Variable gap GMAW welding penetration control method based on rough set knowledge reduction

Abstract

The application discloses a variable gap GMAW welding penetration control method based on rough set knowledge reduction. The control method adopts a visual sensor, a data acquisition card and other components to collect welding experimental data, under a rough set modeling mode, the collected experimental data is processed by a data preprocessing module to construct an original information decision table, and then is processed by an attribute reduction module and a rule reduction module to construct a rough set model. Under an online control mode, the collected experimental data is processed by the data preprocessing module and matched with the rough set model constructed under the rough set modeling mode, a fuzzy membership function is established by using a fuzzy set theory, a current change amount is output by fuzzy logic reasoning to perform penetration control. The application realizes rough modeling of variable gap welding and penetration control by using a fuzzy controller, has simple system composition, strong anti-interference capability and good engineering practicability, and can achieve the purpose of online penetration control under variable gap conditions.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of welding processing. Specifically, it relates to a variable gap GMAW welding penetration control method based on rough set knowledge reduction. BACKGROUND

[0002] GMAW welding, as a kind of efficient welding technology, has been widely used in manufacturing industry. In order to ensure the welding quality and avoid welding quality defects caused by assembly, welding thermal deformation and other problems, it is necessary to ensure that the welding current changes correspondingly after the change of the gap during welding, that is, to carry out real-time and effective welding penetration control. The key is to find out the corresponding relationship between the molten pool information and the penetration. The commonly used control methods at present mainly include welding process control based on sensor monitoring, welding process control based on neural network and welding process control based on fuzzy / PID.

[0003] Among them, the real-time acquisition of the penetration state information based on visual sensing is a key link of online penetration control. Since the penetration state information is difficult to measure directly, the penetration state of the back surface is usually predicted by detecting the front surface molten pool. The penetration state recognition and prediction can be realized by using artificial neural network, classical machine learning and deep learning algorithms. For the penetration state recognition and prediction model based on neural network and other machine learning algorithms, with the increase of input feature parameters, the prediction accuracy of the model is improved to a certain extent, but high-dimensional input inevitably brings redundant features and a large amount of redundant information. The existence of redundant features greatly increases the complexity of problem analysis, making it difficult to obtain the law of the change of the penetration state with the front surface feature of the molten pool through data analysis and processing, so as to establish the corresponding relationship between them

[0004] According to the disclosed technical solutions, the technical solution with publication number CN112507639A proposes a method for visualizing the dynamic process of GMAW welding droplet transfer, which couples the arc and droplet transfer for solving in Fluent software, obtains the instantaneous shape of droplet transfer and the distribution of electromagnetic parameters in the droplet, analyzes and predicts the influence law of the arc on the droplet transfer process, so as to implement accurate control of the discharge arc; the technical solution with publication number WO2013072742A1 sets a current control circuit with multiple on-off switches, so that the discharge current during welding can be accurately controlled; the technical solution with publication number EP3799991A1 proposes a welding system and welding method for realizing ultra-high deposition rate, which realizes the welding process with ultra-high deposition rate by using specially designed consumable flux-cored wire and controlling the temperature and arc current.

[0005] The above technical solutions all propose to collect a large amount of real-time data by using a special processing system, and then perform a large amount of operation to implement control of the GMAW welding process, which has the defects of great implementation difficulty, high requirement for the control system, and cannot be widely applied.

[0006] The foregoing discussion of the background of the application is merely intended to facilitate understanding of the application. It is not acknowledged or admitted that any of the information referred to in this discussion is part of the common general knowledge of the prior art. SUMMARY

[0007] The purpose of the present application is to provide a variable gap GMAW welding penetration control method based on rough set knowledge reduction. The control method uses a visual sensor, a data acquisition card and other components to collect welding experimental data. In the rough set modeling mode, the collected experimental data is processed by a data preprocessing module to construct an original information decision table, and then a rough set model is constructed through an attribute reduction module and a rule reduction module. In the online control mode, the collected experimental data is processed by the data preprocessing module and matched with the rough set model constructed in the rough set modeling mode. A fuzzy membership function is established using the fuzzy set theory, and the current change is output through fuzzy logic reasoning for penetration control. The present application realizes rough modeling of variable gap welding and penetration control through a fuzzy controller. The system has simple structure, strong anti-interference ability and good engineering practicability, and can achieve the purpose of online penetration control under variable gap conditions.

[0008] The present application adopts the following technical solutions:

[0009] A variable gap GMAW welding penetration control method based on rough set knowledge reduction, the control method comprising the following steps:

[0010] S100: Perform a plurality of welding penetration experiments; in the plurality of welding penetration experiments, select different gap parameters and welding machine current parameters for experiments, and collect a plurality of experimental data; the experimental data at least includes current, gap, penetration condition and molten pool size information; and a variable gap-current relationship penetration database is established using the above experimental data;

[0011] S200: Data preprocessing is performed on the penetration database to construct an original information decision system S; the original information decision system S is subjected to attribute reduction through three rough set attribute reduction algorithms, namely FRMAD method, FRBAC method and FRBAS method, to obtain reduced attributes, complete rule extraction and construct a rough set model;

[0012] S300: The obtained reduced attributes are subjected to fuzzy membership function construction through the minimum fuzziness method, and for continuous argument domain, Mamdani reasoning algorithm is used to complete fuzzy set model construction;

[0013] S400: In the actual welding process, Hall sensors are used to measure the welding arc signal, i.e., the welding arc current or welding arc voltage, and visual sensors are used to obtain information about the molten pool.

[0014] S500: The data collected in step S400 is connected to the control system through a data acquisition card, and the current change is output for fusion control through fuzzy logic reasoning.

[0015] The data preprocessing of the melt penetration database in step S200 includes the following steps:

[0016] S210: Initialize data cluster centers;

[0017] S220: Calculate the distance from each point to the cluster center and the membership function matrix;

[0018] S230: Cyclic selection of cluster centers;

[0019] S240: Recalculate the membership function matrix;

[0020] S250: Calculate the objective function. If the efficiency function reaches its minimum value, the data clustering is complete.

[0021] Preferably, the FRMAD method for rough set attribute reduction algorithms includes the following steps:

[0022] E100: Calculate the classification quality of decision attribute d with respect to condition attribute set C in the original decision information system S. Where β is the correct classification rate threshold;

[0023] E200: Sets the simplest reduced set of attributes, RED, and sets RED to an empty set, i.e., RED←φ;

[0024] E300: For each conditional attribute c j ∈C-RED, where j=1,2,……n; calculate the decision attribute d with respect to the condition attribute set RED∪{c j Classification quality

[0025] E400: Select to enable The attribute c with the maximum value j Add it to the set RED, i.e., RED←RED∪{c j};

[0026] E500: If If so, output RED; otherwise, return to step E300 and repeat the loop.

[0027] Among them, classification quality The definition is as follows:

[0028]

[0029] In the above formula, pos C (d) is the positive region of d in the approximation space C;

[0030] Preferably, the FRBAC method for rough set attribute reduction algorithm comprises the following steps:

[0031] F100: Calculate the classification quality of the decision attribute d about the condition attribute set C in the original decision information system S

[0032] F200: Calculate the classification quality of each condition attribute c j about the decision attribute d in the information system S j , and arrange c j in descending order according to sig(c j , C, d), and obtain the attribute core Core = Max(sig(c j , C, d));

[0033] F300: Add Core to the set RED, i.e. RED = Core, and calculate the classification quality of the decision attribute d about RED

[0034] F400: If , output RED; otherwise, jump to the following steps F500;

[0035] Then, in descending order according to sig(c j , C, d), repeat the following steps F500 and F600 for each c j ∈ C;

[0036] F500: Add c j to the set RED, i.e. RED = RED ∪ {c j}, and calculate the classification quality of the decision attribute d about RED

[0037] F600: If , output RED; otherwise, loop to step F400;

[0038] wherein the attribute importance sig(c j , C, d) is defined as:

[0039] sig(c j , C, d) = H({d} | C-{c j}) - H({d} | C);

[0040] In the above formula, H() is the conditional information entropy;

[0041] Preferably, the FRBAS method for rough set attribute reduction algorithms includes the following steps:

[0042] G100: Calculate the classification quality of decision attribute d with respect to condition attribute set C in the original decision information system S.

[0043] G200: Calculate each conditional attribute c j In C, attribute importance sig(c j (C,d), with c j Press sig(c) j Sort the data in descending order (C, d).

[0044] G300: Set RED to an empty set, i.e., RED←φ, and press sig(c j (C,d) Reverse order, for each c j Repeat the following steps for ∈C;

[0045] G400: c j Add to the set RED, i.e., RED←RED∪{c j}, calculate the classification quality of decision attribute d with respect to RED.

[0046] G500: If Then output RED; otherwise, go to step G400.

[0047] Preferably, the rule extraction in step S200 includes the following steps:

[0048] T100: Let X i =U / RED,Y j =U / {d} are the conditional equivalence classes and decision equivalence classes derived by partitioning RED with respect to U and d with respect to U, respectively;

[0049] T200: For any element x∈X i and element y∈Y j The classification rules for determining the weld penetration state based on conditional attributes are described as follows:

[0050] Des(X i →Des(Y) j )·(conf ij );

[0051] in, For equivalence class X i The description, Des(Y) j ) = (d = d(y)) represents the equivalence class Y jThe description of the conf ij =|X i ∩Y j | / |X i | is a rule confidence;

[0052] Preferably, in step S500, the fuzzy logic reasoning comprises the following sub-steps:

[0053] S510: input fuzzification, calculating the membership degree of the input accurate quantity in each fuzzy subset;

[0054] S520: rule evaluation, evaluating the truth value of each rule antecedent according to the fuzzified input;

[0055] S530: rule aggregation, taking the union operation to aggregate the reasoning results of each rule to obtain the total result;

[0056] S540: output defuzzification, using the center of gravity method to calculate the current regulation accurate quantity;

[0057] At the same time, a variable gap GMAW welding penetration control system based on rough set knowledge reduction is proposed, the control system comprises a memory, a processor and a computer program stored on the memory and executable on the processor; wherein, when the processor executes the computer program, the steps of the control method of any one of the above are realized.

[0058] The beneficial effects obtained by the present application are:

[0059] 1. The control method of the present application aims at the problem that neural network algorithm and other algorithms in the welding process control penetration recognition are difficult to establish the corresponding relationship between the molten pool front features and the penetration state due to high-dimensional characteristic parameters, taking the welding knowledge modeling of soft computing theory as the main line, giving the basic framework of rough set soft computing knowledge modeling of welding process, discussing the construction of a simple knowledge model for representing the penetration state by molten pool front features, and avoiding the molten pool recognition and prediction process;

[0060] 2. The control method of the present application proposes a welding forming control scheme based on rough set and fuzzy logic hybrid modeling; taking the low-carbon steel typical structure variable gap welding penetration control as the application background, on the basis of obtaining the molten pool characteristic parameters by the visual sensing device, using the rough set knowledge acquisition method to determine the corresponding molten pool front features under each penetration state, and quickly adjusting the welding parameters through fuzzy control, so as to achieve the real-time control of variable gap welding forming;

[0061] 3. The control method of the present application can be applied to various types of automatic welding equipment or semi-automatic welding equipment, and the original equipment can apply the control method of the present application by updating the system components or algorithms;

[0062] 4. The control system of the present application adopts modular design for each software and hardware part, which facilitates future upgrade or replacement of relevant software and hardware environment, and reduces the cost of use. BRIEF DESCRIPTION OF DRAWINGS

[0063] The present application can be further understood from the following description in conjunction with the drawings. The components in the drawings are not necessarily drawn to scale, but emphasis is instead placed upon illustrating the principles of the embodiments. Like reference numerals designate corresponding parts in different views.

[0064] BRIEF DESCRIPTION OF DRAWINGS

[0065] Figure 1 A schematic diagram of the variable gap GMAW welding penetration control method based on rough set knowledge reduction of the present application;

[0066] Figure 2 A schematic diagram of the algorithm flow of the FRMAD method of the embodiment of the present application;

[0067] Figure 3 A schematic diagram of the algorithm flow of the FRBAC method of the embodiment of the present application;

[0068] Figure 4 A schematic diagram of the algorithm flow of the FRBAS method of the embodiment of the present application;

[0069] Figure 5 A schematic diagram of the variable gap GMAW welding fuzzy control model principle of the embodiment of the present application;

[0070] Figure 6 A schematic diagram of the control system of the embodiment of the present application;

[0071] Figure 7 A schematic diagram of the welding equipment applying the control system of the embodiment of the present application. DETAILED DESCRIPTION

[0072] In order to make the technical solutions and advantages of the present application more clear and understandable, the present application will be further described in detail below in conjunction with its embodiments; it should be understood that the specific embodiments described herein are only used to explain the present application and not to limit the present application. Other systems, methods and / or features of the embodiments will become apparent to those skilled in the art upon inspection of the following detailed description. All such additional systems, methods, features and advantages are intended to be included within the scope of the present application. They are included in the scope of the present application and are protected by the appended claims. Additional features of the disclosed embodiments are described in the following detailed description, and will be apparent to one of ordinary skill in the art upon inspection of the following detailed description.

[0073] The same or similar reference numerals in the drawings of the embodiments of the present application correspond to the same or similar components; in the description of the present application, it should be understood that the orientation or positional relationship indicated by the terms "upper", "lower", "left", "right" and the like is based on the orientation or positional relationship shown in the drawings, and is only for the convenience of describing the present application and simplifying the description, and does not indicate or imply that the device or component referred to must have a particular orientation. The orientation and operation are constructed in a particular orientation, therefore the terms describing the positional relationship in the drawings are only used for exemplary illustration, and cannot be understood as a limitation of the present patent, and for those of ordinary skill in the art, the specific meanings of the above terms can be understood according to the specific circumstances.

[0074] Embodiment one: As the theory has been disclosed, rough set theory (Rough set, RS) as an effective mathematical tool for data representation, learning and induction has been widely used in knowledge discovery, machine learning, pattern recognition, decision support, inductive reasoning and many other fields since Pawlak proposed it. Its application in the field of welding has also gradually attracted attention and made certain progress; the core idea of rough set is to keep the classification ability unchanged, and to derive decision or classification rules through knowledge reduction; the corresponding molten pool front features under each penetration state are determined by the knowledge acquisition method of rough set, and then a fuzzy control model of molten pool front features is established by using the characteristics of fuzzy logic that can simulate human decision-making operation. In the closed-loop control, the penetration recognition and prediction link can be eliminated, and the penetration control target can be achieved by real-time monitoring of the molten pool front parameters;

[0075] Exemplarily, a variable gap GMAW welding penetration control method based on rough set knowledge reduction is proposed, as shown in FIG. 1, the control method comprises the following steps: Figure 1

[0076] ​S100: Perform a plurality of groups of welding penetration experiments; in the plurality of groups of the welding penetration experiments, different gap parameters and welding machine current parameters are selected for experiments, and a plurality of experimental data are collected; the experimental data at least include current, gap, penetration condition, and molten pool size information; and a penetration database of the variable gap-current relationship is established using the above experimental data;

[0077] S200: Data preprocessing is performed on the penetration database to construct an original information decision system S; the original information decision system S is subjected to attribute reduction by three rough set attribute reduction algorithms, namely, the FRMAD method, the FRBAC method, and the FRBAS method, to obtain reduced attributes, complete rule extraction, and construct a rough set model;

[0078] S300: The obtained reduced attributes are subjected to fuzzy membership function construction by the minimum fuzziness method, and a fuzzy set model is constructed by using the Mamdani reasoning algorithm for a continuous domain;

[0079] S400: In an actual welding process, a Hall sensor is used to measure welding arc signals, namely, welding arc current or welding arc voltage, and a visual sensor is used to obtain molten pool information;

[0080] S500: The data collected in step S400 are connected to a control system through a data acquisition card, and a current change amount is output by fuzzy logic reasoning to control penetration;

[0081] In step S200, data preprocessing is performed on the penetration database, including the following steps:

[0082] S210: Initialize data clustering centers;

[0083] S220: Calculate the distance of each point to the clustering center and the membership function matrix;

[0084] S230: Recursively select the clustering center;

[0085] S240: Recalculate the membership function matrix;

[0086] S250: Calculate the objective function, and if the effectiveness function reaches a minimum value, data clustering is completed;

[0087] Preferably, first, an original decision information system S is set:

[0088] S=(U,C∪D,V,f);

[0089] Wherein, U is a domain, which is a non-empty finite object set; C is a non-empty finite condition attribute set, D is a non-empty finite decision attribute set, V is a finite value domain; f is an information function, which gives the attribute value of any object (element) x in U according to the condition attribute and the decision attribute;

[0090] As shown in the accompanying drawings, the FRMAD method for rough set attribute reduction algorithm comprises the following steps: Figure 2

[0091] E100: calculating the classification quality of the decision attribute d about the condition attribute set C in the original decision information system S Wherein, β is a correct classification rate threshold value;

[0092] E200: setting the attribute simplest reduction set RED, and setting RED as an empty set, i.e. RED←φ;

[0093] E300: for each condition attribute c j ∈C-RED, wherein j=1, 2, ……n; calculating the classification quality of the decision attribute d about the condition attribute set RED∪{c j};

[0094] E400: selecting the attribute c that makes the classification quality j maximal, and adding it to the set RED, i.e. RED←RED∪{c j};

[0095] E500: if , then outputting RED; otherwise, returning to step E300 to circulate;

[0096] Wherein, the definition of the classification quality is as follows:

[0097]

[0098] In the above formula, pos C (d) is the positive region about d in the approximation space C;

[0099] The above FRMAD method first defines the attribute simplest reduction set as an empty set RED, and the FRMAD method selects the condition attribute set with the maximal approximation dependency by calculating the classification quality of each condition attribute, and adds it to the set RED one by one;

[0100] Wherein, the termination condition of the FRMAD method is

[0101] ​The FRMAD method takes the maximum approximation dependency as heuristic information, and considers that the condition attribute with the maximum approximation dependency is the most important attribute in the information system. The original information decision system S is added to the set RED one by one through the condition attribute with the maximum approximation dependency, so as to achieve the purpose of fast reduction. With the increase of the number of attributes, the classification quality decreases. The classification ability in the original decision information table is not reduced with the decrease of the condition attribute. The classification ability can be improved through different combinations of condition attributes.

[0102] By adjusting the definition of Ziarko reduction, the condition (1) in the definition of Ziarko reduction is replaced by that is, the classification quality of the simplest reduction set must not be lower than the original decision information table S, that is, the classification ability must be kept unchanged. The FRMAD method adds the attributes by the attribute adding method. The classification quality of the condition attribute has good stability in the reduction process. The simplest reduction is obtained more quickly through the relatively simple combination of condition attributes.

[0103] Further, the FRBAC method for rough set attribute reduction algorithm, as shown in the accompanying drawings, comprises the following steps: Figure 3

[0104] F100: Calculate the classification quality of the decision attribute d about the condition attribute set C in the original decision information system S

[0105] F200: Calculate the importance sig(c j ,C,d) of each condition attribute c j in C, arrange c j in descending order according to sig(c j ,C,d), and obtain the attribute core Core = Max(sig(c j ,C,d));

[0106] F300: Add Core to the set RED, that is, RED <- Core; calculate the classification quality of the decision attribute d about RED

[0107] F400: If , output RED; otherwise, jump to the following steps F500;

[0108] Then, according to the descending order of sig(c j ,C,d), repeat the following steps F500 and F600 for each c j ∈C;

[0109] F500: Add c j ​Add to the set RED, i.e., RED←RED∪{c j}, calculate the classification quality of decision attribute d with respect to RED.

[0110] F600: If Then output RED; otherwise, loop to step F400.

[0111] Among them, the attribute importance sig(c j The definition of ,C,d) is:

[0112] sig(c j ,C,d)=H({d}|C-{c j})-H({d}|C);

[0113] In the above formula, H() represents the conditional information entropy;

[0114] The FRBAC method used above assigns importance sig(c) to all conditional attribute sets C in a raw information decision system S. j Sort the conditional attributes (C, d) in descending order. The conditional attribute with the highest importance must exist in the kernel. First, calculate whether the attribute kernel has the minimum reduction attribute. Then, add other conditional attributes to the conditional attribute with the highest dependence in descending order of importance, until the classification quality of the entire set with respect to the decision attribute set d is greater than or equal to 1.

[0115] The conditional attributes of the original decision information system S are reduced and filtered to obtain the simplest set RED of the original decision information table. The conditional attributes in RED are the set of conditional attributes that can most directly reflect the melting state. The above FRBAC method can be used to establish a knowledge model of the conditional attribute set and melting state in RED.

[0116] Further details are attached. Figure 4 As shown, the FRBAS method for rough set attribute reduction algorithms includes the following steps:

[0117] G100: Calculate the classification quality of decision attribute d with respect to condition attribute set C in the original decision information system S.

[0118] G200: Calculate each conditional attribute c j In C, attribute importance sig(c j (C,d), with c j Press sig(c) j Sort the data in descending order (C, d).

[0119] G300: Set RED to an empty set, i.e., RED←φ, and press sig(c j(C,d) Reverse order, for each c j Repeat the following steps for ∈C;

[0120] G400: c j Add to the set RED, i.e., RED←RED∪{c j}, calculate the classification quality of decision attribute d with respect to RED.

[0121] G500: If Then output RED; otherwise, go to step G400.

[0122] The FRBAS algorithm, a fast reduction algorithm based on attribute importance, is similar in principle to the FRBAC algorithm based on attribute kernels. Attribute importance serves as the heuristic information for this algorithm, and a certain conditional attribute c is removed from C. j Afterwards, the conditional information entropy changes H({d}|C-{c}). j})-H({d}|C) as a metric conditional attribute c j The importance of the original information decision system S; sig(c j The larger the value of (C,d), the more important it is to the original decision information system; each condition attribute c j They are of different importance, so they are sorted in reverse order and added to RED accordingly;

[0123] The termination condition of the FRBAS method is as follows:

[0124] The conditional attributes of the original information decision system S are reduced and filtered to obtain the simplest set RED of the original information decision system S. The conditional attributes in RED are the set of conditional attributes that can most directly reflect the melting state. Thus, a knowledge model of the conditional attribute set and the melting state can be established.

[0125] Preferably, the rule extraction in step S200 includes the following steps:

[0126] T100: Let X i =U / RED,Y j =U / {d} are the conditional equivalence classes and decision equivalence classes derived by partitioning RED with respect to U and d with respect to U, respectively;

[0127] T200: For any element x∈X i and element y∈Y j The classification rules for determining the weld penetration state based on conditional attributes are described as follows:

[0128] Des(X i →Des(Y) j)·(conf ij );

[0129] wherein, is a description of the equivalence class X i , where C(x) represents the attribute values of element x with respect to the attribute set C; Des(Y j ) = (d = D(y)) is a description of the equivalence class Y j , where D(y) represents the attribute values of element y in the attribute set D; conf ij = |X i ∩Y j | / |X i | is a rule confidence;

[0130] Further, in step S300, the flow of the fuzzy reasoning is shown in the attached Figure 5 , including the following sub-steps:

[0131] S510: input fuzzification, calculating the membership degree of the input precise quantity in each fuzzy subset;

[0132] S520: rule evaluation, evaluating the truth value of each rule antecedent according to the fuzzified input;

[0133] S530: rule aggregation, taking the logical sum to aggregate the reasoning results of each rule to obtain the total result;

[0134] S540: output defuzzification, using the center of gravity method to calculate the current regulation precise quantity;

[0135] wherein, the variable gap GMAG welding fuzzy control model finds the simplest reduced set RED in rough set modeling, and determines the tail width coefficient C TW of the molten pool; therefore, a single variable fuzzy control model of the tail width coefficient C TW of the molten pool needs to be established, the input in the fuzzy controller is the error e and the error change rate ec of the tail width coefficient C TW of the molten pool, and the output is the welding current regulation amount ΔI, the error e is the difference between the tail width coefficient C TW of the molten pool at the current time in the welding process and the standard penetration state tail width coefficient (185, 1.2) of the molten pool, which is expressed as:

[0136] Embodiment two: this embodiment should be understood as at least containing all the features of any one of the preceding embodiments, and further improving on the basis thereof;

[0137] Exemplarily, as shown in the attached Figure 6As shown, illustrating an exemplary architecture diagram of the control system; the re-control system includes a computing device 12, which includes a processor 14, a volatile memory 16, an input module 18, an output module 20 and a non-volatile memory 24; the non-volatile memory 24 is used to store the data, application programs or other necessary information required by the control method of the present application;

[0138] Preferably, the computing device 12 includes a communication bus 22, which can operatively couple the processor 14, the input module 18, the output module 20 and the volatile memory 16 to the non-volatile memory 24; although the control system is described as being hosted (i.e., executed) at one computing device 12, it should be understood that the present control system can alternatively be hosted across multiple computing devices, the computing device 12 being communicatively coupled to the other multiple computing devices through a network;

[0139] Wherein the processor 14 includes one or more processors, which can be, for example, one or more of a central processing unit (CPU), a graphics processing unit (GPU), an application-specific integrated circuit (ASIC), a system on a chip (SOC), a field-programmable gate array (FPGA), a logic circuit, or other suitable type of microprocessing module configured to perform the functions described herein;

[0140] Further, the storage module can include the volatile memory 16 and the non-volatile memory 24;

[0141] The volatile memory 16 can be, for example, a random access memory (RAM), a static random access memory (SRAM), a dynamic random access memory (DRAM), etc., which only temporarily stores data during program execution and loses storage function after stopping power supply support; in some examples, a non-volatile random access memory (NVRAM) can be used;

[0142] Preferably, the non-volatile memory 24 is a memory that can retain instruction storage data even without external applied power, such as flash memory, hard disk, read-only memory (ROM), electrically erasable programmable memory (EEPROM), etc.; the non-volatile memory 24 includes programs for instructions for the control system to complete the control method described herein, and data used by these programs sufficient to perform the operations described herein, such as storing multiple sets of experimental data, the original information decision system S, etc.;

[0143] In some embodiments, the input module 18 is connected to the data acquisition card; the data acquisition card is further connected to the visual sensor or other necessary sensors, such as mentioned above, for collecting a series of data required by the control system during the welding process; and the input module 18 can be connected to input devices, such as keyboard, mouse, audio / video recording device, etc., through which the user inputs information such as text, video, audio, etc. to the control system;

[0144] In some embodiments, the output module 20 can be connected to the welding equipment, such as automatic welding machine, welding robot, etc., to output the working parameters to the welding equipment to implement the control method described herein; and the output module 20 can be connected to various types of output devices, such as display, sound, light or other objects that can be used to display information; the user receives the working information from the control system through a set of output devices, and in addition, the output module can be a personal device such as tablet computer or mobile phone, without limiting the specific operation mode here;

[0145] Further, as shown in the accompanying drawings Figure 7 , an exemplary welding equipment applying the control system is shown; wherein the welding equipment 60 is connected to the visual sensor 62 through a connecting component; the visual sensor 62 is preferably kept at a distance from the welding workpiece to avoid damage caused by high temperature or flame; at the same time, the visual sensor 62 can have an automatic focusing function, which can obtain the molten pool data from a certain distance.

[0146] Preferably, the visual sensor 62 moves synchronously with the welding torch 64 through the connecting component and can obtain the molten pool image with the same vision at each collection time, so as to improve the accuracy of the control system in analyzing the molten pool image data collected by the visual sensor 62.

[0147] In the above embodiments, the description of each embodiment has its own focus, and the parts not described or recorded in a certain embodiment can be referred to the related description of other embodiments.

[0148] Although the present application has been described above with reference to various embodiments, it should be understood that many changes and modifications can be made without departing from the scope of the present application. That is, the methods, systems and devices discussed above are examples. Various configurations can omit, substitute, or add various procedures or components as appropriate. For instance, the methods described can be performed in an order different from that described, and / or various steps can be added, omitted, and / or combined. Also, features described with respect to certain configurations can be combined in various other configurations, for instance, different aspects and elements of the configurations can be combined in a similar manner. Also, technology evolves and, thus, many of the elements are examples and do not limit the scope of the disclosure or claims.

[0149] In the description specific details are set forth in order to provide a thorough understanding of the exemplary configurations including implementations. However, configurations can be practiced without these specific details. For example, well known circuits, processes, algorithms, structures, and techniques have not been described in detail, without unnecessary detail, to avoid obscuring the configurations. This description provides example configurations only, and should not be taken as limiting the scope or applicability of the claims. Rather, the preceding description of the configurations will provide enabling descriptions for a skilled artisan to implement the described technology. Various modifications can be made to the functions and arrangements of elements without departing from the scope of the disclosure.

[0150] In view of the above, it will be seen that the details set forth in the preceding description are to be considered merely illustrative of the application and not restrictive. It will be appreciated that, from reading the foregoing description, many modifications will suggest themselves to a skilled artisan in the art, which modifications are intended to be encompassed by the present application. Accordingly, the scope of the present application should be determined by reference to the appended claims and their legal equivalents rather than by reference to the foregoing description.

Claims

1. A method for controlling the penetration of variable gap GMAW welding based on rough set knowledge reduction, characterized in that, The control method includes the following steps: S100: Conduct multiple sets of welding penetration experiments; in the multiple sets of welding penetration experiments, different gap parameters and welding machine current parameters are selected for the experiments, and multiple sets of experimental data are collected; the experimental data include at least current, gap, penetration status, and weld pool size information; and, a penetration database with variable gap-current relationship is established using the above experimental data. S200: Perform data preprocessing on the melt-through database to construct the original information decision system S; The original information decision system S is subjected to attribute reduction by three rough set attribute reduction algorithms, namely FRMAD method, FRBAC method and FRBAS method, to obtain reduced attributes, complete rule extraction, and construct rough set model. S300: The obtained reduced attributes are constructed using the minimum fuzziness method to construct fuzzy membership functions. For the continuous universe of discourse, the Mamdani inference algorithm is used to complete the fuzzy set model construction. S400: In the actual welding process, Hall sensors are used to measure the welding arc signal, i.e., the welding arc current or welding arc voltage, and visual sensors are used to obtain information about the molten pool. S500: The data collected in step S400 is connected to the control system through a data acquisition card, and the current change is output for fusion control through fuzzy logic reasoning. The data preprocessing of the melt penetration database in step S200 includes the following steps: S210: Initialize data cluster centers; S220: Calculate the distance from each point to the cluster center and the membership function matrix; S230: Cyclic selection of cluster centers; S240: Recalculate the membership function matrix; S250: Calculate the objective function. If the effectiveness function reaches its minimum value, the data clustering is complete.

2. The control method as described in claim 1, characterized in that, The FRMAD method for rough set attribute reduction algorithms includes the following steps: E100: Calculate the classification quality of decision attribute d with respect to condition attribute set C in the original decision information system S. Where β is the correct classification rate threshold; E200: Sets the simplest reduced set of attributes, RED, and sets RED to an empty set, i.e., RED←φ; E300: For each conditional attribute c j ∈C-RED, where j=1,2,……n; calculate the decision attribute d with respect to the condition attribute set RED∪{c j Classification quality E400: Select to enable The attribute c with the maximum value j Add it to the set RED, i.e., RED←RED∪{c j }; E500: If If so, output RED; otherwise, return to step E300 and repeat the loop. Among them, classification quality The definition is as follows: In the above formula, pos C (d) is the positive domain of d in the approximate space C; U is the universe of discourse, which is a non-empty finite set of objects.

3. The control method as described in claim 2, characterized in that, The FRBAC method for rough set attribute reduction algorithms includes the following steps: F100: Calculate the classification quality of decision attribute d with respect to condition attribute set C in the original decision information system S. F200: Calculate each conditional attribute c j In C, attribute importance sig(c j (C,d), with c j Press sig(c) j Sort the properties in descending order (C,d) to obtain the attribute core Core = Max(sig(c j ,C,d)); F300: Add Core to the set RED, i.e., RED←Core; calculate the classification quality of the decision attribute d with respect to RED. F400: If Then output RED; otherwise, skip to the next step F500. Then, press sig(c) j The order of C and d) in descending order, for each c j Repeat steps F500 and F600 below for ∈C; F500: will c j Add to the set RED, i.e., RED←RED∪{c j }, calculate the classification quality of decision attribute d with respect to RED. F600: If Then output RED; otherwise, loop to step F400. Among them, the attribute importance sig(c j The definition of ,C,d) is: sig(c j ,C,d)=H({d}|C-{c j })-H({d}|C); In the above formula, H() represents the conditional information entropy.

4. The control method as described in claim 3, characterized in that, The FRBAS method for rough set attribute reduction algorithms includes the following steps: G100: Calculate the classification quality of decision attribute d with respect to condition attribute set C in the original decision information system S. G200: Calculate each conditional attribute c j In C, attribute importance sig(c j (C,d), with c j Press sig(c) j Sort the data in descending order (C, d). G300: Set RED to an empty set, i.e., RED←φ, and press sig(c j (C,d) Reverse order, for each c j Repeat the following steps for ∈C; G400: c j Add to the set RED, i.e., RED←RED∪{c j }, calculate the classification quality of decision attribute d with respect to RED. G500: If If the output is RED, then proceed to step G400.

5. The control method as described in claim 4, characterized in that, The rule extraction in step S200 includes the following steps: T100: Let X i =U / RED,Y j =U / {d} are the conditional equivalence classes and decision equivalence classes derived by partitioning RED with respect to U and d with respect to U, respectively; T200: For any element x∈X i and element y∈Y j The classification rules for determining the weld penetration state based on conditional attributes are described as follows: Des(X i )→Des(Y j )·(conf ij ); in, For equivalence class X i The description, Des(Y) j ) = (d = d(y)) represents the equivalence class Y j Description, conf ij =|X i ∩Y j | / |X i | represents the confidence level of the rule.

6. The control method as described in claim 5, characterized in that, In step S500, the fuzzy logic reasoning includes the following sub-steps: S510: Input fuzzification, calculate the membership degree of the precise input quantity in each fuzzy subset; S520: Rule evaluation, which evaluates the truth value of each rule's antecedent based on the fuzzy input; S530: Rule aggregation, the union operation aggregates the reasoning results of each rule to obtain the total result; S540: Output defuzzification, using the centroid method to perform defuzzification calculations to obtain the precise current regulation value.

7. A variable gap GMAW welding penetration control system based on rough set knowledge reduction, characterized in that, The control system includes a memory, a processor, and a computer program stored in the memory and executable on the processor; wherein, when the processor executes the computer program, it implements the steps of the control method as described in any one of claims 1 to 6.

Images