Automatic process design method and system applied to metal pipe production

Abstract

The application provides an automatic process design method and system applied to metal pipe production, comprising a basic process definition module, a process timing constraint relationship constructed in a network graph data structure, a global optimization engine constructed with a minimum process parameter deviation as a target and meeting material elongation and equipment capacity constraints, reading of a finished product CAD model, analysis of a reading result into geometric parameters based on the basic process definition module, output into a pipe feature parameter table, direct writing of an analysis result into an initial population of the global optimization engine, global optimization of process parameters by the global optimization engine, generation of optimal process parameters, automatic generation of a mold drawing and closed-loop feedback. The application can solve the problems of a broken process chain, parameter dependence on manual work, poor special-shaped adaptability and lack of closed-loop feedback in the prior art.

Description

Technical Field

[0001] This invention relates to the field of intelligent manufacturing technology, and more specifically, to an automated process design method and system for the production of metal pipe fittings. Background Technology

[0002] Current automated process design technologies for metal tubing (especially bicycle frame tubing) can be mainly summarized into three typical solutions:

[0003] (1) The "template-parameter" method based on secondary development of CAD drawings—such as the CAD→BIM automatic design process disclosed in Chinese patent CN202510263231.X. Its core structure is as follows: first, the two-dimensional CAD drawings are cleaned up, the floor information table is parsed, and the components are split and classified. Then, the structured data is output according to the preset BIM parameter standards. Finally, a three-dimensional model is generated in a platform such as Revit. The spatial position and connection relationship between each component (wall, floor, beam and column) depends entirely on the geometric coordinates in the original CAD layer. The process is manifested as a chain of steps: "drawing → layer parsing → family library matching → model assembly".

[0004] (2) The "instantiated reference" method for reinforcement of irregular structures—such as the automatic design of reinforcement for underground power tunnels disclosed in Chinese Patent CN202510748531.7. This method is constructed with a multi-level parametric model as the core, orthogonally classifying structural types (walls, beams, slabs, columns) and shapes (rectangular, circular, irregular); describing the geometric boundaries of irregular shapes through B-spline / NURBS, and then using the instantiated reference algorithm to call the reinforcement template library to output the spatial positioning coordinates of the reinforcement. The components are dynamically associated through the topology mapping module, and the process is "load input → template matching → reinforcement parameter optimization → collision detection → drawing output".

[0005] (3) For rigid-flex PCBs, the “copper-negative overlay” method is used, such as the auxiliary file automatic generation technology disclosed in Chinese Patent CN202411939179.7. Its structure is as follows: first, the outline files of flexible board layers and rigid board layers are distinguished, converted into copper sheets and scaled proportionally; then, the auxiliary layer required for PP milling grooves or cover films is formed by “enlarging / shrinking the copper sheet edge + negative overlay”. The positional relationship between each layer is guaranteed by preset dimensions (1 mm-3 mm, 0.5 mm-1 mm, etc.) and negative Boolean operations; the process is “outline distinction → copper sheet scaling → size compensation → negative overlay → file output”.

[0006] All three technologies mentioned above have achieved automation "from geometric description to manufacturing file", but the following problems still exist:

[0007] (1) The process chain is fragmented and data silos are serious.

[0008] Current technologies generally remain at the "single-process / single-discipline" automation level, leading to a disconnect between process parameters and mold design. For example, the 3D models generated by the CAD-BIM process cannot directly output the cavity parameters of tube-drawing or flaring molds; while PCB copper overlay methods only focus on two-dimensional interlayer compensation, lacking modeling of the three-dimensional plastic deformation laws of metals. The root cause lies in industry barriers: architecture, civil engineering, and electronics develop independently, lacking a unified cross-domain "geometry-process-manufacturing" data model.

[0009] (2) Parameter optimization relies on human experience and lacks a systematic model.

[0010] In the reinforcement design of Chinese patent CN202510748531.7, although the diameter and spacing of the reinforcing bars can be automatically iterated, the template library still uses empirical formulas as boundaries; the dimensional compensation values ​​(such as 1 mm-3 mm) in Chinese patent CN202411939179.7 are also empirical ranges. The forming of metal pipes involves the coupling of multiple physical fields such as material elongation, wall thickness reduction rate, and springback. Existing technologies have not established a global optimization model with the goal of "minimizing process deviation" while meeting equipment capacity constraints, resulting in a high number of trial molding-repair cycles and a persistently high defect rate.

[0011] (3) Irregularly shaped pipe fittings have poor adaptability and long customization cycle.

[0012] For non-circular cross-section or variable diameter pipe fittings, existing CAD-BIM and instantiation referencing technologies cannot directly express features such as "continuous taper variation" or "local flaring"; PCB copper foil methods are also limited to planar 2D shapes. This is because there is a lack of parameterizable 3D surface libraries and feature semantics for metal forming (such as tapered segment length and transition curvature radius). As a result, every time a new specification appears, the mold outline needs to be manually redrawn, significantly lengthening the design cycle.

[0013] (4) Lack of closed-loop feedback makes it difficult to achieve batch automatic design.

[0014] Existing technologies are mostly "one-off" outputs: once the CAD-BIM model is completed, it is not iterated with the process; reinforcement schemes are only manually adjusted after collision detection; and once PCB auxiliary files are generated, they are not corrected based on actual milling results. The lack of a closed-loop mechanism of "process-inspection-optimization" prevents design and manufacturing from forming data-driven batch iterations, thus restricting production efficiency for large-scale customization.

[0015] Metal pipe processing typically involves multiple continuous forming processes such as pipe drawing, tapping, and flaring. Existing technologies have yet to produce a comprehensive solution that couples the "pipe geometry, process parameters, and mold drawings" and can automatically optimize the multi-process route. Summary of the Invention

[0016] To address the four major problems of existing technologies—"fragmented process chain, manual parameter dependence, poor adaptability to irregular shapes, and lack of closed-loop feedback"—this invention aims to propose an automated process design method and system for the production of metal pipe fittings.

[0017] To achieve the above objectives, in a first aspect, the present invention provides an automated process design method for the production of metal pipe fittings, comprising the following steps:

[0018] A basic process definition module and process timing constraints are constructed. The basic process definition module is composed of a cascade of parameterized templates of multiple process units. The process timing constraints are constructed using a directed graph structure, where nodes represent each process and edges represent the predecessor-successor timing constraints between processes.

[0019] A global optimization engine is constructed with the goal of minimizing process parameter deviations while simultaneously satisfying material elongation and equipment capacity constraints. This global optimization engine employs a heuristic optimization algorithm, embedding material elongation constraint sub-models and equipment capacity constraint sub-models, and includes a built-in process optimization objective function.

[0020] f(x) = w1 × flaring ratio deviation + w2 × cross-sectional area change rate + w3 × Taper elongation rate; where w1~w3 are weighting coefficients;

[0021] Read the finished product CAD model; based on the basic process definition module, parse the reading result into pipe fitting geometric feature parameters and output the pipe fitting feature parameter table; then write the pipe fitting feature parameter table and the process timing constraint relationship together as initial data into the initial population of the global optimization engine;

[0022] The global optimization engine is used to globally optimize the process parameters and generate the optimal process parameters.

[0023] Automatic generation of mold drawings: The optimal process parameters are input into the post-processor, and the mold drawings are automatically output; the measurement data from the production site is received and input into the global optimization engine to update the material elongation constraint sub-model and the equipment capacity constraint sub-model, obtain the optimized process parameters, and update the mold drawings accordingly.

[0024] The above technical solution establishes a data chain connecting pipe fitting geometry, process route, and mold drawings, creating a unified data model across processes and resolving the problem of repeated trial molding caused by isolated parameters in multiple processes. It constructs a global optimization model that minimizes process parameter deviations while simultaneously satisfying material elongation and equipment capacity constraints, replacing manual trial-and-error based on experience and reducing defect rates. Through a parametric feature library, it supports arbitrary diameter, tapered, or irregularly shaped pipe fittings, enabling rapid adaptation to new specifications and automated batch design. This forms a complete closed loop of "reading finished product CAD, solving optimization algorithms, automatically generating mold drawings, and correcting process loops," shortening the design-manufacturing cycle and meeting the flexible production needs of industries such as bicycles for multiple varieties and small batches.

[0025] Furthermore, the multi-layered process units include tube drawing process units, taper process units, and flaring process units. The manufacturing experience of bicycle metal tubing is solidified into three types of enumerable processes, providing a unified semantics and computable objects for subsequent algorithms.

[0026] Furthermore, the process timing constraints are stored in the database in the form of a two-dimensional adjacency matrix, where the rows and columns of the matrix correspond to each process node, and the matrix weight represents the minimum or maximum allowed time interval between processes.

[0027] Furthermore, the heuristic optimization algorithm includes a genetic algorithm (GA), a particle swarm optimization algorithm (PSO), a Bayesian optimization algorithm (BO), or a simulated annealing algorithm (SA).

[0028] Furthermore, the material elongation constraint sub-model is embodied in the form of a system of inequalities:

[0029] (d′ / d)²+(D′ / D)²+(L′ / L)² ≤ ε max ;

[0030] Where ε max d′ is the material's ultimate elongation, d′ is the diameter of the smaller end after processing deformation, d is the diameter of the smaller end before processing deformation, D′ is the diameter of the larger end after processing deformation, D is the diameter of the larger end before processing deformation, L′ is the length after processing deformation, and L is the length before processing deformation.

[0031] Furthermore, the material elongation constraint sub-model is embedded through a machine learning-based surrogate model. This improves the fitting accuracy of the nonlinear constraints.

[0032] Furthermore, the measurement data from the production site is uploaded in real time by a coordinate measuring machine or an online laser diameter measuring instrument.

[0033] Secondly, the present invention provides an automated process design system for the production of metal pipe fittings, used to implement the automated process design method described above, comprising:

[0034] The basic process definition module consists of a cascade of parameterized templates for multiple layers of process units;

[0035] The process predecessor-successor relationship model uses a network graph data structure to construct process timing constraints. Nodes represent processes, and edges represent predecessor-successor timing constraints. It communicates bidirectionally with the basic process definition module through an API interface to pass mathematical constraints to the global optimization engine.

[0036] The CAD drawing parser parses the finished CAD drawing results into geometric parameters based on the basic process definition module and outputs a pipe fitting feature parameter table; the parsing results and the process timing constraints are directly written into the initial population of the global optimization engine;

[0037] The global optimization engine is encapsulated using a heuristic optimization algorithm, embedding a material elongation constraint sub-model and an equipment capacity constraint sub-model, and has a built-in process optimization objective function; it optimizes process parameters with the goal of minimizing process parameter deviations while simultaneously satisfying material elongation and equipment capacity constraints.

[0038] The parametric CAD template library stores the master drawing template corresponding to each process unit and receives parameters optimized by the global optimization engine to update the dimensions within the template.

[0039] The post-processor and closed-loop feedback interface enable the automatic generation of drawings and the writing back of deviation data, forming a closed-loop feedback.

[0040] Furthermore, the parametric CAD template library is located in a shared directory on the server and is accessed by the post-processor. This ensures that once the optimization parameters are determined, the 3D cavity of the mold can be generated within one second, achieving zero delay in drawing quality.

[0041] Furthermore, the global optimization engine is deployed on a local CPU or GPU and connected to the production process constraint model via a JSON message bus, which is embedded into the optimization engine as a hard constraint.

[0042] Compared with the prior art, the present invention has the following technical effects:

[0043] (1) Process chain integration: The existing CAD to process and mold design requires multiple independent software, but now it is completed in one go by a unified digital link, reducing the number of data format conversions from 5 to 0.

[0044] (2) Global optimization quality: Measured by the objective function of “process optimization degree”, taking the genetic algorithm as an example, the algorithm can converge in an average of 47 iterations (GA population 150, crossover probability 0.8, mutation probability 0.1), and the final process parameter deviation is reduced from the traditional ±8.5% to ±1.2%.

[0045] (3) Adaptability to irregular shapes, supporting flexible production: For new specifications with varying diameters, curvatures, or elliptical cross sections, only feature parameters need to be added to the CAD template library, reducing the average preparation time for new specifications from 3.5 days to 2.3 hours. It supports mixed production of ≥200 types of small batches and multiple varieties of pipe fittings, meeting the market trend of personalized bicycle customization and driving the technological upgrading of upstream mold factories and equipment factories in the industrial chain.

[0046] (4) Closed-loop correction accuracy: After the laser diameter gauge writes back the data on site, the system can control the wall thickness error within ±0.05 mm in the second iteration, and the number of closed-loop convergence times is ≤3.

[0047] (5) Shorten the design-manufacturing cycle and reduce costs: The time from receiving the finished product CAD to generating mold drawings has been reduced from 7.2 days to 0.9 days, a reduction of 87%. The number of initial mold trials for each specification has been reduced from an average of 4.3 times to 1.2 times. The number of specifications that each process engineer can manage simultaneously has increased from 12 to 68, and labor costs have decreased by 63%.

[0048] (6) Process experience is stored in the enterprise knowledge base in the form of reusable templates, reducing the training cycle of new employees by 50% and forming intelligent manufacturing solutions that can be exported. Attached Figure Description

[0049] The invention, its features and advantages will become more apparent from the following detailed description of non-limiting embodiments with reference to the accompanying drawings.

[0050] Figure 1 This is a schematic diagram of the framework of an automated process design system applied to the production of metal pipe fittings in an embodiment of the present invention;

[0051] Figure 2 This is a schematic diagram of an automated process design method applied to the production of metal pipe fittings in an embodiment of the present invention. Detailed Implementation

[0052] The present invention will be further described below with reference to the accompanying drawings and specific embodiments, but these are not intended to limit the scope of the invention.

[0053] In the following detailed description, numerous specific details are set forth to provide a more thorough understanding of the invention. However, it will be apparent to those skilled in the art that well-known algorithms and models are not shown in detail to avoid obscuring the gist of the invention; and that the techniques not detailed in the following effect examples are readily available prior art.

[0054] Example

[0055] See Figure 1 This embodiment takes the production of bicycle metal tubing as an example and provides an automated process design method and system for the production of metal tubing. Its main structure and connection relationships are as follows:

[0056] The basic process definition module 100, consisting of three cascaded layers—"tube drawing process unit, TAPER process unit, and flaring process unit"—is located at the front end of the system and communicates bidirectionally with the process predecessor-successor relationship model 200 via an API interface. Each unit is represented in the software as a reusable parameterized template. The manufacturing experience of bicycle metal tubing is solidified into three enumerable processes, providing a unified semantics (e.g., "tube drawing" is defined as "mandrel + die" diameter reduction forming), providing computable objects for subsequent algorithms.

[0057] The process predecessor-successor relationship model 200 is implemented using a network graph data structure. Nodes represent processes, and edges represent "predecessor-successor" timing constraints: pipe extraction node → taper node → flaring node; it is represented in the database as a two-dimensional adjacency matrix. The matrix weights store the minimum / maximum allowed time intervals. The process predecessor-successor relationship model 200 communicates bidirectionally with the basic process definition module through an API interface, passing mathematical constraints to the global optimization engine; it transforms traditional experience-based scheduling into explicit mathematical constraints, ensuring that the process route output by the algorithm is physically feasible.

[0058] The production process constraint model 300 includes a material elongation constraint sub-model and an equipment capacity constraint sub-model; these exist in the form of a system of inequalities, such as:

[0059] (d′ / d)²+(D′ / D)²+(L′ / L)² ≤ ε max ; where ε max Let d' be the material's limiting elongation, d′ be the diameter of the smaller end after processing deformation, d be the diameter of the smaller end before processing deformation, D′ be the diameter of the larger end after processing deformation, D be the diameter of the larger end before processing deformation, L′ be the length after processing deformation, and L be the length before processing deformation. As an alternative technical solution, the material elongation inequality can be replaced with a machine learning-based surrogate model to improve the fitting accuracy of the nonlinear constraints.

[0060] The Global Optimization Engine 400 encapsulates other heuristic optimization algorithms such as Genetic Algorithm (GA) or Particle Swarm Optimization (PSO). GA / PSO can be replaced with Bayesian Optimization (BO) or Simulated Annealing (SA), suitable for scenarios more sensitive to computational resources. It includes a built-in "Process Optimization" objective function.

[0061] f(x) = w1 × flaring rate deviation + w2 × cross-sectional area change rate + w3 × Taper elongation rate; where w1~w3 are weighting coefficients, automatically assigned based on industry experience boundary values.

[0062] The global optimization engine 400 can be deployed on a local CPU / GPU and is connected to the production process constraint model 300 and the CAD drawing parser 500 via a JSON message bus. It searches for the optimal combination of process parameters in a multi-dimensional parameter space to achieve global rather than local optima.

[0063] The production process constraint model 300 is embedded as a hard constraint in the global optimization engine 400. Any individual that does not meet the constraints will be immediately eliminated in the global optimization engine 400. The production process constraint model 300 transforms process feasibility from empirical judgment to mathematically determineable conditions, avoiding invalid calculations.

[0064] The CAD drawing parser 500 is a STEP / IGES reading module developed based on the open-source Open CASCADE kernel. It automatically parses finished 3D CAD models into algorithm-recognizable geometric parameters (D1, D2, L, t, etc.), outputting a pipe fitting feature parameter table (CSV), eliminating the need for manual secondary data entry. The parsing results are directly written into the initial population of the optimization engine. As an alternative technical solution, Open CASCADE can be replaced with Autodesk Forge or PTC Creo Toolkit to adapt to the enterprise's existing PLM (Product Lifecycle Management) environment.

[0065] The parametric CAD template library 600 stores the master template (*.dxf template file) corresponding to each process unit. All dimensions within the template are variable, and the template dimensions are updated using parameters optimized by the global optimization engine. As a preferred technical solution, the parametric CAD template library 600 is located in a shared directory on the server and is called by the post-processor, ensuring that once the optimization parameters are determined, the 3D cavity of the mold can be generated within 1 second, achieving zero delay in drawing.

[0066] See Figure 2The design of metal pipe fitting production molds using the above automated process design system includes the following steps:

[0067] S1. Define the basic process and parameters;

[0068] S2. Construct process timing constraints;

[0069] S3. Optimization algorithm calculates optimal process parameters: Read the finished product CAD, parse the finished product CAD reading results into geometric parameters based on the basic process definition module 100, and output them as a pipe fitting feature parameter table; The parsing results and the process timing constraints are directly written into the initial population of the global optimization engine;

[0070] S4. Generate process parameters and input them into the parametric CAD template library;

[0071] S5. Utilize the parametric CAD template library and output mold design drawings through a post-processor;

[0072] S6. Write back the deviation data through the closed-loop feedback interface, return the feedback results to the basic process and parameter definition stage, adjust the process parameters or constraints, and form a closed-loop feedback.

[0073] It should be noted that the device structure and accompanying drawings of this invention mainly describe the principle of this invention. In terms of the technical aspects of this design principle, the configuration of the device's power mechanism, power supply system, and control system is not fully described. However, those skilled in the art who understand the principle of the invention can clearly understand the specifics of its power mechanism, power supply system, and control system.

[0074] The above technical solutions have been validated in multiple batches of pipe fittings produced at Giant (China) Suzhou factory for six consecutive months. The validation results show that: process parameter deviation has been reduced from the traditional ±8.5% to ±1.2%; the average preparation time for new specifications has been shortened from 3.5 days to 2.3 hours; after the on-site laser diameter gauge writes back data, the system can control the wall thickness error within ±0.05 mm in the second iteration, with ≤3 closed-loop convergence times; the time from receiving the finished product CAD to issuing the mold NC code has been reduced from 7.2 days to 0.9 days, a reduction of 87%; the number of initial mold trials per specification has been reduced from an average of 4.3 to 1.2; the number of specifications that a process engineer can manage simultaneously has increased from 12 to 68, resulting in a 63% reduction in labor costs.

[0075] Those skilled in the art should understand that variations can be implemented by combining existing technology with the above embodiments, which will not be elaborated here. Such variations do not affect the essence of the present invention, and will not be elaborated here either.

[0076] The preferred embodiments of the present invention have been described above. It should be understood that the present invention is not limited to the specific embodiments described above. Systems and structures not described in detail should be understood as being implemented in a conventional manner in the art. Any person skilled in the art can make many possible variations and modifications to the technical solutions of the present invention using the methods and techniques disclosed above, or modify them into equivalent embodiments with equivalent changes, without departing from the scope of the present invention. This does not affect the essential content of the present invention. Therefore, any simple modifications, equivalent changes, and modifications made to the above embodiments based on the technical essence of the present invention without departing from the content of the present invention are still within the scope of protection of the present invention.

Claims

1. An automated process design method for the production of metal pipe fittings, characterized in that: Includes the following steps: A basic process definition module and process timing constraints are constructed. The basic process definition module is composed of a cascade of parameterized templates of multiple process units. The process timing constraints are constructed using a directed graph structure, where nodes represent each process and edges represent the predecessor-successor timing constraints between processes. A global optimization engine is constructed with the goal of minimizing process parameter deviations while simultaneously satisfying material elongation and equipment capacity constraints. This global optimization engine employs a heuristic optimization algorithm, embedding material elongation constraint sub-models and equipment capacity constraint sub-models, and includes a built-in process optimization objective function. f(x) = w1 × flaring ratio deviation + w2 × cross-sectional area change rate + w3 × Taper elongation rate; where w1~w3 are weighting coefficients; Read the finished product CAD model; based on the basic process definition module, parse the reading result into pipe fitting geometric feature parameters and output the pipe fitting feature parameter table; write the pipe fitting feature parameter table and the process timing constraint relationship together as initial data into the initial population of the global optimization engine; The global optimization engine is used to globally optimize the process parameters and generate the optimal process parameters. Automatic generation of mold drawings: The optimal process parameters are input into the post-processor, and the mold drawings are automatically output; the measurement data from the production site is received and input into the global optimization engine to update the material elongation constraint sub-model and the equipment capacity constraint sub-model, obtain the optimized process parameters, and update the mold drawings accordingly.

2. The automated process design method for metal pipe fitting production according to claim 1, characterized in that, The multi-layer process unit includes the tube drawing process unit, the TAPER process unit, and the flaring process unit.

3. An automated process design method for the production of metal pipe fittings according to claim 1 or 2, characterized in that, The process timing constraints are stored in the database in the form of a two-dimensional adjacency matrix, where the rows and columns of the matrix correspond to each process node, and the matrix weight represents the minimum or maximum allowed time interval between processes.

4. An automated process design method for the production of metal pipe fittings according to claim 1 or 2, characterized in that, The heuristic optimization algorithm includes genetic algorithm, particle swarm optimization algorithm, Bayesian optimization algorithm, or simulated annealing algorithm.

5. An automated process design method for the production of metal pipe fittings according to claim 1 or 2, characterized in that, The material elongation constraint sub-model is represented by a system of inequalities: (d′ / d)²+(D′ / D)²+(L′ / L)² ≤ ε max ; Where ε max d′ is the material's ultimate elongation, d′ is the diameter of the smaller end after processing deformation, d is the diameter of the smaller end before processing deformation, D′ is the diameter of the larger end after processing deformation, D is the diameter of the larger end before processing deformation, L′ is the length after processing deformation, and L is the length before processing deformation.

6. An automated process design method for the production of metal pipe fittings according to claim 1 or 2, characterized in that, The material elongation constraint sub-model is embedded through a machine learning-based surrogate model.

7. An automated process design method for the production of metal pipe fittings according to claim 1 or 2, characterized in that, Production site measurement data is uploaded in real time by a coordinate measuring machine or an online laser diameter measuring instrument.

8. An automated process design system for the production of metal pipe fittings, characterized in that, For implementing the automated process design method as described in any one of claims 1 to 7, comprising: The basic process definition module consists of a cascade of parameterized templates for multiple layers of process units; The process predecessor-successor relationship model uses a network graph data structure to construct the process timing constraints. Nodes represent processes, and edges represent predecessor-successor timing constraints. It communicates bidirectionally with the basic process definition module through an API interface to pass mathematical constraints to the global optimization engine. The CAD drawing parser parses the finished CAD drawing results into geometric parameters based on the basic process definition module, and outputs a pipe fitting feature parameter table; the parsing results and the process timing constraints are directly written into the initial population of the global optimization engine; The global optimization engine is encapsulated using a heuristic optimization algorithm, embedding a material elongation constraint sub-model and an equipment capacity constraint sub-model as the production process constraint model, and has a built-in process optimization objective function; it optimizes process parameters with the goal of minimizing process parameter deviations while simultaneously satisfying material elongation and equipment capacity constraints. The parametric CAD template library stores the master drawing template corresponding to each process unit and receives parameters optimized by the global optimization engine to update the dimensions within the template. The post-processor and closed-loop feedback interface enable the automatic generation of drawings and the writing back of deviation data, forming a closed-loop feedback.

9. An automated process design system for metal pipe fitting production according to claim 8, characterized in that, The parametric CAD template library is located in a shared directory on the server and is accessed by the post-processor.

10. An automated process design system for metal pipe fitting production according to claim 8, characterized in that, The global optimization engine is deployed on a local CPU or GPU and connects to the production process constraint model via a JSON message bus. The latter is embedded into the optimization engine as a hard constraint.

Images