A method for proposing optimal parameters for parts manufacturing.

Abstract

This provides a method for proposing optimal parameters for parts manufacturing. [Solution] The method includes generating optimization parameters for part manufacturing by training an AI model, generating a simulation dataset containing each set of manufacturing parameters with associated values, generating a training dataset using the simulation dataset generated by performing each manufacturing simulation using each set of manufacturing parameters with associated values, and training an AI model using the generated training dataset. The results of each manufacturing simulation include a representation of the part from each manufacturing simulation and the toolpath of each manufacturing simulation. The training dataset consists of a subset of the results of each manufacturing simulation and each set of manufacturing parameters with associated values ​​corresponding to the subset of results. The trained AI model is configured to determine the optimization manufacturing parameters.

Description

Technical Field

[0001] The present invention relates to a method for proposing optimal parameters for component manufacturing.

Background Art

[0002] Subtractive manufacturing, i.e., machining, and additive manufacturing are common manufacturing techniques. In subtractive manufacturing, raw material is gradually cut away to reveal the desired product shape. In contrast, in additive manufacturing, material is accumulated over time to create an object. Subtractive manufacturing is often performed using computer numerical control (CNC) machines. Machining operations may be classified as roughing, semi-finishing, and finishing operations. Roughing is the initial stage of machining where a large amount of material is removed from the raw material, semi-finishing is the second stage of the machining process, and often uses a more accurate tool path to further refine the rough shape left by the roughing process. Finishing is often the final stage of the process involving accurate tooling and final removal of material to complete the desired product.

Summary of the Invention

[0003] Maximum efficiency is a high priority in both subtractive manufacturing and additive manufacturing. For example, in subtractive manufacturing, efficiency is often related to an optimized tool path for the machine to follow during machining operations, as well as optimized manufacturing parameters. Unfortunately, the numerous parameters for manufacturing and determining an optimized tool path are very complex. Therefore, there is a need for a function to determine optimal tool path parameters for manufacturing, such as subtractive manufacturing and additive manufacturing. Multiple embodiments provide such functionality. Multiple embodiments can maximize the efficiency of manufacturing operations by eliminating the long, human-centered, trial-and-error process used by conventional systems to obtain optimized manufacturing parameter values. Specifically, multiple embodiments can train an artificial intelligence model (AI) to determine optimized parameters for component manufacturing.

[0004] An exemplary embodiment relates to a computer implementation method for training an AI model to generate optimized parameters for part manufacturing. One embodiment of the method generates a simulation dataset containing each set of manufacturing parameters having relevant values. Subsequently, a training dataset is generated using the simulation dataset produced by performing each manufacturing simulation using each set of manufacturing parameters having relevant values. The results of performing each manufacturing simulation may include a representation of the part resulting from each manufacturing simulation and the toolpath of each manufacturing simulation. According to one embodiment, the training dataset may consist of a subset of the results of performing each manufacturing simulation and each set of manufacturing parameters having relevant values ​​corresponding to the subset of results. Furthermore, the method trains an AI model using the generated training dataset. The trained AI model is configured to determine the optimized manufacturing parameters.

[0005] According to one embodiment, a trained AI model may determine optimized manufacturing parameters with optimized relevant values.

[0006] One embodiment may further include receiving a display of key performance indicators (KPIs) and a display of a given part. Such embodiments may further include processing the KPI display and the display of a given part with an AI model trained to determine a given manufacturing parameter for producing a given part that optimizes the KPI. In one embodiment, the KPI may be configured to be at least one of machining time, part surface quality, and amount of material removed. Furthermore, in such embodiments, the determined given manufacturing parameter may include an optimized value for the given manufacturing parameter. One embodiment may further include producing a given part using the determined given manufacturing parameter.

[0007] According to one embodiment, a given representation of a given part resulting from each given manufacturing simulation may be a top view image of the given part. In such an embodiment, the given representation of a given part resulting from each given manufacturing simulation may include depth information.

[0008] In another embodiment, a given representation of a given part resulting from each given manufacturing simulation may be a three-dimensional mesh model representation of the given part, such as a finite element model.

[0009] According to one embodiment, generating a training dataset may include selecting a subset of results from performing each manufacturing simulation based on one or more KPIs.

[0010] In yet another embodiment, each set of given manufacturing parameters may include at least one of the following: cutting tool, material, toolpath strategy, tool diameter ratio, cutting depth, step-over or overlap ratio, feed rate, spindle speed, start point, and end point. In such embodiments, the toolpath strategy may be spiral morphing, forward / backward movement, helical, concentric, or offset on part.

[0011] The embodiments may be used to train an AI model to generate optimized parameters for any type of manufacturing. For example, according to one embodiment, each manufacturing simulation of at least one part simulates additive manufacturing or subtractive manufacturing.

[0012] In one embodiment, the AI ​​model is a convolutional neural network (CNN).

[0013] In another embodiment, the AI ​​model is a first AI model, and the method may further include training a second AI model using the generated training dataset to predict the representation of the part. In such embodiments, the predicted representation may be a machined state representation, for example, a leftover material image. Furthermore, in such embodiments, the second AI model may be trained to predict the representation of the part for each of a plurality of toolpath strategies.

[0014] Another embodiment relates to a computer implementation system for training an AI model to generate optimized parameters for parts manufacturing. One embodiment of the system includes a processor and memory storing computer code instructions. The processor and memory may be configured to use the computer code instructions to cause the system to implement any embodiment or combination of embodiments described herein.

[0015] Another exemplary embodiment relates to a computer program product for training an AI model to generate optimized parameters for part manufacturing. In one exemplary embodiment, the computer program product includes a non-temporary computer-readable medium in which computer code instructions are stored. The computer code instructions, when executed by a processor, are configured to cause a device associated with the processor to implement any embodiment or combination of embodiments described herein.

[0016] It should be noted that embodiments of the Method, System, and Computer Program Product may be any embodiment or combination of embodiments described herein. [Brief explanation of the drawing]

[0017] The foregoing will become clear from the following more specific description of the exemplary embodiments, as similar reference letters throughout the different figures are illustrated in the accompanying drawings to refer to the same parts. The drawings are not necessarily to exact scale and are intended to emphasize that they illustrate embodiments.

[0018] [Figure 1A] Figure 1A shows an example of a graphical user interface (GUI) for manufacturing software. [Figure 1B] Figure 1B shows an example of a graphical user interface (GUI) for manufacturing software. [Figure 1C] Figure 1C shows an example of a graphical user interface (GUI) for manufacturing software. [Figure 1D] Figure 1D shows an example of a graphical user interface (GUI) for manufacturing software. [Figure 2] Figure 2 is a flowchart illustrating a method for training an artificial intelligence (AI) model to generate optimized manufacturing parameters according to one embodiment. [Figure 3] Figure 3 shows an example of the benefits that can be realized by utilizing the embodiments disclosed herein. [Figure 4] Figure 4 shows a workflow for generating an optimal toolpath strategy and manufacturing parameters for a machining operation, according to one embodiment. [Figure 5A] Figure 5A shows several exemplary top-down images used to train an AI model according to one embodiment. [Figure 5B] Figure 5B shows several exemplary top-down images used to train an AI model according to one embodiment. [Figure 6] Figure 6 shows an example of an optimized toolpath strategy determined by the embodiment. [Figure 7] Figure 7 shows an exemplary depth image that may be used by the embodiment. [Figure 8A]FIG. 8A shows a method for determining ground truth data for training an AI model according to an embodiment. [Figure 8B] FIG. 8B shows a method of using a simplified model network that can be used in a method for determining the ground truth of FIG. 8A by training an AI model. [Figure 9] FIG. 9 shows a method for generating a remaining material image and predicting optimized parameters by training an AI model according to an embodiment. [Figure 10] FIG. 10 shows an exemplary implementation of the embodiments disclosed herein on an exemplary software product. [Figure 11] FIG. 11 shows a computer network or similar digital processing environment in which an embodiment can be implemented. [Figure 12] FIG. 12 is a schematic block diagram showing an example of the internal structure of a computer in the environment of FIG. 11.

MODE FOR CARRYING OUT THE INVENTION

[0019] An explanation of exemplary embodiments will be described below.

[0020] Subtraction machining is a complex process produced by gradually removing material from a material, such as a bracket in the real world. Subtraction machining is often performed using computer numerical control (CNC) machining. In CNC, engineers and designers can provide a computer-controlled machining device with the geometric shape of the final product to be machined. The computer may then cause the machine to perform a series of machining operations on the raw material, gradually removing material from the material until the final part is cut out. This process of removing material from a material to reveal the desired part is called subtraction machining or machining. In subtraction machining, the CNC machine may use, for example, a rotary bit fixed to a computer-controlled gantry or arm that gradually removes material from layers of material by following a predetermined path and cutting along that path. The path that the CNC machine follows to cut out the desired product is called the toolpath.

[0021] Several strategies exist for optimizing toolpaths and related manufacturing parameters, which are used to define machining operations. Optimizing these strategies and parameters has traditionally involved a trial-and-error process with many iterations. Optimizing manufacturing parameters is often crucial for meeting key performance indicators (KPIs), such as reduced machining time, which are directly correlated with efficient toolpaths and optimized manufacturing parameters.

[0022] Embodiments disclosed herein provide a novel solution that optimizes manufacturing parameters for part machining by training an artificial intelligence (AI) model, providing determined optimized manufacturing parameters, including an optimized toolpath strategy that satisfies one or more KPIs. By training the AI ​​model to determine the optimized toolpath strategy and manufacturing parameters, the embodiments eliminate the aforementioned human-driven trial-and-error process.

[0023] There are numerous possible combinations of parameters that define a machining operation. For example, pocket machining, a common 2.5-dimensional (2.5D) operation in CNC machining that involves removing material from a workpiece to create a cavity, or pocket, involves a number of parameters that can vary in any number of combinations to define the optimized operation.

[0024] Figures 1A-D show exemplary CNC software graphical user interfaces (GUIs) illustrating numerous parameters used even in simple machining operations such as pocket machining. Figures 1A-D illustrate GUI panels 110, 120, 130, and 140, respectively. Each GUI panel 110, 120, 130, and 140 illustrates a subset of multiple parameters used in an exemplary machining operation.

[0025] The GUI panel 110 in Figure 1A shows the parameters for the toolpath strategy. In the exemplary GUI 110, the parameters include the toolpath strategy 111, e.g., concentricity, cutting direction 112, machining tolerance 113, fixture accuracy 114, pattern 115, movement 116, and maximum discretization 117. Each parameter may include numerous options. For example, the toolpath style parameter includes options shown in the submenu 118.

[0026] The GUI panel 120 in Figure 1B shows the parameters of the radial strategy 121, including the distance between paths 122, the total diameter ratio 123, the overhang ratio 124, the contour machining path 125, the clearance 126, and the supply amount 127. Again, the aforementioned parameters, for example, 121 to 127, may each include a number of options, including various radial strategy 121 options shown in the submenu 128.

[0027] The GUI panel 130 in Figure 1C displays the parameters of the axial strategy 131, including the maximum cutting depth 132, the number of levels 133, the automatic draft angle 134, the breakthrough 135, and the maximum inclination angle 136. Again, the aforementioned parameters, for example, 131-136, may each include a number of options, such as the various axial strategy 131 options shown in the submenu 138.

[0028] Panel 140 in Figure 1D illustrates finishing parameters such as pocket finishing mode 141, side finish thickness 142, number of side finish paths per level 143, bottom thickness on side finish 144, side thickness on bottom finish 145, and bottom finish thickness 146. Again, the aforementioned parameters, for example, 141-146, may each include a number of options, including various finishing mode 141 options shown in submenu 148.

[0029] As shown in Figures 1A-D, exemplary machining operations such as pocket machining involve multiple parameters, each with multiple options. While an example of pocket machining is shown in Figures 1A-D, it will be understood that similar parameter configurations are required for each operation in a manufacturing process, such as a subtractive machining process and an additive manufacturing process. Due to the multiple strategies and parameters, it is generally impossible to optimize manufacturing using existing trial-and-error techniques. The embodiment solves this problem by training an AI model to generate optimized parameters for part manufacturing. Figure 2 shows an example of such an exemplary embodiment.

[0030] Figure 2 is a flowchart of a method 200 for training an AI model to generate optimized parameters for parts manufacturing, according to one embodiment.

[0031] Method 200 begins in step 201 by generating a simulation dataset containing each set of manufacturing parameters having relevant values. Subsequently, in step 202, Method 200 generates a training dataset using the simulation dataset generated in step 201. The training dataset may be configured to be generated in step 202 by performing each manufacturing simulation using each set of manufacturing parameters having relevant values ​​obtained from the simulation dataset. According to one embodiment, the results of performing each manufacturing simulation include a representation of the part resulting from each manufacturing simulation and the toolpath for each manufacturing simulation. Furthermore, according to an embodiment of Method 200, the training dataset may consist of a subset of the results of performing each manufacturing simulation and each set of manufacturing parameters having relevant values ​​corresponding to the subset of results. Then, in step 203, Method 200 continues by training an AI model using the generated training dataset. The AI ​​model trained in step 203 is configured to determine the optimized manufacturing parameters.

[0032] Method 200 is performed on a computer, and therefore, its functionality and effective operation, such as steps 201-203, can be performed automatically by one or more digital processors. Furthermore, Method 200 can be performed using any computer device or combination of computing devices known in the art. Among other embodiments, Method 200 can be performed in particular using a computer network environment described below in connection with Figure 10, and a computer system described below in connection with Figure 11.

[0033] Method 200 begins in step 201 by generating a simulation dataset containing each set of manufacturing parameters having relevant values. According to one embodiment, when generating a simulation dataset for a part, a manufacturing process may be created and simulated for each set of manufacturing parameters. Simulating multiple manufacturing processes may result in multiple respective simulation results. These simulation results may include relevant manufacturing results as well as representations of the part resulting from the manufacturing process. For example, the manufacturing results may include machining time and total time, and the representation of the part may be the machined state of the part (e.g., material residual image). Each of these manufacturing results based on the relevant manufacturing parameters may be collected as a simulation output and later used to train an AI model. Each set of manufacturing parameters may relate to each, for example, a different manufacturing scenario.

[0034] The simulation dataset generated in step 201 may include data for any manufacturing parameters known to those skilled in the art. Each given set of manufacturing parameters generated in step 201 of Method 200 may include at least one of the following: cutting tool, material, toolpath strategy, tool diameter ratio, cutting depth, step over, feed rate, spindle speed, start point, and end point. In such embodiments, the toolpath strategy may be a spiral morphing strategy, a forward / backward movement strategy, a helical strategy, a concentric strategy, or an offset-on-part strategy. Furthermore, the relevant values ​​may be varied over a range of values ​​to generate a dataset that encompasses a changing manufacturing scenario.

[0035] To illustrate step 201, consider an example where the simulation dataset generated in step 201 includes three sets of manufacturing parameters, where the parameters are tolerance, number of levels, and overlap ratio, and each parameter has a corresponding value. In such an example, the simulation dataset includes set A [tolerance = 0.1, number of levels = 1, overlap ratio = 10%], set B [tolerance = 0.01, number of levels = 2, overlap ratio = 20%], and set C [tolerance = 0.001, number of levels = 3, overlap ratio = 30%].

[0036] In step 202, method 200 generates a training dataset using the simulation dataset by performing each manufacturing simulation, e.g., additive manufacturing or subtractive manufacturing, using each set of manufacturing parameters with relevant values. In one embodiment, each manufacturing simulation is performed by using a simulation dataset in an existing simulation / computer-aided engineering tool, such as DELMIA® machining. To illustrate step 202, consider the aforementioned embodiment in which the simulation dataset generated in step 201 consists of each set of manufacturing parameters, including set A [tolerance = 0.1, number of levels = 1, overlap ratio = 10%], set B [tolerance = 0.01, number of levels = 2, overlap ratio = 20%], and set C [tolerance = 0.001, number of levels = 3, overlap ratio = 30%]. In this example, in step 202, three simulations are performed, the first simulation using set A, the second simulation using set B, and the third simulation using set C. The results of each simulation (1st, 2nd, and 3rd) include representations of the part resulting from the manufacturing simulation, and the toolpaths for each simulation (e.g., vectors of points in space). Returning to this embodiment, the simulations yield three partial representations of the removed material and three toolpaths.

[0037] In Method 201, the training dataset consists of a subset of results from performing each respective manufacturing simulation, and a set of manufacturing parameters with associated values ​​corresponding to the subset of results. Note that the subset of results may include all or part of the results from performing the simulation in Step 202. To further illustrate Step 202, consider the aforementioned embodiment in which the results include three subrepresentations and three toolpaths. In this example, we assume that the results generated using Set A and Set B are included in the subset. Thus, the training dataset includes a representation of a part resulting from Set A conditions, and a toolpath from Set A conditions (subset of results) (Set A [tolerance = 0.1, number of levels = 1, overlap ratio = 10%] (manufacturing parameters with associated values ​​corresponding to the subset of results)), a representation of a part resulting from Set B conditions, and a toolpath of the representation of a part resulting from Set B conditions (subset of results) (Set B [tolerance = 0.01, number of levels = 2, overlap ratio = 20%]) (manufacturing parameters with associated values ​​corresponding to the subset of results).

[0038] According to embodiments of Method 200, the representation of a given part resulting from a given manufacturing simulation in step 201 may include a top view of the given part showing the amount of material removed from the material to manufacture the given part. Using the simulation dataset generated in step 201, multiple manufacturing simulations are performed using parameters from simulation datasets in existing simulation / computer-aided engineering tools, such as DELMIA® machining. These manufacturing simulations may generate machined state images of the part, e.g., material removal state images of the part, and may also generate images containing depth information, i.e., “depth images”. In some embodiments relating to removal processes, the “machined state” or “material removal state” image may also be called a “remaining material image”. This generated information may be useful in embodiments for determining optimized machining parameters. Furthermore, the representation resulting from the manufacturing simulation performed in step 201 may also include depth information, e.g., “depth images”. Depth information may be useful in embodiments for determining various depths for removing material from the material to manufacture a given part.

[0039] Furthermore, the given representation from the manufacturing simulation performed in step 201 may include, for each set of manufacturing parameter configurations, a three-dimensional mesh of the given part, for example, a finite element model representation of the given part, i.e., a mosaic representation of the given part.

[0040] In one embodiment of Method 200, generating a training dataset in step 202 may include selecting a subset of results from performing each manufacturing simulation based on one or more KPIs, i.e., selecting results that provide the most optimized manufacturing results with respect to one or more KPIs. Embodiments may perform this selection with respect to any desired KPI, including machining time, total time, and energy consumption, among other embodiments.

[0041] In step 203, the AI ​​model is trained using the generated training dataset. In one embodiment, the AI ​​model may be configured to be trained according to principles known to those skilled in the art. For example, the AI ​​model may be trained using the geometric shape of the area to be machined, e.g., the geometric shape of the material, as well as a number of simulation results generated in step 202. Each simulation result may include, for example, machining time, total time, and the machined state of the part (e.g., as an image or finite mesh) for a set of corresponding manufacturing parameters. As described above, the trained AI model is configured to determine optimized manufacturing parameters. In one embodiment, the AI ​​model is configured to determine optimized manufacturing parameters taking into account a representation of the part to be manufactured.

[0042] Furthermore, according to an embodiment of Method 200, the AI ​​model trained in step 203 may be configured to determine optimized manufacturing parameters with optimized relevant values.

[0043] The AI ​​model trained in step 203 of Method 200 may be, for example, a convolutional neural network (CNN). In such embodiments, the CNN may employ a U-NET-like architecture for generating machined state images of parts with optimized relevant values.

[0044] Method 200 may further include, for example, receiving a display of KPIs and a display of a given part, e.g., a 3D computer-aided design (CAD) model of the part to be manufactured, via user input. Method 200 may process the display of KPIs and the display of the given part with an AI model trained to determine given manufacturing parameters for manufacturing a given part that optimizes the KPIs. According to one embodiment, the given manufacturing parameters may include values ​​optimized for the given manufacturing parameters. The KPIs may be, for example, a display that optimizes the machining time associated with a given part, a display that optimizes quality criteria such as the surface quality of a given part, or a display that optimizes an indicator of the completeness of material removal from the material, or a combination thereof. Furthermore, one embodiment may further include manufacturing a given part using the determined given manufacturing parameters in order to manufacture a given part that optimizes the KPIs. Such embodiments may control a CNC machine according to the given manufacturing parameters, among other embodiments.

[0045] Furthermore, embodiments of Method 200 may train an additional AI model. Such embodiments may train a second AI model to predict a given representation of a given part, such as a machined state image, a top-down image, a remaining material image, etc. The additional AI model may predict representations of the part resulting from manufacturing under various manufacturing conditions, such as different toolpath strategies and / or various manufacturing parameters with varying values. This second AI model can be used in step 202 to perform a simulation. In such embodiments, for example, performing the simulation may be configured to provide manufacturing parameters to the trained model, which provides a representation of the machined state of the part.

[0046] Figure 3 shows an example of the implementation of the benefits realized by utilizing toolpaths and parameters suggested by an AI model trained according to Method 200, according to one embodiment. For example, an operator may configure manufacturing parameters, select a given toolpath 310, and perform pocket machining with a “forward and backward” motion 311. Alternatively, for example, the embodiment may instead determine the AI-recommended toolpath 320 and associated manufacturing parameters via a trained AI model trained according to Method 200, which shows that by using a helical inward toolpath 321, for example, a 30% increase in operator productivity and a 30% to 40% reduction in machining time can be achieved.

[0047] Figure 4 shows a workflow 400 for generating an optimal toolpath strategy and manufacturing parameters for a machining operation, such as pocket machining, according to one embodiment. Method 400 begins in step 410 by determining the profile shape of the part, as well as the associated profile 411, toolpath strategy, and parameters 412. The contour shape of the part (e.g., the area defining the pocket to be machined) may also be determined by the user selecting a face of the part to be pocketed, the contour of which defines the pocket or area to be machined. The information determined in step 410 may be referred to as a simulation dataset (e.g., generated in step 201 in Figure 2) (e.g., generated in step 202 in Figure 2) used to generate a training dataset.

[0048] Subsequently, step 420 initiates the synthesis, i.e., training, dataset generation process. In particular, among other embodiments, such functionality may be performed in step 202 of method 200. In one embodiment, the generation of a training dataset involves a computer program simulating the manufacturing of a given part, e.g., the part defined in step 410, multiple times. Each given simulation utilizes variable manufacturing parameters, such as parameters having values ​​that change (e.g., are repeated) from previous simulations, records the parameters 423 used in the simulation, and records the simulation results, such as associated machining times. The results of each simulation may include individual top-down images 422 of the part manufactured according to the associated manufacturing parameters and images 421, e.g., 2D images, that provide depth information to the part. The associated manufacturing parameters may include multiple different tools, parameters, toolpaths, machining times, and depth information, each having associated values.

[0049] The manufacturing parameters 423 stored in the training dataset generated in step 420 may be a table consisting of rows and columns. Each column in the training dataset may be associated with a given manufacturing parameter among several manufacturing parameters, for example, the distance between each path in a CNC machine path. Each row in the simulation dataset may also be associated with a given value for the corresponding manufacturing parameter, for example, a particular cutting tool having a specific column parameter used to generate a CNC machine toolpath using a selected pocket geometric shape may be associated with a manufacturing parameter. The training dataset may be generated by a processor that performs multiple manufacturing simulations of a given part, records the relevant simulated parameters and associated parameter values, and stores them in the training dataset. Furthermore, there may be hundreds of manufacturing parameters and thousands of potential associated values, and therefore the training dataset used to train the AI ​​model may contain thousands of data points.

[0050] Next, in step 430 of method 400, the training dataset 423 is cleaned for input to the AI ​​model. For illustrative purposes, the generated simulation data may contain out-of-bounds values ​​or images that do not add any benefit to training the AI ​​model. Cleaning may include removing these values. After cleaning, the training dataset is used to train the AI ​​model in step 440. The trained AI model may then determine the optimal machining parameters for optimizing KPIs, such as minimizing machining time. Finally, in step 450, the part to be manufactured is provided to the trained AI model, which determines the optimized parameters for manufacturing the part. A manufacturing device, such as a CNC machine, can then be programmed to manufacture the part using the AI-recommended toolpaths and parameters.

[0051] Figures 5A and 5B show exemplary representations of a given part resulting from their respective manufacturing simulations, e.g., plan view images or “remaining material” images of the given part. The remaining material images show the “machined state” of the part manufactured using various manufacturing parameters, e.g., tool diameter overlap ratio settings. The remaining material images show the amount of material removed by each variation in the manufacturing parameter settings, where the manufacturing parameter in this embodiment is the tool diameter overlap ratio. For example, remaining material images 510a to 510f represent the six increased percentages of the tool diameter overlap ratio as a manufacturing parameter for a complete machining operation in a spiral morphing pattern. Image 510a shows the remaining material when the value associated with the tool diameter overlap ratio is 10%, and it can be observed within the remaining material image 510a that a certain amount of material remains that was not removed by the machine. As the images progress from 510a (tool diameter overlap ratio 10%) to 510b (tool diameter overlap ratio 20%), 510c (tool diameter overlap ratio 30%), 510d (tool diameter overlap ratio 40%), 510e (tool diameter overlap ratio 50%), and 510f (tool diameter overlap ratio 60%), more material is removed.

[0052] Figures 5A and 5B include similar representations of remaining material based on variations in manufacturing parameters for multiple toolpath strategies. For example, images 520a–520f in Figure 5A illustrate material removal for each associated tool diameter overlap ratio parameter value for a forward-backward toolpath strategy. Images 530a–530f represent material removal for each associated tool diameter overlap ratio parameter value for a helical toolpath strategy, images 540a–540f similarly illustrate material removal for each associated tool diameter overlap ratio parameter value for a concentric toolpath strategy, and finally, images 550a–550f show material removal for each associated tool diameter overlap ratio parameter value for an offset-on-part toolpath strategy.

[0053] As shown in Figure 5A, it can be observed that the amount of material removed increases as the distance between each machining pass decreases (i.e., as the tool diameter overlap ratio increases). However, it is important to note that increasing the tool diameter overlap ratio may result in higher machining time to complete the machining operation and therefore require more energy. Thus, by processing these remaining material images for multiple relevant toolpath strategies, multiple embodiments may determine optimized manufacturing parameter values ​​(e.g., tool diameter overlap ratio) that optimize a given KPI while still adhering to several constraints, such as minimizing machining time. (See Figures 8A, 8B, and 9). Accordingly, according to one embodiment, the information provided in the remaining material representations shown in Figures 5A and 5B may be used by an AI model to determine an optimized set of machining parameters to optimize a given KPI, and then used to train the AI ​​model.

[0054] Figure 5B shows similar examples of the remaining material images shown in Figure 5A, but examples 560a-f, 570a-f, 580a-f, 590a-f and 599a-f in Figure 5B show that the same process may be performed in situations where it is not desirable to remove 100% of the material in a given pocketing operation.

[0055] Figure 6 shows an example of an optimized toolpath strategy proposal 600, which may be determined by an AI model trained according to an embodiment, for example, Method 200. For example, the AI ​​model (trained according to an embodiment) may receive reference geometric parameters 610, such as reference machining time 611, reference toolpath strategy 612, reference tool diameter ratio 613, reference pattern 614, reference movement 615, and reference channel width 616. The model may take the received pocket geometry along with other process and tool parameters, predict optimized parameters for each toolpath strategy, and propose the relevant optimized parameters for each toolpath strategy. According to one embodiment, the AI-proposed manufacturing parameters 620, 630, 640, 650, and 660 may include the resulting machining time (621, 631, 641, 651, and 661), toolpath strategies (622, 632, 642, 652, and 662), for example, inward spiral morphing 622 for proposal 620, helical 632 for proposal 630, offset on-part zigzag 642 for proposal 640, forward / backward movement 652 for proposal 650, concentric 662 for proposal 660, and tool diameter ratio (623, 633, 643, 653, 663). Furthermore, the proposal may include additional parameter suggestions (e.g., proposal 660 including a proposed pattern 664, a proposed movement 665 (e.g., zigzag), and a proposed channel width 666), with these parameters relating only to a concentric toolpath strategy.

[0056] Figure 7 shows an exemplary 2D depth image 700 that may be used by the embodiment. For example, the depth image 700 is an example of a representation of a part being manufactured. Furthermore, if a trained AI model is used, according to one embodiment, the depth image may be used as input to the model, and the model may suggest an optimized depth or number of levels, such as an incremental increase in pocket depth.

[0057] In the depth image 700, the x-axis 701 and y-axis 702 may be configured to represent units of measurement that accurately represent the geometry of the part being manufactured when the CNC machine performs a machining operation (e.g., pocketing). The scale 703 may be configured to represent a graded representation indicating the depth of the part represented in the image related to the part being manufactured, i.e., the z-axis of machining. For example, as the gradient of the scale 703 moves from a darker to a lighter tone, the associated machining depth increases. Thus, dark areas in the image 700 indicate parts of the part where little pocketing (i.e., machining depth) is performed, and bright areas in the image 700 indicate parts of the part where deeper pocketing is performed. Each increment change in the gradation of the scale 703 can be associated with a specific machining depth unit (e.g., meters) to accurately communicate the depth to be pocketed to the CNC machine. For example, it can be observed that areas with brighter shadows on the part (e.g., area 704) may be machined to a depth of 0.59 units, while areas with less brighter shadows (e.g., 705) may be machined to a depth of 0.49 units.

[0058] In one embodiment, the AI ​​model may be, for example, a CNN model. The AI ​​model may process a top image of a part (e.g., 700 in Figure 7) as input and predict the optimal manufacturing parameters for each of several machining strategies (e.g., forward / backward, helical, zigzag, spiral, concentric, etc.) for use in manufacturing, such as performing pocket machining. In one embodiment, the ground truth data for training the AI ​​model may be determined by identifying a setup (i.e., manufacturing conditions) that provides the highest quality machining while optimizing a given KPI (e.g., shortest machining time). For example, an embodiment may first identify a subset of parameters from several machining parameters that have shown to yield the best machining quality for a given strategy. From this identified subset, a set of parameters that minimizes machining time and optimizes a given KPI is selected as the optimal setup, e.g., ground truth. The ground truth may be provided to the CNN as an index of parameter values ​​optimized for a given KPI. The CNN model may then be trained to predict this particular optimized configuration, thereby allowing the model to recommend parameter settings configured to balance the efficiency and quality of each strategy.

[0059] Figures 8A and 8B illustrate methods 800a and 800b according to each embodiment for processing a series of top images (e.g., remaining material images 510a-f in Figure 5A) to determine optimal manufacturing parameters (e.g., tool diameter overlap ratio 820-824) and ground truth 825 for training an AI model.

[0060] Figure 8A shows processing 800a of generated residual material images (802, 803, 804, 805, 806, 807, 808, and 809) according to one embodiment to determine the ground truth 808 that optimizes a given KPI (e.g., 811). Figure 8B shows method 800b for determining optimized manufacturing parameters (820, 821, 822, 823, and 824) with respect to the ground truth 825 using an AI model. Figures 8A and 8B show the determination of ground truth 808 and 825 for a helical 801 toolpath strategy, and it will be understood that methods 800a and 800b can be performed to determine ground truth for forward / backward 816, zigzag 817, spiral 818, and concentric 819 toolpath strategies, as well as any other arbitrary toolpath strategy.

[0061] Figure 8A illustrates the determination of ground truth 808 according to one embodiment. Residual material images 802-809 may be generated by performing a manufacturing simulation (e.g., step 202 of Method 200) using various manufacturing parameters (e.g., tool diameter overlap ratio) and may represent a part, e.g., 813, manufactured under various manufacturing parameters. For example, residual material image 802 represents the amount of residual material (i.e., material not removed by machining) on ​​part 813 when the manufacturing parameter value of the tool diameter overlap ratio is set to 10% in the machining process. Similarly, Image 803 shows the amount of remaining material in part 813 where a 20% tool diameter overlap ratio was used as a manufacturing parameter value for the machining process; Image 804 shows the amount of remaining material in part 813 where a 30% tool diameter overlap ratio was used as a manufacturing parameter value; Image 805 shows the amount of remaining material in part 813 where a 40% tool diameter overlap ratio was used as a manufacturing parameter value; Image 806 shows the amount of remaining material in part 813 where a 50% tool diameter overlap ratio was used as a manufacturing parameter value; Image 807 shows the amount of remaining material in part 813 where a 60% tool diameter overlap ratio was used as a manufacturing parameter value; Image 808 shows the amount of remaining material in part 813 where a 70% tool diameter overlap ratio was used as a manufacturing parameter value; and Image 809 shows the amount of remaining material in part 813 where an 80% tool diameter overlap ratio was used as a manufacturing parameter value.

[0062] For each of the toolpath diameter overlap ratio parameter values, from these respective remaining material images 802-809 of machining part 813 using a helical toolpath strategy 801 with each parameter value (i.e., 10% 802, 20% 803, 30% 804, 40% 805, 50% 806, 60% 807, 70% 808, and 80% 809), method 800a determines which toolpath diameter overlap ratio value satisfies a given KPI (e.g., having the best machining time 811) and maintains the best quality of the machined 810. In this example, with respect to the helical toolpath strategy 801, both parameter values ​​70% 808 and 80% 809 represent the best quality of the machined 810. However, a tool diameter overlap ratio value of 70% 808 results in better (i.e., lower) machining time 811 compared to the machining time associated with a tool diameter overlap ratio value of 80% 809. Therefore, method 800a determines that the manufacturing parameter value for the tool diameter overlap ratio is 70% 808, and this manufacturing parameter value can be used as ground truth, e.g. 825, for training an AI model with respect to the helical tool path strategy 801.

[0063] Figure 8B illustrates a method 800b for processing an image of an input 812, i.e., a part 813 to be manufactured, via a CNN model 814 and a feedforward neural network 815, and determining the ground truth 825 for each. According to one embodiment, the ground truth may be optimized manufacturing parameters, such as tool diameter overlap ratio, machining direction, pattern, and channel width, for a plurality of toolpath strategies for machining a given part 813 (e.g., forward / backward movement 816, helical 801, zigzag 817, spiral 818, and concentric 819). Each manufacturing parameter may not be relevant to each respective toolpath strategy, and therefore it should be understood that embodiments may determine only the relevant manufacturing parameters as the ground truth for a given toolpath strategy.

[0064] Next, as illustrated in Figure 8B, an image of the part 813 to be manufactured serves as input 812 to model 814. The image 813 is processed by both a CNN model 814, which may employ an architecture such as RES-NET, as well as a feedforward neural network 815 for classification.

[0065] According to one embodiment, the result of the process predicts optimized manufacturing parameter values ​​as the respective ground truths 825. The ground truths may be used to train an AI model and / or manufacture part 813. In Figure 8A, the optimized tool diameter ratio 821 value for the helical 801 toolpath strategy may be a 70% overlap 808, which may be used as the ground truth 825 for training an AI model with respect to the helical toolpath strategy 801 according to one embodiment. Although only the ground truth 825 for the helical toolpath strategy 801 (i.e., 70% 808) is shown in Figure 8B, it will be understood that methods 800a and 800b can also be applied to each of the toolpath strategies, forward and backward movement 816, zigzag 817, spiral 818, and concentric 819, to determine the ground truth for each toolpath strategy.

[0066] Figure 9 shows a method 900 for determining optimized manufacturing parameters by explicitly generating a residual material image using a neural network 903, a neural network 956, and a feedforward neural network 957, according to one embodiment. In one embodiment of method 900, the neural network 903 may use an architecture such as U-NET, thereby enabling method 900 to predict the machining state of a part having optimized parameters. Without utilizing the neural network 903, method 900 may perform a method similar to methods 800a and 800b shown in Figures 8A and 8B.

[0067] Method 900 begins by receiving image 902 as input 901 to CNN956. Furthermore, the neural network 903 generates a number of remaining material images 910-917, 919-926, 928-935, 937-944, and 946-953 based on input image 901 902. According to one embodiment, the neural network, i.e., CNN903, is implemented using a U-NET-like encoder 904 and decoder (i.e., 905, 906, 907, 908, and 909) structure. Each decoder 905-909 may be configured to generate a respective remaining material image for each toolpath strategy and each manufacturing parameter (e.g., tool diameter overlap ratio). For example, decoder 905 generates the remaining material images for the helical 918 toolpath strategy (i.e., 910-917), decoder 906 generates the remaining material images for the concentric 927 toolpath strategy (i.e., 919-929), decoder 907 generates the remaining material images for the zigzag 936 toolpath strategy (i.e., 928-935), decoder 908 generates the remaining material images for the spiral 945 toolpath strategy (i.e., 937-944), and decoder 909 generates the remaining material images for the forward / backward 954 toolpath strategy (i.e., 946-953).

[0068] The generated remaining material images (i.e., 910-917, 919-926, 928-935, 937-944, and 946-953) represent multiple machining states of part 902 for each of multiple toolpath strategies 918, 927, 936, 945, and 954, each with a multiple percentage of the tool diameter overlap ratio. For example, remaining material images 910, 919, 928, 937, and 946 represent machining states of part 902 with a 10% tool diameter overlap ratio for each of the toolpath strategies 918, 927, 936, 945, and 954, and remaining material images 911, 920, 929, 938, and 947 represent machining states of part 902 with a 20% tool diameter overlap ratio for each of the toolpath strategies 918, 927, 936, 945, and 954. Images 912, 921, 930, 939, and 948 show the machining state of part 902 with a 30% tool diameter overlap ratio for each of the toolpath strategies 918, 927, 936, 945, and 954, respectively, while images 913, 922, 931, 940, and 949 show the machining state of part 902 with a 40% tool diameter overlap ratio for each of the toolpath strategies 918, 927, 936, 945, and 954. The remaining material images 914, 923, 932, 941, and 950 represent the machining state of part 902 with a 50% tool diameter overlap ratio for each of the respective toolpath strategies 918, 927, 936, 945, and 954, while the remaining material images 915, 924, 933, 942, and 951 represent the machining state of part 902 with a 60% tool diameter overlap ratio for each of the respective toolpath strategies 918, 927, 936, 945, and 954. The remaining material images 916, 925, 934, 943, and 952 represent the machining state of part 902 with a 70% tool diameter overlap ratio for each of the respective toolpath strategies 918, 927, 936, 945, and 954, while the remaining material images 917, 926, 935, 944, and 953 represent the machining state of part 902 with an 80% tool diameter overlap ratio for each of the respective toolpath strategies 918, 927, 936, 945, and 954.Method 900 processes residual material images 910-917, 919-926, 928-935, 937-944, and 946-953 generated from a U-NET such as CNN903 using a RES-NET such as CNN956 and a feedforward neural network 957 to generate optimal manufacturing parameters 958, 959, 960, 961, and 962 (e.g., tool diameter overlap ratio) for each toolpath strategy 918, 927, 936, 945, and 954. For example, method 900 may determine that, in order to manufacture part 902 using a helical toolpath strategy 918, the relevant parameter (e.g., tool diameter overlap ratio) should be optimized to a given KPI with a parameter value of 60% 915; in order to manufacture part 902 using a concentric toolpath strategy 927, the relevant parameter should be optimized to a given KPI with a parameter value of 50% 923; in order to manufacture part 902 using a zigzag toolpath strategy 936, the relevant parameter should be optimized to a given KPI with a parameter value of 40% 931; in order to manufacture part 902 using a spiral toolpath strategy 945, the relevant parameter should be optimized to a given KPI with a parameter value of 60% 942; and in order to manufacture part 902 using a forward / backward toolpath strategy 954, the relevant parameter should be optimized to a given KPI with a parameter value of 50% 950. According to one embodiment, once the optimal manufacturing parameters 958, 959, 960, 961, and 962 are determined for each respective toolpath strategy 918, 927, 936, 945, and 954, the parameters determined for each toolpath strategy may then be used to manufacture the part.

[0069] Naturally, while exemplary pocket machining is described herein in relation to embodiments, embodiments may be used in relation to any number of machining operations. For example, for a rough machining operation of a given part, the determination of ground truth used to train the AI ​​model may be similar to that of embodiments described in relation to Figures 2, 4, 8A-8B and 9. However, since rough machining operations involve parts with different depths at different locations, one embodiment may include depth information in both the input image of a given part (e.g., a finite element mesh) and the image of the final part after machining. The overall model architecture remains the same as that described in relation to exemplary pocket machining, but the inputs and outputs may differ due to separate machining parameters for the rough machining operation. In addition, the data processing pipeline may also differ because rough machining operations involve working with a 3D model rather than a two-dimensional (2D) image, as in pocket machining.

[0070] Figure 10 shows an exemplary implementation 1000 of an embodiment disclosed herein on an exemplary software product 1010. For example, a part 1001 may be designed in a given software product 1010. A user of the software product 1010 may select a region 1002 in which a given pocket machining 1003 is performed during manufacturing. The user may then instruct the software product to suggest optimized manufacturing parameters 1004 for a given selected region 1002. The software product 1010 may then be configured to determine the optimized manufacturing parameters according to, for example, methods 200, 400, 800a-b, and / or 900 disclosed herein, and to provide the user with an optimized toolpath strategy, i.e., a suggestion 600 (Figure 6).

[0071] Computer support Figure 11 is a schematic diagram of a computer network in which an embodiment may be implemented. The client computer / device 50 and server computer 60 provide processing, storage, and input / output (I / O) devices for running application programs and the like. The client computer / device 50 can also link to other computing devices, including other client devices / processors 50 and server computers 60, via a communication network 70. The communication network 70 can be part of a remote access network, a global network (e.g., the Internet), a cloud computing server or service, a collection of computers worldwide, a local area or wide area network, and a gateway that communicates with each other using its respective protocol (e.g., TCP / IP, Bluetooth®, etc.). Other electronic device / computer network architectures are also suitable.

[0072] Figure 12 is a block diagram illustrating an exemplary embodiment of computer nodes (e.g., client processor / device 50 or server computer 60) within the computer network 70 of Figure 11. Each computer node 50, 60 contains a system bus 79, which is a set of hardware lines used for data transfer between components of a computer or processing system. The system bus 79 is essentially a shared conduit that connects various different elements of a computer system (e.g., processor, disk storage, memory, I / O ports, network ports, etc.) and enables information transfer between these elements. An I / O device interface 82 is connected to the system bus 79 for connecting various input / output devices (e.g., keyboard, mouse, display, printer, speaker, etc.) to the computer nodes 50, 60. A network interface 86 allows the computer nodes to connect to various other devices connected to the network (e.g., network 70 in Figure 11). Memory 90 is provided with volatile storage for computer software instructions 92a and data 94a used to implement embodiments of the present disclosure. The disk storage 95 provides non-volatile storage for computer software instructions 92b and data 94b used to implement embodiments of the present disclosure, for example, methods 200, 400, 800a-b, and 900. The central processing unit 84 is also connected to the system bus 79 and provided for executing computer instructions.

[0073] In one embodiment, the processor routines 92a-92b and data 94a-94b are a computer program product (generally referred to as 92) comprising a non-temporary computer-readable medium (e.g., a removable storage medium such as a DVD-ROM, CD-ROM, diskette, or tape) that provides at least a portion of the software instructions for the disclosed system. The computer program product 92 can be installed by any preferred software installation procedure, as is well known in the art. In another embodiment, at least a portion of the software instructions may also be downloaded via cable, communication, and / or wireless connection. In yet another embodiment, the program of the Disclosure is a computer program propagated signal product embodied in a propagated signal on a propagating medium (e.g., radio waves, infrared waves, laser waves, sound waves, or electrical waves propagated over a global network such as the Internet or other networks). Such a carrier medium or signal provides at least a portion of the software instructions for the routines / programs 92 of the Disclosure.

[0074] In alternative embodiments, the propagated signal is an analog carrier wave or a digital signal carried on a propagation medium. For example, the propagated signal may be a digitized signal propagated on a global network (e.g., the Internet), a telecommunications network, or another network (such as network 70 in Figure 11). In one embodiment, the propagated signal is a signal transmitted over a period of time via the propagation medium, such as instructions for a software application sent in a packet over a network over a period of milliseconds, seconds, minutes, or more. In another embodiment, the computer-readable medium of the computer program product 92 is a propagation medium that the computer system 50 can receive and read, for example, by receiving the propagation medium and identifying the propagated signal embodied within the propagation medium, as described above for the computer program propagated signal product.

[0075] Generally, the term "carrier" or "transient carrier wave" encompasses the aforementioned transient signals, propagating signals, propagation media, storage media, and so on.

[0076] In other embodiments, the program product 92 may be implemented as so-called Software as a Service (SaaS), or as other installations or communications that support the end user.

[0077] Embodiments or aspects thereof may be implemented in the form of hardware, including but not limited to hardware circuits, firmware, or software. When implemented in software, the software may be stored on any non-temporary computer-readable medium configured to allow a processor to read the software or a subset of its instructions. The processor is then configured to execute instructions and operate a device, or to cause a device to operate in the manner described herein.

[0078] Furthermore, hardware, firmware, software, routines, or instructions may be described herein as performing specific operations and / or functions of a data processor. However, naturally, such descriptions included herein are merely for convenience, and such operations are actually the responsibility of the computing device, processor, controller, or other device that performs firmware, software, routines, instructions, etc.

[0079] Naturally, flowcharts, block diagrams, and network diagrams may contain more or fewer elements, be arranged differently, or be represented differently. However, even more naturally, a particular implementation may carry out in a particular way the number of block diagrams and network diagrams, as well as the number of block diagrams and network diagrams illustrating the execution of the embodiment, are determined.

[0080] Therefore, other embodiments may also be implemented in various computer architectures, physical computers, virtual computers, cloud computers, and / or some combination thereof, and thus the data processors described herein are for illustrative purposes only and not to limit the embodiments.

[0081] While exemplary embodiments have been specifically shown and described, those skilled in the art will understand that various modifications of form and detail can be made therein without departing from the scope of embodiments included in the appended claims.

[0082] For example, the foregoing description and details of the embodiments refer to, but are not limited to, the tools and platforms of the applicant and assignee (Dassault Systems Americas Corporation) and Dassault Systemes for illustrative purposes. Other similar tools and platforms are also suitable.

Claims

1. A computer implementation method for training an artificial intelligence (AI) model to generate optimized parameters for parts manufacturing, wherein the method is performed by a processor, To generate a simulation dataset containing each set of manufacturing parameters with related values, A training dataset is generated using a simulation dataset produced by performing each manufacturing simulation using each set of manufacturing parameters having related values, wherein (i) the results of performing each manufacturing simulation include representations of parts resulting from each manufacturing simulation and toolpaths for each manufacturing simulation, and (ii) the training dataset consists of a subset of the results of performing each manufacturing simulation and each set of manufacturing parameters having related values ​​corresponding to the subset of results. The process involves training an AI model using the generated training dataset, wherein the trained AI model is configured to determine optimized manufacturing parameters. Computer implementation methods, including those mentioned above.

2. The computer implementation method according to claim 1, wherein the trained AI model is configured to determine the optimized manufacturing parameters with optimized relevant values.

3. (i) receiving the display of key performance indicators (KPIs) and (ii) receiving the display of a given component, The representation of the KPI and the representation of the given part are processed by the AI ​​model trained to determine given manufacturing parameters for manufacturing the given part that optimizes the KPI. The computer implementation method according to claim 1, further comprising:

4. The computer mounting method according to claim 3, wherein the KPI is at least one of machining time, part surface quality, and amount of material removed.

5. The computer implementation method according to claim 3, wherein the given manufacturing parameters include values ​​optimized for the given manufacturing parameters.

6. The computer mounting method according to claim 3, further comprising manufacturing the given component using the determined given manufacturing parameters.

7. The computer mounting method according to claim 1, wherein a given representation of a given part resulting from each of the given manufacturing simulations is a top view image of the given part.

8. The computer mounting method according to claim 7, wherein the given representation of the given component resulting from each of the given manufacturing simulations includes depth information.

9. The computer mounting method according to claim 1, wherein a given representation of a given part resulting from each of the given manufacturing simulations is a three-dimensional mesh model representation of the given part.

10. The computer implementation method according to claim 1, wherein generating the training dataset includes selecting a subset of the results of performing each manufacturing simulation based on one or more KPIs.

11. The computer implementation method according to claim 1, wherein each given set of manufacturing parameters includes at least one of cutting tool, material, toolpath strategy, tool diameter ratio, cutting depth, step-over, feed rate, spindle speed, start point, and end point.

12. The computer implementation method according to claim 11, wherein the toolpath strategy is spiral morphing, forward / backward movement, helical, concentric, or offset onpart.

13. The computer mounting method according to claim 1, wherein each of at least one manufacturing simulations of at least one component simulates additive manufacturing or subtractive manufacturing.

14. The computer implementation method according to claim 1, wherein the AI ​​model is a convolutional neural network (CNN).

15. The AI ​​model is the first AI model, and the method is Using the generated training dataset, a second AI model is trained to predict the representation of the parts. The method according to claim 1, further comprising:

16. A computer implementation system for training an artificial intelligence (AI) model to generate optimized parameters for parts manufacturing, wherein the system Processor and The system comprises a memory on which computer code instructions are stored, and the processor and the memory are configured in the system. To generate a simulation dataset containing each set of manufacturing parameters with related values, A training dataset is generated using a simulation dataset produced by performing each manufacturing simulation using each set of manufacturing parameters having related values, wherein (i) the results of performing each manufacturing simulation include representations of parts resulting from each manufacturing simulation and toolpaths for each manufacturing simulation, and (ii) the training dataset consists of a subset of the results of performing each manufacturing simulation and each set of manufacturing parameters having related values ​​corresponding to the subset of results. The process involves training an AI model using the generated training dataset, wherein the trained AI model is configured to determine optimized manufacturing parameters. A computer implementation system configured to perform the following.

17. The processor and the memory use the computer code instructions stored in the memory to configure the system. (i) receiving the display of key performance indicators (KPIs) and (ii) receiving the display of a given component, The representation of the KPI and the representation of the given part are processed by the AI ​​model trained to determine given manufacturing parameters for manufacturing the given part that optimizes the KPI. The computer implementation system according to claim 16, further configured to perform the following:

18. The processor and the memory use the computer code instructions stored in the memory to configure the system. To manufacture the given part using the determined given manufacturing parameters, The computer implementation system according to claim 17, further configured to perform the following:

19. When generating the training dataset, the processor and the memory use the computer code instructions stored in the memory to the system, Selecting a subset of the results of performing each manufacturing simulation based on one or more KPIs, A computer implementation system according to claim 16, configured to perform the following:

20. A computer program product for training an artificial intelligence (AI) model to generate optimized parameters for parts manufacturing, wherein the computer program product comprises a non-temporary computer-readable medium on which computer code instructions are stored, and when the computer code instructions are executed by a processor, the device associated with the processor, To generate a simulation dataset containing each set of manufacturing parameters with related values, A training dataset is generated using a simulation dataset produced by performing each manufacturing simulation using each set of manufacturing parameters having related values, wherein (i) the results of performing each manufacturing simulation include representations of parts resulting from each manufacturing simulation and toolpaths for each manufacturing simulation, and (ii) the training dataset consists of a subset of the results of performing each manufacturing simulation and each set of manufacturing parameters having related values ​​corresponding to the subset of results. The process involves training an AI model using the generated training dataset, wherein the trained AI model is configured to determine optimized manufacturing parameters. A computer program product configured to execute [something].

Images