A dynamic balance calibration welding machine system

Abstract

The invention discloses a dynamic balance calibrated welding machine system, which relates to the technical field of welding, and comprises a global parameter module, a welding machine monitoring module, a power supply balance module, a pressurization balance module, an overheat protection module and a welding module; the global parameter module automatically generates a global instruction based on a trajectory planning algorithm and a workpiece matching strategy analysis of a three-dimensional model and workpiece information; the welding machine monitoring module collects real-time welding data; the power supply balance module uses an intelligent reactive power compensation method to adjust the three-phase compensation admittance in real time to achieve three-phase power supply balance, and uses a fuzzy control algorithm to adjust the transformer to ensure a constant welding voltage; the pressurization balance module autonomously decides whether to determine the electrode working power based on an electrode working equation, and performs servo pressurization to maintain welding pressure balance; the overheat protection module cuts off power supply when the welding machine is overheated; the welding module controls the welding machine to perform welding in a specified manner at a corresponding welding point, which effectively improves the automation and balance of the welding of the welding machine.

Description

Technical Field

[0001] The invention relates to the technical field of welding, in particular to a dynamic balance calibrated welding machine system. Background Art

[0002] Traditional welding is a process of manual control, which requires high proficiency and experience of welding technicians. With the development of automatic control technology, automatic welding machines have gradually become popular. However, due to the poor feedback control ability of automatic welding machines for welding status, existing automatic welding machines can only be applied to some simple welding conditions.

[0003] The existing Chinese patent with announcement number CN109967829B discloses a welding machine control system and a control method thereof, including a welding current control module for controlling the welding current; an arc voltage control module for controlling the arc voltage; a welding speed control module for controlling the welding speed; a welding angle control module for controlling the welding angle; a shielding gas flow control module for controlling the shielding gas flow; a database module for storing tabular data of weld states and welding parameter adjustment methods; a weld image shooting module for shooting weld images; and a weld image analysis module for analyzing the shot weld images, thereby improving the control accuracy of the welding machine in complex working conditions.

[0004] However, the existing technology analyzes the weld image to retrieve the corresponding welding parameters in the database for control. The control method has a delay and cannot provide timely control for emergencies during the welding process. In addition, the existing technology does not fully consider problems such as pressure imbalance and three-phase imbalance of the spot welding machine, which brings safety hazards to the operation of the spot welding machine. Summary of the invention

[0005] In view of the deficiencies in the prior art, the present invention proposes a dynamically balanced and calibrated welding machine system, which can dynamically adjust and control during the welding process to achieve three-phase power supply balance, welding pressure balance and welding voltage balance.

[0006] The technical solution to achieve the purpose of the present invention is:

[0007] A dynamic balance calibration welding machine system is applied to a rotary table welding machine device, comprising a global parameter module, a welding machine monitoring module, a power supply balance module, a pressurization balance module, an overheat protection module and a welding module;

[0008] The global parameter module obtains the workpiece information of the workpiece to be welded, generates global welding parameters based on the trajectory planning algorithm and the workpiece matching strategy, integrates them into global instructions and sends them to the welding module;

[0009] The welding machine monitoring module collects the real-time welding pressure, real-time welding voltage and real-time welding temperature at each welding point, and feeds them back to the power supply balance module, pressure balance module and overheat protection module respectively;

[0010] The power supply balance module uses intelligent reactive power compensation method to achieve dynamic three-phase power balance and maintains constant welding voltage based on fuzzy control algorithm;

[0011] The pressurization balance module decides whether to use servo pressurization to meet the welding pressure based on the real-time welding pressure;

[0012] The overheat protection module decides whether to urgently cut off the three-phase power supply based on the real-time welding temperature;

[0013] The welding module receives and parses the global instructions and and pressurization direction sequence Control welding The welding points are welded in the specified manner.

[0014] Further, the global parameter module includes a three-dimensional modeling unit, a workpiece measurement unit, an analysis and matching unit, and a global instruction unit;

[0015] The three-dimensional modeling unit establishes a reference space coordinate system, scans and obtains the three-dimensional model of the workpiece to be welded, and transmits it to the analysis and matching unit;

[0016] The workpiece measurement unit captures the appearance image of the workpiece to be welded , generate pulse current on the surface of the workpiece to be welded, and measure the temperature of the workpiece to be welded when the pulse current is generated ,Voltage and current , further calculate and obtain the resistance of the workpiece to be welded , the appearance image of the workpiece to be welded ,resistance and temperature Transmit to the analysis and matching unit;

[0017] The analysis and matching unit imports the three-dimensional model of the workpiece to be welded, and generates global welding parameters based on the trajectory planning algorithm and workpiece matching strategy in the reference space coordinate system;

[0018] The global instruction unit receives global welding parameters, integrates and generates global instructions, and sends them to the welding execution module.

[0019] Furthermore, the trajectory planning algorithm used in the analysis matching unit is used to generate the solder joint position sequence and pressurization direction sequence , including the following specific steps:

[0020] Acquire a three-dimensional model of the workpiece to be welded, and mark a common welding edge of the first sub-portion and the second sub-portion of the three-dimensional model to confirm a welding line;

[0021] Select one end of the welding line as the starting point, and number the triangles that the welding line passes through. Triangles , is the total number of face numbers;

[0022] Connect the welding wire to the Triangles Coordinates of two intersection points and The midpoint coordinates of Triangles Corresponding solder joint position ;

[0023] Get it in sequence The solder point positions corresponding to the triangular facets are used to construct a solder point position sequence ;

[0024] Get the Triangles The normal vector , calculate the normal vector The spatial angle with the reference space coordinate system As the Triangles Corresponding pressurization direction;

[0025] Get it in sequence The pressure direction corresponding to the triangles is used to construct a pressure direction sequence .

[0026] Furthermore, the workpiece matching strategy adopted in the analysis matching unit is used to confirm the single welding time, single welding pressure and initial welding voltage, including the following specific steps:

[0027] Determine the average thickness of the workpiece to be welded based on the 3D model of the workpiece to be welded ;

[0028] Get the appearance image of the workpiece to be welded ,resistance and temperature , the material of the workpiece to be welded is confirmed by using multi-modal feature joint matching network analysis;

[0029] Query the internal database, combined with the material and average thickness of the workpiece to be welded Get the single welding time according to the thickness range , Single welding pressure and single welding voltage .

[0030] Furthermore, the multi-modal feature joint matching network confirms the material of the workpiece to be welded, including the following specific steps:

[0031] The appearance image of the workpiece to be welded The pixel values ​​of the three channels are unified into a single channel grayscale value through weighted summation, making the appearance image Convert to grayscale image ;

[0032] The first convolutional layer and the second convolutional layer are used to respectively transform the grayscale image from a high-dimensional perspective. Perform global extraction and pixel-by-pixel extraction to generate the texture feature vector of the welded workpiece ;

[0033] The resistance of the workpiece to be welded is processed by a nonlinear mapping consisting of linear expansion and ReLU activation function. and temperature , generating and texture feature vector Resistance eigenvectors of the same dimension and the temperature eigenvector ;

[0034] Using the linear weights of the fully connected layer, weighted combination of texture feature vectors , resistance characteristic vector and the temperature eigenvector , generate multimodal feature vector ;

[0035] Use Softmax logistic regression classifier to reduce dimension and normalize multimodal feature vectors , generating the probability vector , and the probability vector The material with the highest matching probability is selected as the material of the workpiece to be welded.

[0036] Furthermore, the global instruction unit receives the global welding parameters and integrates and generates the global instruction, including the following specific steps:

[0037] Get the solder joint position sequence , pressurization direction sequence , Single welding time , Single welding pressure and single welding voltage ;

[0038] Initialize the sequence of solder joint positions Welding duration sequence with the same dimension and all values ​​0 , welding pressure sequence and welding voltage sequence ;

[0039] Single welding time , Single welding pressure and single welding voltage Replace welding duration sequence separately , welding pressure sequence and welding voltage sequence All 0s in

[0040] Corresponding welding point position sequence , pressurization direction sequence , welding time sequence , welding pressure sequence and welding voltage sequence , generate global instructions.

[0041] Furthermore, the power supply balancing module includes a three-phase balancing unit and a voltage-stabilizing power supply unit;

[0042] The three-phase balancing unit collects three-phase voltage and three-phase current in real time, and uses intelligent reactive power compensation method to achieve dynamic three-phase power balance;

[0043] The voltage-stabilizing power supply unit obtains the real-time welding voltage and adjusts the transformer in real time based on the fuzzy control algorithm to maintain a constant welding voltage.

[0044] Furthermore, the three-phase balancing unit adopts the intelligent reactive power compensation method to achieve dynamic three-phase balancing, including the following specific steps:

[0045] Real-time acquisition of three-phase voltage vectors , , and the three-phase line current vector , , , and the phase difference between any two phases of the three-phase voltage is 120 degrees, the three-phase voltage vector , and They can be expressed mutually, and the specific formula is as follows:

[0046] ,

[0047] ,

[0048] in, represents the phase offset vector, , represents an imaginary number, ;

[0049] Obtaining the three-phase load admittance vector based on the triangle load balancing formula , and , the triangle load balancing formula is as follows

[0050] ,

[0051] ,

[0052] ;

[0053] Based on the positive sequence current vector and negative sequence current vector Calculating Current Unbalance , where the positive sequence current vector and negative sequence current vector It can be obtained by deducing the three-phase load admittance. The specific formula is as follows

[0054] ,

[0055] ;

[0056] Determining the current imbalance Is it greater than the current unbalance reference value? If the current imbalance Less than or equal to , the three-phase power supply is in a three-phase load balance state. If the current is unbalanced Greater than , the three-phase power supply is in a three-phase load unbalanced state, and the symmetrical component method is used for load compensation to confirm the three-phase compensation admittance vector , and .

[0057] Furthermore, the three-phase compensation admittance vector is determined based on the symmetrical component method. , and The specific steps include:

[0058] After load compensation, the negative sequence current vector of the three-phase power is required to be becomes 0, and the imaginary part of the positive sequence current vector is becomes 0, where Indicates the operation of taking the imaginary part to confirm the compensation of the positive sequence current vector satisfy , compensate the negative sequence current vector ;

[0059] Derivation of three-phase compensation admittance vector based on triangle compensation balance formula , and The triangle compensation balance formula is as follows

[0060] ,

[0061] ;

[0062] Combined three-phase voltage vector , , The phase correlation between them can be obtained by sorting out the three-phase compensation admittance vector , and The calculation formula is as follows

[0063] ,

[0064] ,

[0065] .

[0066] Furthermore, the voltage stabilizing power supply unit adjusts the transformer in real time based on the fuzzy control algorithm, including the following specific steps:

[0067] Get the Real-time welding voltage of each welding point and single welding voltage , ;

[0068] The real-time welding voltage and single welding voltage The difference is taken as the real-time voltage error , the real-time voltage error Single welding time The ratio of the real-time error change rate ;

[0069] The real-time voltage error and real-time error change rate Input fuzzy neural network respectively, fuzzy neural network output and real-time voltage error and real-time error change rate The three best matching PID parameters are: Optimal ratio parameters for each solder joint , optimal integration parameters and the optimal differential parameters ;

[0070] PID combined with optimal proportional parameters , optimal integration parameters and the optimal differential parameters Calculate the proportion items in sequence , integral term and the differential term , and output the control parameters of the transformer .

[0071] Furthermore, the fuzzy neural network includes an input layer, a membership function layer, a rule layer, and a defuzzification layer;

[0072] The input layer obtains real-time voltage error and real-time error change rate , respectively, with the error quantization factor and rate of change quantification factor Multiply to get the first input and the second input ;

[0073] The membership function layer is used to characterize the first input and the second input Regarding the membership of the seven fuzzy sets, output the first membership vector and the second membership vector , among which, Membership vector The Membership The specific calculation formula is as follows

[0074] ,

[0075] in, and Respectively enter With The mean and standard deviation of the fuzzy sets are learnable parameters, , ;

[0076] The rule layer converts the first membership vector and the second membership vector The membership matrix is ​​obtained by transposing and multiplying ;

[0077] The defuzzification layer weights the membership matrix by connecting the weight coefficients To generate the optimal PID parameter vector ,in, , and They are the optimal proportional parameter, optimal integral parameter and optimal differential parameter respectively.

[0078] Furthermore, the pre-training of the fuzzy neural network includes the following specific steps:

[0079] Defining indicator functions Indicates the welding voltage obtained by PID adjustment and the single welding voltage The error, Fuzzy neural network During round training, the indicator function is used The calculated error;

[0080] The hyperparameters involved in the fuzzy neural network are uniformly defined as , No. Hyperparameters for training rounds Update based on the Delta rule. The specific formula is as follows

[0081] ,

[0082] in, Represents hyperparameters In the The amount of hyperparameter updates in a training round, For the Hyperparameters during rounds of training;

[0083] No. The amount of hyperparameter updates in a training round Combined with learning efficiency and momentum factor Determine, the specific calculation formula is as follows

[0084] ,

[0085] in, Represents hyperparameters In the The amount of hyperparameter updates in a training round.

[0086] Furthermore, the pressurization balance module determines whether to use servo pressurization based on the real-time welding pressure, including the following specific steps:

[0087] Get the Real-time welding pressure of each welding point , and calculate the deviation of the pressurization direction , ;

[0088] Determine the deviation of the pressurization direction Is it greater than the standard deviation? , if the pressure direction deviates Less than or equal to the standard deviation , it is considered as acceptable deviation error;

[0089] If the pressure direction deviates Greater than the standard deviation , calculate the real-time pressurization pressure ;

[0090] The real-time pressure Confirming the electrode working power based on the electrode working equation The electrode working equation is as follows:

[0091] ,

[0092] in, is the electrode acceleration, is the electrode conversion efficiency.

[0093] Furthermore, the overheat protection module obtains the real-time welding temperature and compares it with the welding machine warning temperature. If it is greater than the welding machine warning temperature, the three-phase power supply is immediately cut off. If it is less than or equal to the welding machine warning temperature, no processing is performed.

[0094] Compared with the prior art, the present invention has the following significant advantages:

[0095] 1. Design a global parameter module, use trajectory planning algorithm to analyze the three-dimensional model of the workpiece to be welded, accurately generate the welding point trajectory and pressure direction, use the workpiece matching strategy to analyze and confirm the material of the workpiece to be welded, and query the internal database in combination with the thickness to obtain the appropriate single welding time, single welding pressure and single welding voltage, thus realizing the automatic setting of welding parameters without manual work;

[0096] 2. Design a power supply balance module. On the one hand, it analyzes the three-phase voltage and three-phase current in real time, and adjusts them in real time through intelligent reactive power compensation method to ensure the stability of the three-phase power supply. On the other hand, it obtains the real-time welding voltage, compares and analyzes it with the single welding voltage, and uses fuzzy control algorithm to control the transformer to ensure that the welding voltage is always stable at the single welding voltage to achieve constant voltage control;

[0097] 3. Design a pressurization balance module to obtain real-time welding pressure and independently decide whether to determine the electrode working power through the electrode working equation to perform servo pressurization to ensure the dynamic balance of welding pressure. BRIEF DESCRIPTION OF THE DRAWINGS

[0098] Figure 1 A schematic diagram of a welding machine used in the present invention;

[0099] Figure 2 for Figure 1 A partial enlarged view of part A;

[0100] Figure 3 The schematic diagram of the structure of the turntable for the welding machine;

[0101] Figure 4 for Figure 3A partial enlarged view of part B in FIG.

[0102] Figure 5 To reflect the structural diagram of the welding station;

[0103] Figure 6 for Figure 5 A partial enlarged view of part C in FIG.

[0104] Figure 7 A dynamically balanced and calibrated welding machine system;

[0105] Figure 8 This is a multi-modal feature joint matching network model diagram in the present invention;

[0106] Fig. 9 It is the flow chart of the fuzzy control algorithm in the present invention;

[0107] Fig.10 This is a fuzzy neural network model diagram in the present invention.

[0108] Reference numerals

[0109] 1. Machine body; 2. Turntable; 3. Loading station; 31. Loading cylinder; 32. Guide wheel; 4. Welding station; 41. Truss; 42. Wire clamp; 43. Bottom mold; 44. Groove; 6. Positioning tooling; 61. Damper; 62. Positioning splint; 7. Pressing mechanism; 71. Pressing cylinder; 72. Fixed mold; 73. Moving mold; 74. Mold frame; 75. Reset damping; 8. Fixed mold mechanism; 81. Push wedge block; 82. Push cylinder; 9. Feeding mechanism; 91. Vibrating screen; 92. Feeding channel. DETAILED DESCRIPTION

[0110] The present invention is further described in detail below in conjunction with the accompanying drawings and embodiments.

[0111] Example 1

[0112] like Figures 1 to 6As shown, a specific embodiment of the present invention discloses a dynamic balance calibration welding machine system, which is applied to a turntable welding machine device. The turntable welding machine device includes a body 1, a turntable 2, a loading station 3 and a welding station 4. The turntable 2 is arranged on the body 1 and is connected to a driver. The driver is used to drive the turntable 2 to rotate. A plurality of positioning fixtures 6 are arranged on the turntable 2 along the circumferential direction. The positioning fixtures 6 are used to clamp the parts loaded through the loading station 3. The loading station 3 includes a loading cylinder 31 and a guide wheel 32. The positioning fixture 6 includes a damper 61 and a positioning clamping plate 62. A positioning groove for installing the positioning clamping plate 62 is opened on the turntable 2. One side of the positioning clamping plate 62 is located on the turntable 2 toward the body 1. The guide wheel 32 is located on one side of the turntable 2 away from the machine body 1, and the other side passes through the positioning groove to the side of the turntable 2 away from the machine body 1. The damper 61 is arranged between the positioning clamping plate 62 and the turntable 2, and is located on the side of the turntable 2 facing the machine body 1. When the turntable 2 is driven by the driver to drive the positioning tooling 6 to rotate to the loading station 3, the loading cylinder 31 pushes the guide wheel 32 to contact the positioning clamping plate 62, and pushes the positioning clamping plate 62 to move, so as to drive the damper 61 to compress and deform to accumulate force. At this time, the positioning clamping plate 62 is located on the side of the turntable 2 away from the machine body 1 and the gap between the turntable 2 and the turntable 2 is increased to provide space for placing parts for welding. When the loading cylinder 31 drives the guide wheel 32 to reset, the damper 61 is restored and drives the positioning clamping plate 62 to clamp the parts and the turntable 2.

[0113] like Figure 1 , Figure 2 , Figure 5 and Figure 6 As shown, the welding station 4 includes a truss 41, a welding wire clamp 42, a pressing mechanism 7, a bottom mold 43 and a fixed mold mechanism 8. The truss 41 is erected on the machine body 1. The pressing mechanism 7 is arranged on the truss 41 and connected to the welding wire clamp 42. The welding wire clamp 42 is used to clamp the welding wire. The pressing mechanism 7 is used to control the downward movement of the welding wire, and to form a welding point after the welding wire contacts the welding position of the part. The fixed mold mechanism 8 is arranged on the machine body 1 and is located at the bottom of the turntable 2. The fixed mold mechanism 8 is used to push the bottom mold 43 to move. The bottom mold 43 is provided with a part groove 44, so that when the positioning tool 6 holding the part is rotated to the welding station 4, the fixed mold mechanism 8 pushes the bottom die 43 to move toward the turntable 2, and contacts with one side of the damper 61 connected to the positioning clamp 62, so as to push the positioning clamp 62 to release the clamping of the part. At this time, the position of the part is located in the groove 44 in the bottom die 43, and the bottom die 43 forms a positioning for the part. A feeding mechanism 9 is also arranged on the machine body 1. The feeding mechanism 9 includes a vibrating screen 91 and a feeding channel 92. The vibrating screen 91 is used to transport the parts to the feeding channel 92 after being separated by the vibrating screen 91, so that the parts to be welded can move to the bottom die 43 through the feeding channel 92, and the pressing mechanism 7 welds and fixes the parts to be welded and the parts located in the bottom die 43 by pressing the welding wire.

[0114] like Figure 5 and Figure 6 As shown, the fixed mold mechanism 8 includes a pushing wedge block 81 and a pushing cylinder 82. The pushing wedge block 81 is connected to the end of the piston rod of the pushing cylinder 82. The pushing wedge block 81 pushes against the end of the bottom mold 43 away from the turntable 2, so that when the pushing cylinder 82 pushes the pushing wedge block 81 to move, the acting surfaces of the bottom mold 43 and the pushing wedge block 81 are wedge-shaped surfaces, thereby driving the bottom mold 43 to move up and down.

[0115] like Figure 2 , Figure 5 and Figure 6 As shown, the pressing mechanism 7 includes a pressing cylinder 71, a fixed die 72, a movable die 73 and a die frame 74, the fixed die 72 is connected to the truss 41, the pressing cylinder 71 is arranged on the side of the truss 41 that is away from the state, the movable die 73 is slidably connected to the fixed die 72, the piston rod end of the pressing cylinder 71 extends into the truss 41 and is connected to the movable die 73, the side of the movable die 73 away from the pressing cylinder 71 is connected to the die frame 74, the welding wire clamp 42 is fixedly connected to the die frame 74, and a reset damper 75 is also arranged between the movable die 73 and the fixed die 72, so that when the pressing cylinder 71 pushes the movable die 73 to move, the welding wire clamp 42 holding the welding wire is pressed down and moved, and at this time the reset damper 75 is deformed and stored, so that when the force of the pressing cylinder 71 is released, the reset damper 75 drives the movable die 73 to reset, thereby driving the welding wire clamp 42 to reset.

[0116] like Figures 7 to 10 As shown, the dynamic balance calibration welding machine system includes a global parameter module, a welding machine monitoring module, a power supply balance module, a pressurization balance module, an overheat protection module and a welding module;

[0117] The global parameter module obtains the workpiece information of the workpiece to be welded, generates global welding parameters based on the trajectory planning algorithm and the workpiece matching strategy, integrates them into global instructions and sends them to the welding module. The workpiece information includes the three-dimensional model and resistance of the workpiece to be welded;

[0118] The welding machine monitoring module collects real-time welding pressure, real-time welding voltage and real-time welding temperature at each welding point through multiple built-in sensors or intelligent detection equipment, and feeds them back to the power supply balance module, pressure balance module and overheat protection module respectively;

[0119] The power supply balance module uses intelligent reactive power compensation method to adjust the three-phase power to achieve dynamic three-phase power balance, and adjusts the transformer in real time based on fuzzy control algorithm to maintain constant welding voltage;

[0120] The pressurization balance module decides whether to use servo pressurization to pressurize in real time to meet the welding pressure based on the real-time welding pressure;

[0121] The overheat protection module decides whether to urgently cut off the three-phase power supply based on the real-time welding temperature;

[0122] The welding module receives and parses the global instructions and and pressurization direction sequence Control welding The welding points are welded in a specified manner, and the specified manner specifically refers to the welding gun at the Solder point locations In the direction of pressure With single welding pressure Welding is performed, and the welding time is the single welding time .

[0123] Further, the global parameter module includes a three-dimensional modeling unit, a workpiece measurement unit, an analysis and matching unit, and a global instruction unit;

[0124] The three-dimensional modeling unit establishes a reference space coordinate system, uses a laser scanning device to scan and obtain a three-dimensional model of the workpiece to be welded fixed on the platform and transmits it to the analysis and matching unit;

[0125] The workpiece measurement unit uses high-definition camera equipment to capture the appearance image of the workpiece to be welded. , and generate pulse current on the surface of the workpiece to be welded, and use intelligent measuring equipment to obtain the temperature of the workpiece to be welded when the pulse current is generated ,Voltage and current , based on voltage and current Calculate the resistance of the workpiece to be welded , the appearance image of the workpiece to be welded ,resistance and temperature Transmit to the analysis and matching unit;

[0126] The analysis and matching unit imports the 3D model of the workpiece to be welded into the 3D software, and generates global welding parameters based on the trajectory planning algorithm and workpiece matching strategy in the reference space coordinate system. The global welding parameters include the sequence of welding point positions. , pressurization direction sequence , Single welding time , Single welding pressure and single welding voltage ;

[0127] The global instruction unit receives global welding parameters, integrates and generates global instructions, and sends them to the welding execution module.

[0128] Furthermore, the trajectory planning algorithm used in the analysis matching unit is used to generate the solder joint position sequence and pressurization direction sequence , including the following specific steps:

[0129] Acquire a three-dimensional model of the workpiece to be welded, wherein the three-dimensional model of the workpiece to be welded includes a first sub-portion and a second sub-portion, and mark common welding edges of the first sub-portion and the second sub-portion respectively to confirm a welding line;

[0130] Select one end of the welding line as the starting point, and number the triangles that the welding line passes through. Triangles , is the total number of face numbers;

[0131] Get welding line and Triangles The coordinates of the two intersection points and , the intersection coordinates and The midpoint coordinates of Triangles Corresponding solder joint position It is worth noting that since the welding line is continuous, the triangle patch The intersection coordinates That is the Triangles The intersection coordinates , and the triangle patch The coordinates of another intersection point That is the Triangles The intersection coordinates ;

[0132] Get it in sequence The solder point positions corresponding to the triangular facets are used to construct a solder point position sequence ;

[0133] Get the Triangles The normal vector , calculate the normal vector The spatial angle with the reference space coordinate system As the Triangles Corresponding pressurization direction;

[0134] Get it in sequence The pressure direction corresponding to the triangles is used to construct a pressure direction sequence .

[0135] Furthermore, the workpiece matching strategy adopted in the analysis matching unit is used to confirm the single welding time, single welding pressure and initial welding voltage, including the following specific steps:

[0136] Determine the average thickness of the workpiece to be welded based on the 3D model of the workpiece to be welded ;

[0137] Get the appearance image of the workpiece to be welded ,resistance and temperature , using a multimodal feature joint matching network to jointly analyze the appearance image ,resistance and temperature To confirm the material of the workpiece to be welded;

[0138] Query the internal database, combined with the material and average thickness of the workpiece to be welded Get the single welding time in the thickness range , Single welding pressure and single welding voltage The internal database is pre-established to store the single welding time, single welding pressure and single welding voltage corresponding to various materials of welded workpieces within various thickness ranges.

[0139] like Figure 2 As shown, further, the multi-modal feature joint matching network confirms the material of the workpiece to be welded, including the following specific steps:

[0140] The appearance image of the workpiece to be welded The pixel values ​​of the three channels of R, G, and B are unified into a single channel grayscale value through weighted summation, making the appearance image Convert to grayscale image , where the weights assigned to the three channels R, G, and B are 0.299, 0.587, and 0.114, respectively, which simulates the sensitivity of the human eye to red, green, and blue;

[0141] The first convolutional layer and the second convolutional layer are used to extract grayscale images from a high-dimensional perspective. , generate the texture feature vector of the welded workpiece , where the first convolutional layer uses The large convolution kernel is used for global extraction with a stride of 3, aiming to filter out the grayscale image. The interference noise in the second convolution layer is The small convolution kernel is extracted pixel by pixel with a step size of 1 to perceive the grayscale image Pixel changes in ;

[0142] The resistance of the workpiece to be welded is processed by a nonlinear mapping consisting of linear expansion and ReLU activation function. and temperature , generating and texture feature vector Resistance eigenvectors of the same dimension and the temperature eigenvector ;

[0143] Using the linear weights of the fully connected layer, weighted combination of texture feature vectors , resistance characteristic vector and the temperature eigenvector , generate multimodal feature vector ,in, , and The texture feature vectors are , resistance characteristic vector and the temperature eigenvector The corresponding linear weights;

[0144] Use Softmax logistic regression classifier to process multimodal feature vectors , the multimodal feature vector The dimensionality is reduced to a specific dimension through linear processing, and then normalized by Softmax (i.e., normalized exponential function) to generate a probability vector , the probability vector is the matching probability between the workpiece to be welded and various materials in the internal database, and the probability vector The material with the highest matching probability is used as the material of the workpiece to be welded, wherein the specific dimension specifically refers to the total number of materials stored in the internal database.

[0145] Furthermore, the global instruction unit receives the global welding parameters and integrates and generates the global instruction, including the following specific steps:

[0146] Get the solder joint position sequence , pressurization direction sequence , Single welding time , Single welding pressure and single welding voltage ;

[0147] Initialize the sequence of solder joint positions Welding duration sequence with the same dimension and all values ​​0 , welding pressure sequence and welding voltage sequence ;

[0148] Single welding time , Single welding pressure and single welding voltage Replace welding duration sequence separately , welding pressure sequence and welding voltage sequence All 0s in the welding time sequence , welding pressure sequence , welding voltage sequence ;

[0149] Corresponding welding point position sequence , pressurization direction sequence , welding time sequence , welding pressure sequence and welding voltage sequence ,generate Dimensional global directive.

[0150] Furthermore, the power supply balancing module includes a three-phase balancing unit and a voltage-stabilizing power supply unit;

[0151] The three-phase balancing unit is connected to the three-phase power supply, collects the three-phase voltage and three-phase current in real time, and uses the intelligent reactive power compensation method to achieve dynamic three-phase power supply balance to ensure the safety and stability of the three-phase power supply;

[0152] The voltage-stabilized power supply unit obtains the real-time welding voltage and adjusts the transformer in real time based on the fuzzy control algorithm to maintain the welding voltage at a single welding voltage. .

[0153] Furthermore, the three-phase balancing unit adopts the intelligent reactive power compensation method to achieve dynamic three-phase balancing, including the following specific steps:

[0154] Real-time acquisition of three-phase voltage vectors through power measurement equipment , , and the three-phase line current vector , , It is worth noting that the phase difference between any two phases of the three-phase voltage is 120 degrees, so the three-phase voltage vector , and They can be expressed mutually, and the specific formula is as follows:

[0155] ,

[0156] ,

[0157] in, represents the phase offset vector, , represents an imaginary number, ;

[0158] Based on the triangle load balancing formula combined with the three-phase voltage vector , , and the three-phase line current vector , and Get the three-phase load admittance vector , and , the triangle load balancing formula is as follows

[0159] ,

[0160] ,

[0161] ;

[0162] Based on the positive sequence current vector and negative sequence current vector Calculating Current Unbalance , where the positive sequence current vector and negative sequence current vector It is derived based on the three-phase load admittance. The specific formula is as follows

[0163] ,

[0164] ;

[0165] Determining the current imbalance Is it greater than the current unbalance reference value? If the current imbalance Less than or equal to , it is considered that the current three-phase power is in a three-phase load balance state. If the current imbalance degree Greater than , it is considered that the current three-phase power is in a three-phase load unbalanced state, and the symmetrical component method is used for load compensation to confirm the three-phase compensation admittance vector required for load compensation. , and .

[0166] Furthermore, the three-phase compensation admittance vector is determined based on the symmetrical component method. , and The specific steps include:

[0167] Since the symmetrical component method requires that the negative sequence current vector of the three-phase power after load compensation is becomes 0, and the imaginary part of the positive sequence current vector is becomes 0, where Indicates the operation of taking the imaginary part, therefore, the compensated positive sequence current vector in load compensation is satisfy , compensate the negative sequence current vector ;

[0168] Based on the triangle compensation balance formula combined with the compensation positive sequence current vector and compensated negative sequence current vector , the three-phase compensation admittance vector can be derived , and The triangle compensation balance formula is as follows

[0169] ,

[0170] ;

[0171] Combined three-phase voltage vector , , The phase correlation between them can be obtained by sorting out the three-phase compensation admittance vector , and The calculation formula is as follows

[0172] ,

[0173] ,

[0174] .

[0175] like Figure 3 As shown, further, the voltage stabilizing power supply unit adjusts the transformer in real time based on the fuzzy control algorithm, including the following specific steps:

[0176] Get the Real-time welding voltage of each welding point and single welding voltage , ;

[0177] Calculate and obtain real-time voltage error and real-time error change rate , real-time voltage error Real-time welding voltage and single welding voltage The difference, that is , real-time error change rate The real-time voltage error Single welding time The ratio of ;

[0178] The real-time voltage error and real-time error change rate Input the fuzzy neural network respectively, and the fuzzy neural network corrects the value of the adjustable parameter in real time to obtain the real-time voltage error and real-time error change rate The three best matching PID parameters are: Optimal ratio parameters for each solder joint , optimal integration parameters and the optimal differential parameters ;

[0179] PID combined with optimal proportional parameters , optimal integration parameters and the optimal differential parameters Calculate the proportion items in sequence , integral term and the differential term It is worth noting that PID uses proportional terms and the integral term Make a single large adjustment and multiple small adjustments to reduce real-time voltage errors , while using the differential term , control the pace of small adjustments to prevent over-adjustment, and finally output the control parameters of the transformer The control parameter is essentially the transformer turns ratio, which is used to steadily step down the three-phase voltage to the required single welding voltage. .

[0180] like Figure 4 As shown, further, the fuzzy neural network includes an input layer, a membership function layer, a rule layer and a defuzzification layer;

[0181] The input layer obtains real-time voltage error and real-time error change rate , and respectively with the error quantization factor and rate of change quantification factor Multiply to get the first input and the second input ;

[0182] The membership function layer has 14 nodes, which are used to represent the first input and the second input Regarding the membership of the seven fuzzy sets, the final output is the first membership vector and the second membership vector , among which, Membership vector The Membership The specific calculation formula is as follows

[0183] ,

[0184] in, and are the first enter With The mean and standard deviation of the fuzzy sets are learnable parameters, , ;

[0185] The rule layer has 49 nodes in total. and the second membership vector Multiply the transpose of to get the dimension The membership matrix ;

[0186] The defuzzification layer weights the membership matrix by connecting the weight coefficients To generate the optimal PID parameter vector ,in, and are the linear weight matrix and linear weight vector that can be learned in the defuzzification layer, and their dimensions are and , , and They are the optimal proportional parameter, optimal integral parameter and optimal differential parameter respectively.

[0187] Furthermore, the pre-training of the fuzzy neural network includes the following specific steps:

[0188] Defining indicator functions Indicates the welding voltage and single welding voltage obtained by PID based on the optimal PID parameter vector adjustment The error of Fuzzy neural network During round training, the indicator function is used The calculated error;

[0189] The hyperparameters involved in fuzzy neural networks are uniformly defined as , hyperparameters Including the membership function layer enter With The mean of fuzzy sets and standard deviation and the learnable linear weight matrix in the defuzzification layer and the linear weight vector ;

[0190] Fuzzy Neural Network Hyperparameters for training rounds Update based on the Delta rule. The specific formula is as follows

[0191] ,

[0192] in, Represents hyperparameters In the The amount of hyperparameter updates in a training round, Fuzzy neural network Hyperparameters during rounds of training;

[0193] No. The amount of hyperparameter updates in a training round In the indicator function About Hyperparameters Based on the gradient, the learning efficiency is combined and momentum factor Determine, the specific calculation formula is as follows

[0194] ,

[0195] in, Represents hyperparameters In the The hyperparameter update amount in the round of training is adaptively and dynamically adjusted, which reduces the oscillation trend during pre-training and improves the convergence speed of the fuzzy neural network.

[0196] Furthermore, the pressurization balance module determines whether to use servo pressurization based on the real-time welding pressure, including the following specific steps:

[0197] Get the Real-time welding pressure of each welding point , based on real-time welding pressure Get the deviation of the pressurization direction , that is, the welding machine is The welding points are along the pressure direction With single welding pressure When welding, the real-time pressure direction is in the pressure direction Based on deviation , resulting in real-time welding pressure Less than single welding pressure , ;

[0198] Determine the deviation of the pressurization direction Is it greater than the standard deviation? , if the pressure direction deviates Less than or equal to the standard deviation , it is considered as acceptable deviation error;

[0199] If the pressure direction deviates Greater than the standard deviation , then the single welding pressure With real-time welding pressure The difference is taken as the real-time pressurization pressure ;

[0200] The real-time pressure Confirming the electrode working power based on the electrode working equation The electrode working equation is as follows:

[0201] ,

[0202] in, is the electrode acceleration, is the electrode conversion efficiency. When the electrode pressure device is determined, the electrode acceleration and electrode conversion efficiency are all known constants, and are calculated by using the electrode working power By starting the electrode pressurizing device, insufficient welding pressure can be supplemented to achieve servo pressurization.

[0203] Furthermore, the overheat protection module obtains the real-time welding temperature and compares it with the welding machine warning temperature. If it is greater than the welding machine warning temperature, the three-phase power supply is immediately cut off. If it is less than or equal to the welding machine warning temperature, no processing is performed.

[0204] The invention discloses a welding machine system with dynamic balance calibration, comprising a global parameter module, a welding machine monitoring module, a power supply balance module, a pressurization balance module, an overheat protection module and a welding module; the global parameter module analyzes a three-dimensional model and workpiece information based on a trajectory planning algorithm and a workpiece matching strategy, and automatically generates global welding parameters; the welding machine monitoring module collects real-time welding pressure, real-time welding voltage and real-time welding temperature; the power supply balance module adopts an intelligent reactive power compensation method to analyze three-phase voltage and three-phase current in real time, and realizes three-phase power supply balance by regulating the three-phase compensation admittance to ensure power supply safety and stability, and adopts a fuzzy control algorithm to regulate the transformer to ensure constant welding voltage; the pressurization balance module independently decides whether to determine the electrode working power based on the electrode working equation based on the real-time welding pressure, and performs servo pressurization to maintain welding pressure balance; the overheat protection module cuts off power supply when the welding machine is overheated; the welding module controls the welding machine to weld in a specified manner at the corresponding welding point, effectively improving the automation and balance of the welding of the welding machine.

[0205] The above is only a preferred embodiment of the present invention, and the protection scope of the present invention is not limited to the above embodiments. All technical solutions under the concept of the present invention belong to the protection scope of the present invention. It should be pointed out that for ordinary technicians in this technical field, some improvements and modifications without departing from the principle of the present invention should also be regarded as the protection scope of the present invention.

Claims

1. A dynamic balance calibration welding machine system, applied to a rotary table welding machine, characterized in that: Including global parameter module, welding machine monitoring module, power supply balance module, pressure balance module, overheat protection module and welding module; The global parameter module analyzes the workpiece information based on the trajectory planning algorithm and the workpiece matching strategy to generate global welding parameters, and integrates them into global instructions and sends them to the welding module; The welding machine monitoring module collects real-time welding pressure, real-time welding voltage and real-time welding temperature at each welding point, and feeds them back to the power supply balance module, the pressure balance module and the overheat protection module respectively; The power supply balancing module adopts intelligent reactive power compensation method and fuzzy control algorithm to achieve dynamic three-phase power supply balance and constant welding voltage; The pressurization balancing module determines whether to use servo pressurization based on the real-time welding pressure; The overheat protection module decides whether to cut off the power supply urgently based on the real-time welding temperature; The welding module receives and parses the global instruction, and controls the welding to be performed in a specified manner at each welding point according to the welding point position sequence and the pressurization direction sequence; The global parameter module includes a three-dimensional modeling unit, a workpiece measurement unit, an analysis and matching unit, and a global instruction unit; The three-dimensional modeling unit establishes a reference space coordinate system, obtains a three-dimensional model of the workpiece to be welded, and transmits it to the analysis and matching unit; The workpiece measurement unit obtains the appearance image, resistance and temperature of the workpiece to be welded and transmits them to the analysis and matching unit; The analysis and matching unit imports a three-dimensional model of the workpiece to be welded, and generates global welding parameters based on a trajectory planning algorithm and a workpiece matching strategy; The global instruction unit receives global welding parameters, integrates and generates global instructions, and sends them to the welding execution module.

2. A dynamic balance calibration welding machine system as claimed in claim 1, characterized in that: The trajectory planning algorithm is used to generate a welding point position sequence and a pressurizing direction sequence, and includes the following specific steps: Obtain and mark the common welding edge of the three-dimensional model of the workpiece to be welded to confirm the welding line, and sequentially number the triangular facets through which the welding line passes; The midpoint coordinates of the two intersection points of the welding line and the single triangle facet are used as the welding point position corresponding to the single triangle facet; Sequentially obtain the welding point positions corresponding to the triangular facets that the welding line passes through, and construct a welding point position sequence; The spatial angle between the normal vector of the triangular face and the reference space coordinate system is taken as the pressurizing direction, and the pressurizing directions corresponding to the triangular facets through which the welding line passes are obtained in sequence to construct a pressurizing direction sequence.

3. A dynamic balance calibration welding machine system as claimed in claim 1, characterized in that: The workpiece matching strategy adopted in the analysis and matching unit is used to confirm the single welding time, single welding pressure and initial welding voltage, and includes the following specific steps: Determine the average thickness of the workpiece to be welded based on the three-dimensional model of the workpiece to be welded; Obtain the appearance image, resistance and temperature of the workpiece to be welded, and use a multi-modal feature joint matching network to confirm the material of the workpiece to be welded; The internal database is queried, and the single welding time, single welding pressure and single welding voltage are obtained in combination with the material of the workpiece to be welded and the thickness range of the average thickness.

4. A dynamic balance calibration welding machine system as claimed in claim 3, characterized in that: The multi-modal feature joint matching network confirms the material of the workpiece to be welded, including the following specific steps: The appearance image is converted into a single-channel grayscale image through weighted summation; The first convolution layer and the second convolution layer are used to perform global extraction and pixel-by-pixel extraction on the grayscale image from a high-dimensional perspective to generate a texture feature vector of the welded workpiece; The resistance and temperature of the workpiece to be welded are processed by nonlinear mapping to generate a resistance characteristic vector and a temperature characteristic vector; The texture feature vector, the resistance feature vector and the temperature feature vector are weightedly combined using a fully connected layer to generate a multimodal feature vector; The Softmax logistic regression classifier is used to reduce the dimension and normalize the multimodal feature vector to generate a probability vector, and the material with the highest matching probability in the probability vector is taken as the material of the workpiece to be welded.

5. A dynamic balance calibration welding machine system as claimed in claim 1, characterized in that: The power supply balancing module includes a three-phase balancing unit and a voltage-stabilizing power supply unit; The three-phase balancing unit collects three-phase voltage and three-phase current in real time, and uses intelligent reactive power compensation method to achieve dynamic three-phase power balance; The voltage-stabilizing power supply unit obtains the real-time welding voltage and adjusts the transformer in real time based on the fuzzy control algorithm to maintain a constant welding voltage.

6. A dynamic balance calibration welding machine system as claimed in claim 5, characterized in that: The three-phase balancing unit adopts the intelligent reactive power compensation method to realize dynamic three-phase balancing, which includes the following specific steps: The three-phase voltage vector and the three-phase line current vector are obtained in real time, and the three-phase voltage vector can be mutually represented based on the phase offset vector; Obtain the three-phase load admittance vector based on the triangle load balancing formula; Calculating the current unbalance degree based on the positive sequence current vector and the negative sequence current vector, wherein the positive sequence current vector and the negative sequence current vector can be obtained by deriving the three-phase load admittance; Determine whether the current imbalance is greater than a current imbalance reference value. When the current imbalance is greater than a reference value, confirm the compensation positive sequence current vector and the compensation negative sequence current based on the symmetrical component method. Based on the mutual representation between the triangle compensation balance formula and the three-phase voltage vectors, the three-phase compensation admittance vector is obtained.

7. A dynamic balance calibration welding machine system as claimed in claim 5, characterized in that: The voltage-stabilized power supply unit adjusts the transformer in real time based on the fuzzy control algorithm, including the following specific steps: Get the real-time welding voltage and single welding voltage of a single welding point; The difference between the real-time welding voltage and the single welding voltage is taken as the real-time voltage error, and the ratio of the real-time voltage error to the single welding time is taken as the real-time error change rate; The real-time voltage error and the real-time error change rate are input into the fuzzy neural network respectively, and the fuzzy neural network outputs three PID parameters that best match the real-time voltage error and the real-time error change rate; PID combines the three PID parameters to calculate the proportional term, integral term and differential term in sequence, and outputs the control parameters of the transformer.

8. A dynamic balance calibration welding machine system as claimed in claim 7, characterized in that: The fuzzy neural network includes an input layer, a membership function layer, a rule layer and an anti-fuzzification layer; The input layer multiplies the real-time voltage error and the real-time error change rate by the error quantization factor and the change rate quantization factor respectively to obtain a first input and a second input; The membership function layer is used to characterize the membership of the first input and the second input with respect to the fuzzy set, and outputs a first membership vector and a second membership vector; The rule layer multiplies the first membership vector and the transpose of the second membership vector to obtain a membership matrix; The defuzzification layer generates the optimal PID parameter vector by connecting the weight coefficients of the weighted reduced dimension membership matrix. The optimal PID parameter vector includes the three best matching PID parameters.

9. A dynamic balance calibration welding machine system as claimed in claim 1, characterized in that: The pressurization balancing module determines whether to use servo pressurization based on the real-time welding pressure, including the following specific steps: Obtain the real-time welding pressure of a single welding point and calculate the deviation of the pressurizing direction; Determine whether the deviation of the pressurization direction is greater than the standard deviation, and when the deviation of the pressurization direction is greater than the standard deviation, calculate the real-time pressurization pressure; The electrode working power is confirmed based on the electrode working equation combined with the real-time pressurization pressure to achieve servo pressurization.

Images