Intelligent assembly station reconstruction method for final assembly

Abstract

The intelligent assembly station reconstruction method solves the technical problems that manual station planning is long in period, low in accuracy, insufficient in efficiency and the like. Comprising the following steps of 1, extracting and representing assembly information; 2, solving an assembly sequence based on an intelligent algorithm; and step 3, constructing, resolving and optimizing evaluation indexes. According to the method, unmanned aerial vehicle product structures and assembly process characteristics are analyzed, and assembly station evaluation indexes are constructed from the two levels of assembly procedures and occupied space; designing an assembly station evaluation index calculation mode, and constructing an assembly station evaluation index calculation method so as to support the evaluation of the assembly station; based on the assembly station evaluation indexes, a multi-criterion assembly station optimization method is established, and the acquisition of the optimal assembly station is supported. According to the method, the assembly station planning efficiency can be improved, and finally the effectiveness of the station planning method is verified through assembly simulation.

Description

Technical Field

[0001] The invention belongs to the technical field of overall assembly of aircraft products, and relates to an assembly station reconstruction planning method for a complex highly integrated system. Background Art

[0002] Reconstructing assembly stations is a prerequisite for efficient final assembly. The assembly cost (time and space occupied), assembly efficiency, and assembly quality of products are all affected by the quality of station planning. At present, the positioning of product assembly stations is mainly completed by manual subjectivity, and stations are usually determined based on the professional knowledge and experience of process personnel. At present, the final assembly of drone products is characterized by large size, complex shape, and a large number of parts and connectors. At this time, manual assembly stations cannot perfectly utilize space and achieve the highest assembly efficiency, which will lead to an extension of the product production cycle and increased costs. Secondly, different personnel have different positioning criteria for assembly stations, making it difficult to ensure standardization and consistency when manually setting stations, which causes difficulties for subsequent actual assembly. Therefore, the manual station setting method is difficult to apply to increasingly complex drone products, and an automated assembly station setting solution is urgently needed.

[0003] The establishment of assembly stations should be based on the given design scheme, taking into account the constraints such as assembly environment, assembly process, and assembly process. It should not only meet all geometric constraints but also shorten the assembly time and reduce the complexity of assembly as much as possible, so as to find reasonable and feasible dynamic assembly stations, so as to reduce the assembly cost and optimize product performance. Through the current research and analysis, there are several difficulties in the establishment of assembly stations: First, the representation of the assembly model. The assembly model needs to intuitively and comprehensively express the part information, matching information and process information; second, there are few evaluation indicators for assembly stations. The assembly station should not only meet the most basic geometric feasibility, but also further evaluate the feasible assembly stations from other aspects, and finally obtain both feasible and optimized assembly stations; third, the generation of assembly sequences. With the product assembly process, the structure becomes more and more complex, and the assembly stations conflict with each other and are prone to errors. Therefore, it is necessary to find a more efficient and accurate way to generate assembly sequences. Summary of the invention

[0004] In order to solve the technical problems of long planning cycle, low accuracy and insufficient efficiency of manual workstation planning, the present invention develops assembly information extraction and characterization technology for assembly processes, assembly workstation evaluation index construction, solution and optimization technology, and assembly sequence solution technology based on intelligent algorithms.

[0005] The present invention is achieved by the following technical solutions:

[0006] A method for reconfiguring an intelligent assembly station for general assembly, comprising the following steps:

[0007] Step 1: Extraction and representation of assembly information: Considering the assembly process decomposition and model hierarchy of complex products, the information demand analysis of assembly station planning is carried out, and the extraction method and standardized expression method of the three types of object information in the process information model, namely, parts, assembly process, and assembly constraints, are established. The product assembly information representation framework is established using the association graph model to form an assembly process information model that integrates product structure, assembly process, and assembly tools, providing an information source for product assembly station planning;

[0008] Step 1 specifically includes:

[0009] 1) Create the assembly information model of the product: First, establish the part information model of the product, including part quality, symmetry, matching clearance, and connection type information. Secondly, establish the assembly process information and assembly resource information. The assembly process information includes the number of assembly direction changes of each part, the type of assembly tool, the number of assembly tool changes, the assembly priority relationship of parts, and assembly action information. This type of information is stored in the form of a matrix to facilitate the input of the solution algorithm. Finally, establish the assembly constraint information model between parts, including fixed constraints, coincidence constraints, offset constraints, fixed connection constraints, surface contact, straight line contact, contact constraints, and angle constraints. Map various types of constraints to corresponding values ​​and store them in the form of a matrix;

[0010] 2) Extract assembly information: For assemblies, the inherent information of the model includes two aspects: one is the part information, such as part name, part geometry, part quality, etc.; the other is the assembly information between parts, such as matching relationship, etc. The depth-first traversal algorithm is used to obtain relevant information for the UG model assembly tree;

[0011] 3) Characterizing assembly information: combining the part information involved in the complex product assembly process design process with the assembly information, and mapping them to a directed graph framework to form an assembly relationship graph model. The graph model is used to express the positioning relationship information, assembly resource information, and assembly operation information of the assembly object.

[0012] Step 2: Solving assembly sequence based on intelligent algorithm: sort out relevant assembly process knowledge and resources, form interference matrix, connection matrix and support matrix as input and constraint conditions for assembly station planning, clarify the basic conditions that feasible assembly stations need to meet, form assembly station solution space, in the station solution space, take product assembly stability and assembly aggregation factors as evaluation indicators, give constraint parameters and calculation method of objective function, take maximizing target benefit as solution target of assembly station planning problem, finally form mathematical model of assembly station planning, select heuristic algorithm, generate feasible assembly stations and optimize according to corresponding fitness function and evaluation index, and finally generate candidate set of assembly stations;

[0013] In step 2, when constructing the interference matrix, the assembly is fixed in the three-dimensional space, and the posture does not change. The active part changes its posture continuously during the disassembly process through the user-controlled step length and posture transformation operation. In the process of changing posture, real-time collision detection is performed with other parts, interference information is recorded, and the interference matrix is ​​obtained;

[0014] In step 2, the specific steps of constructing the connection matrix are:

[0015] Step 1: Obtain the overall 3D model information of the product assembly and record the 3D geometric information of each part in the model;

[0016] Step 2: Set the static interference inspection order of the parts, do not inspect the inspected parts again, assign values ​​based on the original inspection results, and perform clearance analysis on the uninspected parts;

[0017] Step 3: Determine whether the gap Dc between the parts is greater than 0. If Dc>0, it is determined that the two parts have no connection relationship and go to step 5. Otherwise, go to step 4.

[0018] Step 4: According to the part properties, determine whether the parts are in a stable connection relationship. If so, go to step 5. Otherwise, determine that the two parts are in a contact connection relationship, and go to step 5.

[0019] Step 5: Determine whether all parts have been inspected. If so, output the connection matrix. Otherwise, go to step 2.

[0020] In step 2, the specific steps of constructing the support matrix are:

[0021] Step 1: Obtain the overall 3D model information of the product assembly and record the 3D geometric information of each part in the model;

[0022] Step 2: Select the support direction and set the step length;

[0023] Step 3: Set the static interference inspection order of the parts and perform clearance analysis on the parts;

[0024] Step 4: Determine whether the parts interfere with each other. If there is no interference between the two parts, it is determined that there is no support relationship between the parts, and the support matrix elements between the parts are assigned a value of 0;

[0025] Step 5: When the motion step length is set, implement the step-by-step interference test;

[0026] Step 6: Determine whether the normal vector of the contact surface between part i and part j is consistent with the direction of gravity. If the two directions are consistent, it is determined that part i has a stable connection relationship with part j, s ij =1, otherwise s ij =0;

[0027] Step 7: Determine whether all parts have been identified. If so, output the support matrix. Otherwise, go to step 3.

[0028] Step 3. Construction, calculation and optimization of evaluation indicators: Analyze the product structure and assembly process characteristics of UAVs, and start from the two levels of assembly process and occupied space to construct evaluation indicators for assembly stations; design the calculation method for assembly station evaluation indicators, and construct a calculation method for assembly station evaluation indicators to support the evaluation of assembly stations; based on assembly station evaluation indicators, establish a multi-criteria assembly station optimization method to support the acquisition of the best assembly station.

[0029] In step 3, the calculation method of the assembly station evaluation index is the DPSO algorithm, and the specific steps are:

[0030] Step 1: Input interference matrix, connection matrix, support matrix, and assembly tool set;

[0031] Step 2: Set the population size n size , inertia weight w and learning factors c1, c2; set the maximum number of iterations G max The initial iteration step is t=1, and the velocity vector V and position vector X are randomly generated in the initial population;

[0032] Step 3: Calculate the fitness function value of each assembly sequence. For each assembly sequence in a certain iteration process, if the comprehensive evaluation function value of the current assembly sequence is better than the individual historical optimal assembly sequence P i If it is excellent, the current assembly sequence is used as the new individual historical optimal solution and the original assembly sequence is updated;

[0033] Step 4: Compare all individual historical optimal assembly sequences P i The fitness function value of the optimal assembly sequence P is selected i As the global optimal assembly sequence P g ;

[0034] Step 5: Update the position and speed of each assembly sequence;

[0035] Step 6: Evaluate the fitness function value of the updated assembly sequence. If the comprehensive evaluation function value of the existing assembly sequence is better, use the current assembly sequence to update the individual historical optimal assembly sequence P i , otherwise it will not be updated;

[0036] Step 7: Compare all current assembly sequences P i , and take the optimal assembly sequence as the global optimal assembly sequence P g ;

[0037] Step 8: Update the number of iterations t = t + 1, if t ≤ G max, then go to step 3 and repeat the operation, otherwise, go to step 9 and repeat the operation;

[0038] Step 9: Output the optimal assembly sequence P g And the fitness function value f, the optimal assembly sequence obtained by DPSO search is the reasonable assembly sequence.

[0039] In step three, the evaluation indicators of assembly stations are divided into two categories: part-level evaluation indicators and assembly-level evaluation indicators.

[0040] In step 3, the assembly station optimization method is to calculate the part-level assembly sequence evaluation index and the assembly-level sequence evaluation index respectively based on the existing reasonable assembly and part parameters, and finally calculate the weight matrix W = (w1, w2, ... w 12 ) and each index membership matrix F = (f 1 ,f 2 ,…f 12 ) T The comprehensive evaluation value of each reasonable assembly sequence is calculated by multiplying the product of , so as to select the best assembly sequence.

[0041] Furthermore, according to the existing optimal assembly sequence, a dynamic assembly station change sequence is established correspondingly to the optimal assembly sequence, thereby realizing intelligent planning of assembly stations.

[0042] The present invention focuses on solving the problems of incomplete or non-intuitive assembly model representation, few assembly station evaluation indicators, and low efficiency of manually formulating assembly stations, improves the efficiency of assembly station planning, and ultimately verifies the effectiveness of the station planning method through assembly simulation. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] Figure 1 : Schematic diagram of parts information requirements;

[0044] Figure 2 : Schematic diagram of assembly depth-first traversal flow chart;

[0045] Figure 3 : Connection matrix generation algorithm flow chart;

[0046] Figure 4 : Support matrix generation algorithm flow chart;

[0047] Figure 5 Wherein the DPSO algorithm flow chart is shown;

[0048] Figure 6 : Assembly sequence evaluation index;

[0049] Figure 7 : Assembly sequence comprehensive evaluation flow chart;

[0050] Figure 8 : Assembly station reconstruction flow chart. DETAILED DESCRIPTION

[0051] In order to fully describe the information related to assembly and provide complete and correct basic information support for the subsequent assembly station planning, the assembly station planning problem first requires a reasonable product assembly information model. On this basis, appropriate methods are selected to plan and optimize the product assembly stations, judge their feasibility and rationality, and finally conduct a comprehensive evaluation of the optimized assembly stations to select the best assembly station.

[0052] The technical solution of the present invention is further described in detail below through the accompanying drawings and implementation examples.

[0053] like Figure 1-8 As shown, the implementation process of the present invention is:

[0054] 1) Create an assembly information model for the product: First, establish a part information model for the product, including information such as part quality, symmetry, fit clearance, connection type, etc. Secondly, establish assembly process information and assembly resource information. Assembly process information includes the number of assembly direction changes for each part, assembly tool type, number of assembly tool changes, parts assembly priority relationship, assembly actions, etc. This type of information is stored in a matrix form to facilitate the input of the solution algorithm. Finally, establish an assembly constraint information model between parts, including fixed constraints, coincident constraints, offset constraints, fixed connection constraints, surface contact, straight line contact, contact constraints, and angle constraints. Map various types of constraints to corresponding values ​​and store them in a matrix form (attached Figure 1 ).

[0055] 2) Extract assembly information: For assemblies, the inherent information of the model includes two aspects: one is the part information, such as part name, part geometry, part quality, etc.; the other is the assembly information between parts, such as the matching relationship, etc. The depth-first traversal algorithm is used to obtain relevant information on the UG model assembly tree (see Appendix). Figure 2 ).

[0056] 3) Characterizing assembly information: The part information involved in the complex product assembly process design process is combined with the assembly information, and mapped to a directed graph framework to form an assembly relationship graph model. The graph model is used to express the positioning relationship information, assembly resource information, and assembly operation information of the assembly object. For assembly a, its corresponding descriptor G a It can be expressed as:

[0057] G a =(V a ,E a )

[0058] Where:

[0059] V a——A set of descriptor vertices, and V a = {v a1 ,v a2 ,…,v an}, where n is the number of vertices;

[0060] E a ——A set of descriptor edges, and E a ={e a1 ,e a2 ,…,e am}, where m is the number of edges;

[0061] 4) Build a relationship matrix

[0062] Construct interference matrix: Human-computer interaction simulation assembly means that the assembly is fixed in three-dimensional space and its posture does not change. The active parts are controlled by the user through step size and posture transformation operations, such as translation and rotation, so that the active parts continuously change their posture during the disassembly process. In the process of changing posture, real-time collision detection is performed with other parts, interference information is recorded, and the interference matrix is ​​obtained.

[0063] Construct the connectivity matrix:

[0064] Step 1: Obtain the overall 3D model information of the product assembly and record the 3D geometric information of each part in the model;

[0065] Step 2: Set the static interference inspection order of the parts, do not inspect the inspected parts again, assign values ​​based on the original inspection results, and perform clearance analysis on the uninspected parts;

[0066] Step 3: Determine whether the gap Dc between the parts is greater than 0. If Dc>0, it is determined that the two parts have no connection relationship and go to step 5. Otherwise, go to step 4.

[0067] Step 4: According to the part properties, determine whether the parts are in a stable connection relationship. If so, go to step 5. Otherwise, determine that the two parts are in a contact connection relationship, and go to step 5.

[0068] Step 5: Determine whether all parts have been inspected. If so, output the connection matrix; otherwise, go to step 2.

[0069] The specific process of the algorithm is as follows Figure 3 shown.

[0070] Construct the support matrix:

[0071] Step 1: Obtain the overall 3D model information of the product assembly and record the 3D geometric information of each part in the model;

[0072] Step 2: Select the support direction and set the step length;

[0073] Step 3: Set the static interference inspection order of the parts and perform clearance analysis on the parts;

[0074] Step 4: Determine whether the parts interfere with each other. If there is no interference between the two parts, it is determined that there is no support relationship between the parts, and the support matrix elements between the parts are assigned a value of 0;

[0075] Step 5: When the motion step length is set, implement the step-by-step interference test;

[0076] Step 6: Determine whether the normal vector of the contact surface between part i and part j is consistent with the direction of gravity. If the two directions are consistent, it is determined that part i has a stable connection relationship with part j, s ij =1, otherwise s ij =0;

[0077] Step 7: Determine whether all parts have been identified. If so, output the support matrix. Otherwise, go to step 3.

[0078] The specific process of the algorithm is as follows Figure 4 shown.

[0079] 5) Constructing mathematical models

[0080] For an assembly consisting of n parts, A = {c1, c2, ..., c n}, in order to obtain its reasonable assembly sequence set S = {s1,s2,…,s n The key to} is the following two points:

[0081] The assembly sequence is geometrically feasible. Geometric feasibility is a strong constraint on the assembly sequence. Only when the assembly sequence is geometrically feasible, it is meaningful to optimize the target constraints.

[0082] Assembly sequence optimization. Assembly sequence optimization is a series of indicators that constitute the objective function or fitness function. When the feasible sequence makes the fitness function optimal, the sequence is considered to be a reasonable assembly sequence.

[0083] 6) Design intelligent algorithms

[0084] 7) Assembly sequence planning process based on DPSO algorithm (Appendix Figure 5 )

[0085] The specific steps of the discrete particle swarm algorithm applied to the assembly sequence planning problem are as follows:

[0086] Step 1: Input interference matrix, connection matrix, support matrix, and assembly toolset.

[0087] Step 2: Set the population size n size, inertia weight w and learning factors c1, c2; set the maximum number of iterations G max The initial iteration step number t = 1. The velocity vector V and the position vector X are randomly generated in the initial population.

[0088] Step 3: Calculate the fitness function value of each assembly sequence. For each assembly sequence in a certain iteration process, if the comprehensive evaluation function value of the current assembly sequence is better than the individual historical optimal assembly sequence P i If it is excellent, the current assembly sequence is used as the new individual historical optimal solution and the original assembly sequence is updated.

[0089] Step 4: Compare all individual historical optimal assembly sequences P i The fitness function value of the optimal assembly sequence P is selected i As the global optimal assembly sequence P g .

[0090] Step 5: Update the position and speed of each assembly sequence.

[0091] Step 6: Evaluate the fitness function value of the updated assembly sequence. If the comprehensive evaluation function value of the existing assembly sequence is better, use the current assembly sequence to update the individual historical optimal assembly sequence P i , otherwise it will not be updated.

[0092] Step 7: Compare all current assembly sequences P i , and take the optimal assembly sequence as the global optimal assembly sequence P g .

[0093] Step 8: Update the number of iterations t = t + 1, if t ≤ G max , then go to step 3 and repeat the operation; otherwise, go to step 9 and repeat the operation.

[0094] Step 9: Output the optimal assembly sequence P g And the fitness function value f, the optimal assembly sequence obtained by DPSO search is the reasonable assembly sequence.

[0095] 8) Construct evaluation indicators (attached Figure 6 )

[0096] The factors affecting the process of assembly are analyzed. Starting from the two levels of key parts and assembly, the assembly sequence evaluation index is constructed. 12 assembly sequence evaluation indexes are described, which are divided into two categories: part-level evaluation index and assembly-level evaluation index. The index code is U i Indicated by, where superscript i is the index number, for part-level evaluation index, the independent attribute value of each part under the part-level index is expressed as Indicates, i represents the index number, and k is the part number in the sequence).

[0097] 9) Assembly sequence optimization guided by evaluation index (Appendix Figure 7 )

[0098] According to the available reasonable assembly and part parameters, the part-level assembly sequence evaluation index and assembly-level sequence evaluation index are calculated respectively. Finally, the weight matrix W = (w1, w2, …w 12 ) and each index membership matrix F = (f 1 ,f 2 ,…f 12 ) T The comprehensive evaluation value of each reasonable assembly sequence is calculated by multiplying the product of , so as to select the best assembly sequence.

[0099] 10) Assembly sequence guide station reconstruction (attached Figure 8 )

[0100] According to the existing optimal assembly sequence, a dynamic assembly station change sequence is established accordingly to achieve intelligent planning of assembly stations.

[0101] The method of the present invention is an intelligent assembly station planning method proposed to solve the problems of complex assembly process, difficult assembly path planning, high rework rate, and high cycle risk of highly integrated systems for batch assembly of unmanned aerial vehicles. It can effectively avoid the technical problems of insufficient efficiency of on-site station allocation due to the large number of product loading parts, insufficient experience of process personnel, and the development of multiple products and multiple lines in the same cycle.

Claims

1. A method for reconfiguring an intelligent assembly station for general assembly, characterized by: The following steps are involved: Step 1: Extraction and representation of assembly information: Considering the assembly process decomposition and model hierarchy of complex products, the information demand analysis of assembly station planning is carried out, and the extraction method and standardized expression method of the three types of object information in the process information model, namely, parts, assembly process, and assembly constraints, are established. The product assembly information representation framework is established using the association graph model to form an assembly process information model that integrates product structure, assembly process, and assembly tools, providing an information source for product assembly station planning; Step 2: Solving assembly sequence based on intelligent algorithm: sort out relevant assembly process knowledge and resources, form interference matrix, connection matrix and support matrix as input and constraint conditions for assembly station planning, clarify the basic conditions that feasible assembly stations need to meet, form assembly station solution space, in the station solution space, take product assembly stability and assembly aggregation factors as evaluation indicators, give constraint parameters and calculation method of objective function, take maximizing target benefit as solution target of assembly station planning problem, finally form mathematical model of assembly station planning, select heuristic algorithm, generate feasible assembly stations and optimize according to corresponding fitness function and evaluation index, and finally generate candidate set of assembly stations; Step 3. Construction, calculation and optimization of evaluation indicators: Analyze the product structure and assembly process characteristics of UAVs, and start from the two levels of assembly process and occupied space to construct evaluation indicators for assembly stations; design the calculation method for assembly station evaluation indicators, and construct a calculation method for assembly station evaluation indicators to support the evaluation of assembly stations; based on assembly station evaluation indicators, establish a multi-criteria assembly station optimization method to support the acquisition of the best assembly station.

2. According to claim 1, a method for reconfiguring an intelligent assembly station for final assembly is characterized by: Step 1 specifically includes: 1) Create the assembly information model of the product: First, establish the part information model of the product, including part quality, symmetry, matching clearance, and connection type information. Secondly, establish the assembly process information and assembly resource information. The assembly process information includes the number of assembly direction changes of each part, the type of assembly tool, the number of assembly tool changes, the assembly priority relationship of parts, and assembly action information. This type of information is stored in the form of a matrix to facilitate the input of the solution algorithm. Finally, establish the assembly constraint information model between parts, including fixed constraints, coincidence constraints, offset constraints, fixed connection constraints, surface contact, straight line contact, contact constraints, and angle constraints. Map various types of constraints to corresponding values ​​and store them in the form of a matrix; 2) Extract assembly information: For assemblies, the inherent information of the model includes two aspects: one is the part information, such as part name, part geometry, part quality, etc.; the other is the assembly information between parts, such as matching relationship, etc. The depth-first traversal algorithm is used to obtain relevant information for the UG model assembly tree; 3) Characterize assembly information: Combine the parts information and assembly information involved in the complex product assembly process design process, and map them to a directed graph framework to form an assembly relationship graph model. Use the graph model to express the positioning relationship information, assembly resource information, and assembly operation information of the assembly object.

3. The method for reconfiguring an intelligent assembly station for final assembly according to claim 1 is characterized by: In step 2, when constructing the interference matrix, the assembly is fixed in three-dimensional space and its posture does not change. The active parts continuously change their posture during the disassembly process through user-controlled step size and posture transformation operations. In the process of changing posture, real-time collision detection is performed with other parts, interference information is recorded, and the interference matrix is ​​obtained.

4. The method for reconfiguring an intelligent assembly station for final assembly according to claim 1 is characterized by: In step 2, the specific steps of constructing the connection matrix are: Step 1: Obtain the overall 3D model information of the product assembly and record the 3D geometric information of each part in the model; Step 2: Set the static interference inspection order of the parts, do not inspect the inspected parts again, assign values ​​based on the original inspection results, and perform clearance analysis on the uninspected parts; Step 3: Determine whether the gap Dc between the parts is greater than 0. If Dc>0, it is determined that the two parts have no connection relationship and go to step 5. Otherwise, go to step 4. Step 4: According to the part properties, determine whether the parts are in a stable connection relationship. If so, go to step 5. Otherwise, determine that the two parts are in a contact connection relationship, and go to step 5. Step 5: Determine whether all parts have been inspected. If so, output the connection matrix; otherwise, go to step 2.

5. The method for reconfiguring an intelligent assembly station for final assembly according to claim 1 is characterized by: In step 2, the specific steps of constructing the support matrix are: Step 1: Obtain the overall 3D model information of the product assembly and record the 3D geometric information of each part in the model; Step 2: Select the support direction and set the step length; Step 3: Set the static interference inspection order of the parts and perform clearance analysis on the parts; Step 4: Determine whether the parts interfere with each other. If there is no interference between the two parts, it is determined that there is no support relationship between the parts, and the support matrix elements between the parts are assigned a value of 0; Step 5: When the motion step length is set, implement the step-by-step interference test; Step 6: Determine whether the normal vector of the contact surface between part i and part j is consistent with the direction of gravity. If the two directions are consistent, it is determined that part i has a stable connection relationship with part j, s ij =1, otherwise s ij =0; Step 7: Determine whether all parts have been identified. If so, output the support matrix. Otherwise, go to step 3.

6. The method for reconfiguring an intelligent assembly station for final assembly according to claim 1 is characterized by: In step 3, the calculation method of the assembly station evaluation index is the DPSO algorithm, and the specific steps are: Step 1: Input interference matrix, connection matrix, support matrix, and assembly tool set; Step 2: Set the population size n size , inertia weight w and learning factors c1, c2; set the maximum number of iterations G max The initial iteration step is t=1, and the velocity vector V and position vector X are randomly generated in the initial population; Step 3: Calculate the fitness function value of each assembly sequence. For each assembly sequence in a certain iteration process, if the comprehensive evaluation function value of the current assembly sequence is better than the individual historical optimal assembly sequence P i If it is excellent, the current assembly sequence is used as the new individual historical optimal solution and the original assembly sequence is updated; Step 4: Compare all individual historical optimal assembly sequences P i The fitness function value of the optimal assembly sequence P is selected i As the global optimal assembly sequence P g ; Step 5: Update the position and speed of each assembly sequence; Step 6: Evaluate the fitness function value of the updated assembly sequence. If the comprehensive evaluation function value of the existing assembly sequence is better, use the current assembly sequence to update the individual historical optimal assembly sequence P i , otherwise it will not be updated; Step 7: Compare all current assembly sequences P i , and take the optimal assembly sequence as the global optimal assembly sequence P g ; Step 8: Update the number of iterations t = t + 1, if t ≤ G max , then go to step 3 and repeat the operation, otherwise, go to step 9 and repeat the operation; Step 9: Output the optimal assembly sequence P g And the fitness function value f, the optimal assembly sequence obtained by DPSO search is the reasonable assembly sequence.

7. The method for reconfiguring an intelligent assembly station for final assembly according to claim 1 is characterized by: In step three, the evaluation indicators of assembly stations are divided into two categories: part-level evaluation indicators and assembly-level evaluation indicators.

8. The method for reconfiguring an intelligent assembly station for final assembly according to claim 1 is characterized by: In step 3, the assembly station optimization method is to calculate the part-level assembly sequence evaluation index and the assembly-level sequence evaluation index respectively based on the existing reasonable assembly and part parameters, and finally calculate the weight matrix W = (w1, w2, ... w 12 ) and each index membership matrix F = (f 1 ,f 2 ,…f 12 ) T The comprehensive evaluation value of each reasonable assembly sequence is calculated by multiplying the product of , so as to select the best assembly sequence.

9. The method for reconfiguring an intelligent assembly station for final assembly according to claim 1 is characterized by: According to the existing optimal assembly sequence, a dynamic assembly station change sequence is established accordingly to achieve intelligent planning of assembly stations.

Images