particular, the error of 2.1 mm is amplified to a very high risk level by exponential logic. Since the gap error is within the allowable range, the third intermediate evaluation value is calculated as low. Finally, the first, second, and third intermediate evaluation values ​​are combined, and due to the exponentially amplified second evaluation value, the final alignment index is determined to exceed the 'emergency correction' threshold of 15.0. Accordingly, the control module immediately stops the welding arc generation and automatically transmits a precision drive signal to the moving guide unit to correct the x-axis position error and rotation angle error. After the correction is completed, a command is transmitted to reconfirm the alignment status by performing the procedure again from the beginning.

[0065] When the control module acquires data, if the value received from the sensor falls within a physically impossible range or shows an abnormal difference from the previous value, it determines the sensor as 'error' and replaces it with data from a backup sensor or sends a sensor inspection notification to the operator. If the final alignment index does not fall below the threshold value despite several attempts at automatic calibration, for example, three times, it considers this to be a problem that the system cannot resolve, such as permanent deformation of the workpiece, determines it as 'unable to calibrate', stops all operations, and issues a warning requesting manual intervention from the manager.

[0066] In a simple position error scenario, that is, when the x-axis position error is 3mm and there are no other errors, the prior art detects the position error and proceeds with normal welding after x-axis correction, whereas the present invention detects the position error, determines the final index to a 'warning' level, and controls the process to proceed with normal welding after x-axis correction. On the other hand, in a complex fine error scenario, that is, when a combination of an x-axis position error of 1mm, a rotation angle error of 1.5 degrees, and a critical point deviation of 1.2mm occurs, the prior art misjudges the process as 'normal' because each error is within the individual allowable limit and proceeds with welding, resulting in defects, whereas the present invention transmits a command to produce good products by preemptively correcting the final index to an 'urgent correction' level by 'mutual multiplication' and 'exponential' logic, even if each error is small, and proceeds with welding.

[0067] The operation of the present invention is verified through simulation. In Simulation 1, if an x-axis position error of 0.5 mm, a rotation angle error of 0.2 degrees, a maximum critical point deviation of 0.3 mm, and a gap error of 0.1 mm are input, the final alignment index is calculated as 1.2, and it is determined to be normal. In Simulation 2, if only the x-axis position error is 3.0 mm, the index is calculated as 7.5, and it is determined to be normal as it approaches the correction threshold. In Simulation 3, if a combination of an x-axis position error of 1.0 mm and a rotation angle error of 3.5 degrees occurs, the index is calculated as 8.9, and a warning is generated. In Simulation 4, if only the maximum critical point deviation is 2.5 mm, the index is calculated as 16.2, and an emergency correction due to the critical point error is executed. In Simulation 5, if a combination of an x-axis position error of 2.5 mm and a rotation angle error of 4.0 degrees occurs, the index is calculated as 18.5, and an emergency correction due to the combination of basic errors is executed. In Simulation 7, if only the gap error is 3.0mm, the index is calculated as 9.5, triggering a warning for a gap defect. In Simulation 8, if the x-axis position error of 2.2mm, rotation angle error of 2.8 degrees, maximum critical point deviation of 1.9mm, and gap error of 1.1mm all accumulate, the index is calculated as 22.4, triggering an emergency correction due to the total error accumulation. In Simulation 10, in the worst-case scenario where the x-axis position error of 4.0mm, rotation angle error of 5.0 degrees, maximum critical point deviation of 2.8mm, and gap error of 1.5mm occur, the index rises to 35.1, controlling the system to execute an immediate emergency correction.

[0068] According to the judgment rules of the present invention, for basic alignment errors, namely x-axis position error and rotation angle error, a logic rule of logarithmic relationship is applied for the purpose of ensuring sensitivity to minute errors and preventing over-correction, and a logic rule of mutual multiplication is applied for the purpose of reflecting the amplification of risk when complex errors occur. For critical point deviations, a logic rule of exponential relationship is applied for the purpose of a rapid and strong response to local errors that cause fatal defects, and for multiple deviation groups, a logic rule of vector summation is applied to synthesize the overall magnitude of multiple errors without distortion. The final index is calculated to make a final decision by synthesizing all risk factors through the synthesis of each intermediate evaluation value.

[0069] The judgment rules adopted in this invention are the result of technically modeling the physical phenomena and statistical correlations that cause welding defects. While a simple summation method may lead to the error of evaluating the risk of 'position error of 5mm' and 'position error of 2.5mm + rotation of 2.5 degrees' as equal, the 'mutual multiplication relationship' and 'exponential relationship' of this invention are inevitable choices to objectively reflect non-linear risk increases, thereby ensuring that a person of ordinary skill in the field clearly understands their validity based on knowledge of welding engineering and quality control. By adopting these components, it is possible to accurately predict complex error scenarios and respond preemptively, thereby eliminating potential defect factors that could not be detected by existing technologies and providing the effect of dramatically improving the reliability of welding quality in the final product. Furthermore, by minimizing unnecessary rework and avoiding overreaction to minute errors within the normal range, it prevents the loss of overall takt time. Additionally, by replacing the intuitive judgment of skilled welders with quantitative and logical data-based judgment, it elevates the level of complete automation of the welding process to the next level. The 'final alignment index' used in this invention is a quantified value representing the risk level of the alignment state calculated according to a logical processing procedure, and the 'threshold' is a criterion for distinguishing states; rather than being a fixed value, it is set by the manager through an economic analysis considering the quality standards of the welding target, the results of statistical analysis of past welding defect data, and rework costs, or automatically transmits a command that the system dynamically optimizes based on learning data.

[0070] The technical concept of the present invention described above can be implemented through the following examples of logical operations. However, the following operation formula is merely an example of implementation to explain the logical relationship of the present invention, and weights or detailed functions may be tuned according to the characteristics of the workpiece targeted by the system and the required quality level.

[0071] In calculating the first intermediate evaluation value (basic alignment error),

[0072] Evaluation_Position = Weight_x * log(1 + |x-axis position error|)

[0073] Evaluation_Angle = Weight_θ* log(1 + |Rotation_Angle_Error|)

[0074] 1st Intermediate Evaluation Value = 1 + (Evaluation_Position * Evaluation_Angle)

[0075] Here, log is a common logarithm function with base 10, and weights (e.g., weight_x=1.5, weight_θ represents the importance of each error). Adding 1 to the final value is to prevent the value from becoming 0 during subsequent multiplication operations and the influence of other evaluation values ​​from disappearing.

[0076] In calculating the second intermediate evaluation value (important point deviation),

[0077] Combined deviation = √(Σ_k_severity_point_deviation)²)

[0078] Here, k represents the unique number of an individual 'important point'.

[0079] Second Intermediate Evaluation = exp(weight_δ * integrated deviation)

[0080] Here, exp is an exponential function with base e, and weights δ (e.g., 0.8) control the amplification rate of the combined deviation on the total risk.

[0081] In calculating the final alignment index,

[0082] Final Alignment Index = 1st Intermediate Evaluation Value * 2nd Intermediate Evaluation Value

[0083] You can use it.

[0084] The 'emergency correction' threshold of the present invention (e.g., 15.0) is derived through the following data-based analysis and simulation.

[0085] In terms of data collection, an analysis of defect data from 10,000 past automobile body welding processes statistically confirmed that when the x-axis position error exceeds 2.5 mm and the deviation at a critical point of the A-pillar simultaneously exceeds 1.8 mm, the probability of reduced body rigidity and door misalignment defects reaches 95%. This is defined as a 'critical defect condition' that must be avoided.

[0086] Based on the above 'critical failure condition', 10 simulations for various error combinations were performed using the above 'logical operation example', and the results are as shown in the table below.

[0087] [Table 1-1]

[0088] As a result of the above simulation, the index in Simulation No. 6, which corresponds to a 'critical defect condition,' surges to 17.55, showing a clear difference from the previous one, No. 5 (13.98). Therefore, to reliably prevent such critical defects in advance, a safety margin considering system stability is applied, and the 'emergency correction' threshold value is set to 15.0. In addition, since the risk level increases significantly from an index of approximately 7.0 or higher, the 'warning' threshold value is set to 7.0. As such, the threshold values ​​of the present invention are technical values ​​set based on actual process data and statistical analysis.

[0089] To verify the validity of the set thresholds (Warning: 7.0, Emergency Correction: 15.0), a simulation was performed on 5 actual past failure and normal cases, and the results are as shown in the table below.

[0090] [Table 1-2]

[0091]

[0092] As a result of the verification simulation above, the present invention predicted and determined actual process results with 100% accuracy based on the set threshold values. In particular, even for cases 3 (minor rework) and 5 (critical defect), which are difficult to judge based solely on simple error values, the risk level was accurately calculated through complex logical operations to make a correct judgment. This demonstrates that the threshold setting method and control logic of the present invention are highly effective and reliable.

[0094] The judgment criteria used by the control module of the present invention to determine the safety risk level of a welding workspace in real time consist of a path prediction mismatch group, a path deviation, and an object-specific risk group, and each criterion is selected and managed to predict the probability and severity of potential safety accidents. The path prediction mismatch group includes the number of allowable movement paths, which is the total number of paths permitted for safe movement of workers or logistics equipment and set in advance by a work manager, and the number of actual movement paths, which is the total number of movement trajectories of objects actually detected within the workspace by a vision camera or LiDAR sensor. The number of allowable movement paths serves as a baseline for the system to recognize normal conditions. Additionally, the judgment criterion includes the path deviation, which is the difference between the 'number of allowable movement paths' and the 'number of actual movement paths.' This value functions as a direct indicator of how much unexpected movement is occurring within the workspace and as a key criterion for quantifying the magnitude of potential risk. Furthermore, the types and number of unauthorized objects, which identify whether the detected object is a person, a forklift, an unmanned transport vehicle, etc., and determine the number of objects of each type, and the object-specific risk group, which includes the object's intrinsic risk score—a quantified score assigned to each identified object according to predefined rules regarding the severity of the potential risk the object poses to welding operations—are utilized as judgment criteria. The object's intrinsic risk score is controlled to evaluate the qualitative aspects of the risk beyond merely the presence or absence of the object.

[0095] The present invention transmits a command to determine whether to perform a final emergency stop through a sophisticated logical structure for detecting complex risk situations that occur when the above judgment criteria are combined, going beyond evaluating the criteria individually. For 'path deviation,' the value is first squared, and then processed by applying an 'exponential relationship.' Squaring the path deviation reflects the reality that risk increases non-linearly as unpredictability increases, by ensuring that the increase in risk when the number of unpredictable paths increases from one to two is much greater than the increase when they increase from three to four. Furthermore, by additionally applying an exponential relationship to further amplify this non-linear increase, it provides a technical basis for the system to react very conservatively and rapidly to unpredictable situations. The 'inherent risk scores' of multiple unauthorized objects simultaneously detected within the workspace are integrated using 'logic similar to vector summation,' that is, processed by squaring the risk score of each object, summing them, and then taking the square root. This allows for determining that a situation with one forklift with a high risk score is more dangerous than a situation with three workers with low risk scores, thereby accurately evaluating the actual level of risk by assigning greater weight to the mere presence of a high-risk object rather than simply adding risk scores. Finally, the risk derived from 'path prediction discrepancy' and the risk derived from 'object-specific risk' are combined through a 'mutual multiplication relationship.' This quantitatively evaluates that it is much more dangerous for a forklift to approach via an unpredictable path by reflecting the 'synergy effect,' where the two risks amplify each other and the probability of an accident increases explosively when unpredictable movement and a high-risk object occur simultaneously.

[0096] The control module collects and analyzes safety-related data in real time from wide-angle cameras, 3D depth cameras, LiDAR sensors, etc., installed at key points in the workplace. During the system setup phase, the work manager designates two main movement paths for workers and one path for a material transport AGV on the drawing; for example, a total of three allowed movement paths are set in the system. The actual number of movement paths is determined by a video analysis algorithm that tracks the trajectories of objects moving within the workspace in real time and counts them. For example, assume that one worker is currently moving along allowed path A and one external forklift has entered a new, unallowed path C. In this case, the system determines the 'total currently active paths' as the sum of the 'number of currently used allowed paths' and the 'number of newly appeared unallowed paths'. If allowed path B is empty, the system recognizes that a total of two paths are active by adding the one currently used allowed path (A) and the new unallowed path (C). Therefore, since there are 2 actual active paths when the allowed paths are set to 3, the path deviation is calculated as 1 and transmitted. The unique risk score for each object is calculated by a pre-trained machine learning model or rule-based logic based on the type, size, and movement speed of the identified object, and is managed by assigning scores such as 3 points for a worker wearing a safety helmet, 9 points for a forklift moving at high speed, and 1 point for a small cart in a stationary state.

[0097] The 'Final Emergency Stop Index,' comprehensively calculated through the above logical processing procedure, is compared with a pre-set safety threshold and utilized to automatically perform step-by-step measures. If the Final Emergency Stop Index is below Safety Threshold A, e.g., 150, it is determined to be in a 'Safe' state, and all systems operate normally. If it is above Safety Threshold A but below Safety Threshold B, e.g., 250, it is determined to be in a 'Caution' state, the warning lights installed around the welding torch flash, and a 'Nearby Hazard Detected' notification is sent to the operator's terminal. If the Final Emergency Stop Index is above Safety Threshold B, it is determined to be in a 'Dangerous' state, an 'E-Stop' signal is generated to immediately halt all welding torch operations, and a loud warning sound is emitted throughout the entire workplace to ensure all personnel are aware of the situation. The system is controlled so that it can only be restarted after the hazard disappears and the index returns to a 'Safe' state, following a manual verification procedure by a manager.

[0098] In a scenario where two allowed paths are set, and one worker is moving along an allowed path while one forklift from the outside rapidly approaches the welding robot via an unallowed path, the control module collects data on the number of allowed movement paths (2) and the detected objects: one worker (risk level 3) and one forklift (risk level 9). At this time, the actual active paths are recognized as a total of two by adding the one allowed path currently in use and the new unallowed path, and thus the path deviation is calculated as 0. Next, the risk levels of the worker (3 points) and the forklift (9 points) are integrated using a vector summation method to calculate a second intermediate evaluation value for a 'cumulative risk level' of approximately 9.5. Finally, the first intermediate evaluation value and the second intermediate evaluation value are combined in a mutual multiplication relationship to calculate the final emergency stop index; since unpredictable movement and the presence of high-risk objects occurred simultaneously, the final emergency stop index is determined by exploding to a value far exceeding the 'risk' threshold of 250. Accordingly, the control module immediately generates an E-Stop signal to halt all operations of the welding robot and transmits a command to sound a loud siren alarm to spread awareness of the dangerous situation.

[0099] The system fuses data from multiple cameras and LiDAR sensors using 'sensor fusion' technology. This ensures the reliability of judgments by continuously tracking the location and path of an object based on information detected by other sensors, even if one sensor is obscured by a pillar or other object and misses it. In cases where an operator is standing still or a tool cart is temporarily placed, the system performs exception handling logic to reduce unnecessary warnings by reclassifying the object as a 'fixed obstacle' and excluding it from the 'actual movement path' count if there is no movement for a set time, for example, 5 seconds or longer.

[0100] In a scenario where two workers move along an allowed path, the prior art detects two objects within the danger zone and generates an unnecessary 'caution' alert, whereas the present invention controls the system to maintain a 'safe' state because the number of allowed paths matches the number of actual paths, resulting in a path deviation of 0. On the other hand, in a scenario where one forklift approaches via an unexpected path, the prior art detects one object within the danger zone and is limited to a 'caution' alert level, whereas the present invention determines the situation as 'dangerous' because the path deviation and the high risk level of the forklift are multiplied to cause the final index to skyrocket, and immediately issues an E-Stop command to prevent an accident.

[0101] According to the judgment rules of the present invention, in order to respond quickly by non-linearly amplifying the risk of unpredictable movements, logical rules of squared deviation and exponential relationships are applied to path deviations, and in order to evaluate the existence of high-risk objects more heavily and to synthesize the risk of multiple objects, logical rules of vector summation are applied to multi-object risk. The final index is calculated by applying logical rules of mutual multiplication relationships to detect complex risk situations by reflecting the synergistic effect of unpredictability and object risk.

[0102] The safety judgment rule of the present invention is based on a statistical analysis of industrial site accident types. Taking into account that most serious accidents occur when "high-risk equipment" enters via "unpredictable paths," the present invention adopts an inevitable logical structure to objectively identify specific scenarios with a high probability of accident occurrence and to respond preemptively by combining "path deviation" and "object risk" through "multiplication." By adopting these components, the invention accurately targets "unpredictable high-risk situations" with the highest potential for accidents and immediately halts the system, thereby fundamentally preventing the possibility of serious accidents. Furthermore, by not generating unnecessary alarms for normal and safe worker movements, it prevents a decline in worker concentration and maintains productivity by not disrupting the overall workflow. It also achieves the effect of implementing an intelligent, autonomous safety management system that determines risk levels by analyzing dynamic changes in the actual work environment in real time. The 'final emergency stop index' used in this invention is a quantified value representing a complex risk level calculated according to safety judgment logic, and the 'safety threshold' is a criterion for distinguishing states. It automatically transmits a command that is set by a safety manager by comprehensively considering engineering data such as the safety management regulation grade of the workplace, analysis of past near-miss data, and the maximum movement speed and braking distance of the welding robot, or is set variably depending on the type of work or equipment deployed.

[0103] The safety judgment logic of the present invention described above can be implemented through the following examples of logical operations. However, the following formula is merely an example of implementation to explain the technical concept of the present invention, and detailed parameters may be adjusted according to the safety regulations and working environment of each workplace.

[0104] In calculating the first intermediate evaluation value (unpredictability),

[0105] Path deviation = |Actual number of active paths - Number of allowed movement paths|

[0106] First Intermediate Evaluation = exp(weight_α * (path deviation)²)

[0107] Here, weight_α (e.g., 0.5) is a safety factor that adjusts the sensitivity of unpredictability to the overall risk.

[0108] In calculating the second intermediate evaluation value (cumulative risk),

[0109] Risk per object = Base risk * Speed ​​weight * Size weight

[0110] Second Intermediate Assessment Value = √(ΣObject-specific Risk_u)²)

[0111] Here, the base risk level is determined by the type of object (person=3, forklift=8), the speed and size weights reflect the dynamic state of the detected object, and u represents the unique number of an individual 'object' detected within the workspace.

[0112] In calculating the final emergency stop index,

[0113] Final Emergency Stop Index = 1st Intermediate Evaluation Value * 2nd Intermediate Evaluation Value

[0114] You can use it.

[0115] The threshold value for determining the 'risk' state of the present invention (e.g., 250) is derived through the Industrial Safety and Health Act and an analysis of serious accident occurrence scenarios within the workplace.

[0116] In terms of data analysis, an analysis of near miss data from the past three years within the workplace revealed that the most frequent and fatal risk scenario was when a forklift traveling at a speed of 10 km / h or more approached within a radius of 3 meters of a welding robot via an unauthorized route. This is defined as a "condition of imminent serious accident."

[0117] Based on the above 'imminent serious accident condition', 10 simulations for various risk combinations were performed using the above 'logical operation example', and the results are as shown in the table below.

[0118] [Table 2-1]

[0119]

[0120] As a result of the above simulation, the index surges to 280.5 in Simulation No. 6, which corresponds to the 'condition of imminent serious accident,' showing a clear difference from the previous 'caution' level situation (No. 5, 145.2). Therefore, to reliably prevent such serious accidents, a safety margin considering the system response speed is applied, and the threshold for determining the 'danger' state is set to 250. Additionally, since it is determined that management is required when the index approaches approximately 150, the threshold for determining the 'caution' state is set to 150.

[0121] To verify the validity of the set thresholds (Caution: 150, Danger: 250), simulations were performed on five actual near misses and normal operation cases that occurred in the past, and the results are shown in the table below.

[0122] [Table 2-2]

[0123]

[0124] As a result of the verification simulation above, the present invention accurately determined the risk level of actual near-miss situations with 100% accuracy based on the set threshold. In particular, regarding the serious near-miss situations in cases 4 and 5, which would have resulted in a simple 'caution' alert under prior art, the present invention was able to accurately calculate the complex risk level, determine it as 'dangerous,' and issue an E-Stop. This strongly demonstrates that the threshold setting and safety logic of the present invention can dramatically improve safety in actual industrial sites.

[0126] The overall operation process of the welding system according to the present invention is explained through a specific scenario assuming a process of welding a side panel of an automobile body. Worker A logs into a control application installed on their user terminal, selects the SUV model 'GX-5' as the body model to be produced today, enters a production quantity of 100 units, and then transmits a 'Start Work' command to the server. The server retrieves the CAD data and standard welding procedure for the 'GX-5' model from the database and, based on the said data, generates a work plan that includes welding path data, which is a 3D coordinate path where the welding torch must move; quality standard data including an allowable x-axis position error of ±1.0mm, an allowable rotation angle error of ±1.5 degrees, and an allowable deviation of ±0.8mm for critical points of the A-pillar hole; and safety standard data including three allowed movement paths and a basic risk score for each object. The server transmits the generated work plan to the control module at the site via a wireless network.

[0127] When the material transport AGV transports and lowers the side panel to be welded onto the support frame, the hydraulic clamp of the support frame operates to initially secure the panel. The control module recognizes the securing signal and immediately activates the alignment correction logic, measuring the current state of the panel through a vision camera and a laser sensor. For example, if the measurement results show an x-axis position error of -2.8mm, a rotation angle error of +2.1 degrees, and an A-pillar hole deviation of 1.6mm, the control module calculates the 'final alignment index' as 16.8 according to the built-in 'logical operation example'. Since this value exceeds the set 'emergency correction' threshold of 15.0, the control module transmits a precision control signal to the movement guide unit. Accordingly, the servo motor of the movement guide unit operates to move the support frame unit by +2.8mm in the x-axis direction and simultaneously rotate it to correct the angle by -2.1 degrees. After calibration, the control module re-measures and confirms that all errors are within the allowable range and the final alignment index is 2.5, which is within the 'normal' range, and sends a command to the server to report the 'alignment complete and welding ready' status.

[0128] Upon receiving a welding start command from the server, the control module simultaneously controls the welding torch and the moving guide to initiate automatic welding along the planned welding path, at which point a welding arc is generated and a weld bead begins to form. During this process, Operator A monitors information in real time through their user terminal screen, including live camera footage showing the current position of the welding torch, current welding parameters, and graphs of fluctuations in the 'final alignment index' and 'final emergency stop index' calculated in real time.

[0129] While welding is in full swing, an unexpected situation occurs where an external forklift transporting parts deviates from its designated path and rapidly approaches the welding robot. The control module's LiDAR sensor and camera immediately detect this situation and activate the emergency stop logic at high speed. For example, if the detected data shows a path deviation of 1, the object type is a forklift (basic risk level 8), and the speed is 12 km / h (high speed weighting), the control module calculates a 'final emergency stop index' of 321.4 based on a built-in 'logical operation example'. Since this value exceeds the set 'danger' state judgment threshold of 250, the control module immediately sends an E-Stop signal to the welding torch to stop the arc generation and all movements of the robot arm, sounds a siren alarm throughout the entire workplace, and simultaneously performs the action of reporting the occurrence of an 'emergency stop' event to the server.

[0130] The server receives an emergency stop signal and immediately transmits a video to Worker A's user terminal screen displaying a pop-up message stating, "Danger! Unauthorized high-speed object approach! The system has been emergency stopped!" along with the location where the forklift was detected marked in red. Upon hearing the alarm, the forklift operator immediately stops and reverses to a safe area, and Worker A confirms via the user terminal that the hazard has disappeared. The control module detects that the forklift has moved out of the safe area and, after confirming that the 'final emergency stop index' has returned to the 'safe' range, transmits the 'hazard situation cleared, waiting to resume work' status to the server and the user terminal. A message stating, "The hazard situation has been cleared. Please check the site and press the 'Resume Work' button," is displayed on the user terminal screen, and Worker A presses the 'Resume Work' button on the screen after visually confirming the safety of the site. This command is transmitted to the control module via the server, and the control module controls the welding torch and the moving guide unit to restart welding from the point where it was stopped. This ensures that the entire welding cycle is safely completed, and after completion, all work logs, such as alignment correction records and emergency stop event records, are stored in the server's database and automatically transmitted to be utilized as data for quality tracking and process improvement.

[0131] Although preferred embodiments of the present invention have been illustrated and described above, the present invention is not limited to the specific embodiments described above. Various modifications are possible by those skilled in the art without departing from the essence of the invention as claimed in the claims, and such modifications should not be understood individually from the technical spirit or perspective of the present invention. Explanation of the symbols

[0132] Support frame part (10) Welding torch (20) Movement guide section (30) control module (40) Server (100) User terminal (200)

Claims

Claim 1 A support frame portion for fixing a workpiece; a welding torch for welding the workpiece; a moving guide portion for moving the support frame portion or the welding torch; A welding system comprising a control module for controlling a moving guide unit based on data received from a sensor; wherein the control module comprises the steps of: obtaining from the sensor a first error group including an x-axis position error and a rotation angle error of a workpiece, and a second error group including deviations of each of a plurality of pre-set critical points on the workpiece; calculating a first intermediate evaluation value by applying a mutual multiplication relationship in which the influence of the other error is weighted when either the x-axis position error or the rotation angle error increases; calculating a second intermediate evaluation value by applying an exponential relationship in which the risk level increases exponentially as the plurality of deviations included in the second error group accumulate; calculating a final alignment index by multiplying the first intermediate evaluation value and the second intermediate evaluation value, and driving the moving guide unit to automatically correct the alignment of the workpiece when the final alignment index exceeds a pre-set threshold value. Claim 2 A welding system comprising a control module for workpiece alignment correction according to claim 1, wherein the control module calculates the first intermediate evaluation value by additionally applying a logarithmic relationship to the error value so that the value changes sensitively even when the x-axis position error and the rotation angle error change within a small range, and the threshold value is variably set according to the results of a statistical analysis of past welding defect data or the quality standards of the workpiece.

Images