Method and device for parameterising a production process
Patent Information
- Application Number
- EP2023702067
- Authority / Receiving Office
- EP · EP
- Patent Type
- Applications
- Current Assignee / Owner
- Priority Date
- 2022-01-28
- Filing Date
- 2023-01-23
- Publication Date
- 2024-12-04
AI Technical Summary
In manufacturing processes, especially those with multiple complex steps, it is challenging to adjust process parameters optimally due to diverse influencing factors and interactions, making it difficult to achieve high component quality, and existing methods often require expert intervention and are not reproducible.
A computer-implemented method using a data-based process parameter model that trains on measured and state variables to optimize process parameters, leveraging a quality model to determine the best parameters for each step, allowing for automated optimization of control variables and parameters through a deep neural network trained with varying data points and quality functions.
This approach enables the optimization of process parameters across complex manufacturing processes, improving component quality by accounting for various influencing factors and interactions, and allows for automated parameter adjustment without requiring expert knowledge, enhancing reproducibility and accuracy.
Smart Images

Figure 1.1
Abstract
Description
[0001] Description
[0002] title
[0003] Method and device for parameterizing a manufacturing process
[0004] Technical area
[0005] The invention relates to manufacturing processes with one or more process steps for producing a component. The invention further relates to methods for adjusting the one or more process steps to increase the quality of the manufactured component.
[0006] Technical background
[0007] In the manufacture of components, manufacturing processes are typically used that create the final product from one or more intermediate products through processing or finishing. The manufacturing processes used can generally be carried out in many different ways and are usually adjusted using process parameters. These process parameters typically specify a constant or variable control, or in the case of a closed-loop control, control parameters for controlling a tool in the manufacturing process with a corresponding manipulated variable. In a closed-loop control system, the control of a machining tool is linked to the manufacturing progress.
[0008] Disclosure of the invention
[0009] According to the invention, a method for parameterizing a manufacturing process for producing a component according to claim 1 as well as a device and a manufacturing arrangement according to the independent claims are provided.
[0010] Further embodiments are specified in the dependent claims.
[0011] According to a first aspect, a computer-implemented method for providing a process parameter model for parameterizing one or more process steps of a manufacturing process for producing or processing a component is provided, comprising the following steps:
[0012] Providing a quality model for determining a quality, wherein the quality model is designed to specify the quality of the resulting component directly or with the aid of a predetermined quality function depending on one or more predetermined measured variables and / or one or more predetermined state variables, each of which specifies a property of at least one precursor product of the component to be manufactured and / or of a manufacturing device for carrying out a process step and / or at least one environmental condition, and on one or more process parameters that control a corresponding one of the process steps;
[0013] Training a data-based process parameter model to output one or more process parameters depending on one or more sensor-detected measured variables and / or one or more specified state variables by optimizing the quality.
[0014] To manufacture a component, one or more sequential process steps are typically used, each of which involves the processing or finishing of one or more intermediate products or the semi-finished component in its current state of production. The processing and finishing is carried out in each process step using production equipment.
[0015] The process step performed by a manufacturing device can be controlled using one or more process parameters. The process parameters can specify specific control variables, such as the current of an electric actuator, the opening angle of a valve, the intensity of a laser beam, the contact force, and the like.
[0016] It can be provided that the process parameters comprise a constant control variable for a process step, a time profile of a control variable for a process step, a control parameter of a control for a process step and / or a target manipulated variable for a control for a process step.
[0017] Accordingly, such a process parameter can be specified as a constant control variable or a temporal progression of a control variable relating to the same physical quantity. Furthermore, process parameters can also include control parameters of a closed-loop control system in order to achieve optimal control behavior for the manipulated variables to be controlled.
[0018] The goal is to adapt the process parameters as optimally as possible to the current state of the component or its precursor(s), to the ambient conditions, and to the conditions of one or more production devices. These are usually provided using one or more sensor-detected measured variables and / or one or more state variables specified in some other way.
[0019] Process parameters are difficult to adjust easily, particularly in complex processes or with multiple, successive process steps, due to the diverse influencing factors and interactions. Furthermore, physical modeling using a physically motivated model is often not readily possible due to the complexity of the individual process steps. To achieve optimal component quality, it is necessary to coordinate the process parameters, especially when there are multiple process steps. The current approach of having experts perform process modeling and controller design is complex and not reproducible.
[0020] The process parameters, be they control variables, temporal control variable profiles, or control parameters for the control functions executed in the manufacturing process for processing the component, can be optimized in an automated manner according to the above method. For this purpose, the process parameters are mapped to one or more process parameters depending on at least one measured variable comprising a measurable state of the component, the environment, and / or the relevant manufacturing device, and / or depending on at least one state variable comprising a predetermined state of the component, the environment, and / or the relevant manufacturing device.
[0021] The at least one process parameter is determined using a data-based process parameter model that is trained to map the at least one measured variable and / or the at least one state variable to the corresponding at least one process parameter.
[0022] For the training of a data-based model, training data is usually required that assigns a training data point to one or more corresponding labels in the form of, for example, process parameters.
[0023] A fundamental difficulty lies in optimally specifying the process parameters required for the optimized execution of one or more process steps as labels for the various initial situations regarding the properties / condition of the component / precursors to be processed, the properties / condition of the manufacturing equipment, and the environmental conditions. Determining such process parameters for a specific initial situation would be complex, and comprehensive training of the data-based model would require a large number of training data sets for different initial situations.
[0024] In this regard, the above method provides for creating the process parameter model using a quality model. The data-based process parameter model is trained based on the quality model. The quality model maps the at least one measured variable and / or the at least one state variable that describe or characterize the initial situation for the component to be manufactured, and the at least one process parameter that specifies the manner in which the manufacturing process is to be carried out, to a component quality.
[0025] According to one embodiment, the quality model can comprise a data-based model, wherein training data sets are determined for training the quality model, wherein the training data sets are determined by training data points which are determined by varying values of the one or more measured variables and / or the one or more state variables and varying values of the process parameters within respectively predetermined permissible value ranges, with respectively assigned qualities as labels, wherein the qualities each result from at least one property of the manufactured component and / or costs of the manufacturing process with the aid of a quality function, wherein the quality model is trained with the training data sets.
[0026] According to a further embodiment, the quality model can comprise a data-based model, wherein training data sets are determined for training the quality model, wherein the training data sets are formed by training data points which are determined by varying values of the one or more measured variables and / or the one or more state variables and varying values of the process parameters within respectively predetermined permissible value ranges, each with an associated at least one property of the manufactured component and / or costs of the manufacturing process as labels, from which the quality can be determined using a quality function, wherein the quality model is trained with the training data sets.
[0027] The quality model can comprise a physical model, a heuristic or a data-based model and can be designed to evaluate properties of the manufactured component and / or costs of the manufacturing process depending on an initial situation which is specified by the at least one measured variable and / or the at least one state variable and the at least one process parameter which characterises the implementation of the manufacturing process, wherein the quality is determined using the quality function with regard to the properties of the manufactured component and / or the costs of the manufacturing process.
[0028] To determine quality, a quality model can generally be used, which can be designed as a physical model, a heuristic, or a data-based model. The quality model can directly evaluate quality depending on the initial situation, which is specified by the at least one measured variable and / or the at least one state variable and the at least one process parameter that characterizes the execution of the manufacturing process. Alternatively, the quality can evaluate properties of the manufactured component and costs of the manufacturing process according to a quality function. The properties of the manufactured component and / or the costs of the manufacturing process are provided accordingly by the quality model.Properties of the manufactured component include, for example, geometric dimensions and dimensional tolerances, surface properties of the manufactured component, electrical properties of the manufactured component, robustness of the manufactured component, and the like. Regarding the costs of the manufacturing process, tool wear, manufacturing process duration, energy consumption, and material consumption are all factors to consider.
[0029] The quality model can also be designed as a data-based model, e.g., in the form of a neural network, and trained appropriately before using and creating the process parameter model. In particular, the training of the data-based quality model can be based on simulations and / or measurements, with the initial situation and the resulting quality, or the initial situation and one or more component properties of the component determined by the initial situation, each determining a training data set.
[0030] The quality model is now trained based on the training data sets. If the quality model is trained on the output of component properties, a resulting quality can be determined using a predefined quality function. The quality function can essentially be defined based on the component properties mentioned above.
[0031] If the quality model is trained accordingly or otherwise specified, it can be used as the basis for training the process parameter model. The quality model is preferably designed to be differentiable, so that the data-based process parameter model can be trained based on the quality. In other words, the loss function for training the process parameter model corresponds to a maximization of the quality, so that the data-based process parameter model can be easily trained in this way.
[0032] It may be provided that training data points are used by varying the one or more measured variables and / or the one or more predetermined state variables within their respective value ranges in order to train the data-based process parameter model depending on a loss function that is determined by the quality resulting from the application of the quality model.
[0033] By linking the quality model with the process parameter model, the process parameter model can be easily trained. During training, at least one measured variable and / or at least one state variable are varied within their respective possible value ranges in order to account for possible combinations of one or more properties / states of the component or its precursor(s), one or more properties / states of the manufacturing device, and / or one or more environmental conditions.
[0034] The above approach makes it possible to create a process parameter model that considers highly complex relationships within one or between multiple process steps of a manufacturing process and optimizes the quality of a component as the resulting end product by specifying at least one process parameter. Using the quality model and, if necessary, the quality function to train the process parameter model makes it possible to extract knowledge from data without understanding the underlying physical models. The high fault tolerance during training of data-based models also allows for the consideration of noisy measurement and state variables, which, in physically motivated models, would often lead to deviations in the process parameters to be determined and degrade the product quality of the manufactured component.
[0035] The process parameter model can be used prior to the manufacture of a component to parameterize the one or more process steps, in particular depending on at least one sensor-detected measurement variable of a property or state of one or more precursors, a property or state of one or more manufacturing devices for the one or more process steps, and / or one or more environmental conditions using the trained process parameter model. After parameterizing the one or more process steps, the component to be manufactured can be manufactured by executing the parameterized process steps.
[0036] Brief description of the drawings
[0037] Embodiments are explained in more detail below with reference to the attached drawings. They show:
[0038] Figure 1 is a schematic representation of a manufacturing process for
[0039] Manufacturing of a component;
[0040] Figure 2 is a block diagram of a system for providing a
[0041] Process parameter model for carrying out the manufacturing process of Figure 1; and
[0042] Figure 3 is a flowchart illustrating a method for providing a process parameter model.
[0043] Description of embodiments
[0044] Figure 1 schematically shows a manufacturing process 1 with a number of consecutive process steps 2 for producing or further processing a product. The process steps 2 serve to successively process or process one or more precursors V to obtain a component B. For this purpose, at least one manufacturing device 3 is provided for each of the process steps 2, which is equipped with a tool or other device for processing the component B.
[0045] Component B can be a molded component, an electronic component, an assembly, a group with component arrangements, or a finished product. Precursors V include corresponding workpiece blanks, individual parts assembled by process steps 2, such as electronic components, and a printed circuit board in a process for creating a populated printed circuit board, an integrated component and a lead frame, and a housing mold for producing a housed integrated circuit, and the like.
[0046] The process steps can include a variety of machining and processing steps, such as all types of mechanical machining such as drilling, turning, milling, grinding, cutting, punching, sawing and the like, joining techniques such as gluing, bonding, welding, laser welding and the like, joining techniques, thermal treatments, surface treatments, chemical processing, contacting techniques such as bonding and the like.
[0047] The manufacturing process 1 comprises one or more process steps 2, which successively produce the finished component from one or more preliminary products. Each of the process steps 2 is controlled by one or more process parameters P. Process parameters P are understood herein to be parameters that can be specified for the execution of a mechanical, thermal, or chemical process in order to influence at least one property of the component B to be manufactured. Each of the process steps 2 can be characterized by one or more process parameters P, such as temperatures, supplied power, tool or workpiece, speeds or rotational speeds, or also target specifications for controlling a mechanical process and the like.
[0048] The process parameters P can generally include control variables for the processing techniques. The control variables can be specified as constants or can define varying curves of a process parameter, which can be parameterized accordingly. For example, the temporal curves of a model parameter can be specified by several subparameters in the form of time periods and corresponding gradients of the respective process parameter P provided within the time periods. The subparameters and the process parameters P are hereinafter referred to as process parameters P.
[0049] The control variables can also include a target specification for a closed-loop control of a manipulated variable of a production device 3, which is used to guide a tool or other device. Furthermore, the process parameters P can include one or more control parameters of a closed-loop control system in order to parameterize a corresponding control system for one or more manipulated variables. Such a control system can include a PID controller, state controller, adaptive controller, model-based predictive controller, or the like.
[0050] During the manufacture of a component, one or more process parameters are specified and the component is manufactured accordingly while maintaining the relevant process parameters P.
[0051] Figure 2 schematically shows a system for providing a process parameter model 11, with which the individual process steps 2 and the manufacturing devices 3 assigned to the process steps can be preset in order to manufacture a component B from one or more precursors V. In Figure 2, a process parameter model 11 is initially provided which, in the trained state, is intended to determine one or more process parameters P from one or more measured variables M and one or more state variables Z.
[0052] The process parameter model 11 is preferably designed as a data-based model and, in particular, is provided as a deep neural network. The measured variables can correspond to sensor-detected variables and can indicate material properties, geometric properties, chemical properties, and characteristics of the one or more intermediate products. The state variables can indicate states of the intermediate products (materials or material parameters) and / or the production equipment (such as default settings or properties of the production equipment that are known in advance without measurement) and environmental conditions. Overall, the measured variables and state variables provide a description of the system state before the start of the production process.
[0053] Once the process parameter model 11 has been trained, process parameters P are obtained on the output side, which are optimized for controlling the individual process steps 2 of the manufacturing process 1.
[0054] For training the process parameter model 11, its output is coupled to a quality model 12, which maps the one or more measured variables M, the one or more state variables Z, and the resulting model parameters to a quality G of the resulting manufactured component B. The quality model can be configured in a variety of ways depending on the component to be manufactured and the processes used. The quality model 12 can, for example, comprise a physically motivated model, a heuristic model, or a data-based model. The quality G corresponds to an assessment of the quality of the manufactured component and / or the costs of the manufacturing process.
[0055] The quality of the manufactured component is characterized by its properties or usability for the intended purpose and can include: manufacturing tolerances, dimensions, a
[0056] Surface quality, robustness, and the like. The costs of the manufacturing process can include, for example, the use of materials, the duration of the manufacturing process, the energy consumption, and the wear and tear of the manufacturing equipment.
[0057] The quality can be provided directly by the quality model 12 or determined using a quality function 13, which usually combines one or more properties of the manufactured component and / or the costs of the manufacturing process quantified in an evaluation measure of the quality G.
[0058] Figure 3 shows a flowchart illustrating a method for creating the process parameter model 11 according to the function shown in Figure 2. The method is computer-implemented and serves to easily optimize the process parameter model 11 without expert knowledge.
[0059] The core of the procedure is the linking of the process parameter model 11 with the quality model 12.
[0060] For this purpose, in step S1, a manufacturing process 1 is first defined which, as described in connection with Figure 1, comprises a series of process steps 2, each of which can be parameterized by one or more process parameters P. With the aid of a suitable simulation model, a heuristic or measured values, properties of the component thus manufactured and / or the costs of the manufacturing process can be evaluated based on real or virtual process specifications, namely concrete values for the one or more measured variables M, for the one or more state variables Z and for possible process parameters P.
[0061] The evaluation can be based on the actual production of a component and a subsequent measurement or assessment of the manufactured component and / or a determination of the costs of the manufacturing process, taking into account the duration of the manufacturing process, wear and tear, material expenditure, and energy expenditure. This results in a quality G in each case according to the specified quality function 13. The process specifications M, Z, P and the quality as a label define training data sets in step S2. To determine training data sets, the process specifications M, Z, P are varied, and the corresponding resulting quality G is determined as a label. The process specifications are varied in such a way that the input data space for the quality model is represented as completely as possible. Alternatively, the process specifications can also be varied based on domain knowledge.
[0062] In this way, the training data points are labeled and trained for the value ranges of the one or more measured variables, the one or more state variables and the one or more model parameters that are realistic in practice.
[0063] The training data sets can be used in step S3 to parameterize a physical quality model 12 or to train a data-based quality model 12. Alternatively, the quality G can also be derived directly from the corresponding process specifications M, Z, P by applying a heuristic.
[0064] A deep neural network is used as the data-based quality model 12, which can be trained in a conventional manner using a gradient-based method. The data-based quality model 12 can, in particular, be trained with qualities G resulting from measurement or simulation methods relating to the component B to be manufactured or manufactured. This results in a data-based quality model 12 that can reliably determine a quality G of the resulting component B in a large input data space. The quality model 12 is preferably selected such that it is differentiable, so that the resulting quality G can be used in a loss term for training the process parameter model 11. In particular, the quality G can be considered directly as a loss, or an inverse of the quality G can be considered as a loss.
[0065] A heuristic can combine the corresponding process specifications M, Z, P in a rule-based manner to directly obtain a quality G of the component B to be manufactured.
[0066] In step S4, the process parameter model 11 is trained in a corresponding manner by varying input variables for the process parameter model 11, wherein the one or more measured variables M and the one or more state variables Z are each varied over their entire value range in order to completely cover the input data space for the process parameter model 11. This results in process parameters P which, together with the one or more measured variables M and the one or more state variables Z, define the process specifications and are evaluated for evaluation by the quality model 12 and, if applicable, by the quality function 13.
[0067] The resulting quality factor G now provides the loss for training the network parameters of the data-based process parameter model 11. This loss can be used in a conventional manner within a gradient-based training procedure. Based on this, the process parameter model 11 can be trained automatically based on a large number of combinations of varying measured variables M and / or state variables Z, using an evaluation by the quality model 12.
[0068] The process parameter model 11 thus modeled can now be used for the manufacturing process described above. To do this, before the start of manufacturing a component B, the measured variables M and the state variables Z are recorded, and the corresponding process parameters P are determined using the process parameter model. These process parameters P are then implemented in the manufacturing devices 3 of the individual process steps.
[0069] An example of a manufacturing process is the bonding process of a bond wire to a contact surface. During the bonding process, a bond wire is pressed onto the surface of the contact surface using an ultrasonic head. The contact force and the transmitted vibrations bond the bond wire to the surface. This can be used, for example, to electrically connect electronic components to the circuit board. With regard to the above definitions, the measured variables correspond to surface roughness, contamination, etc.; the state variables correspond to a material, a component geometry (with regard to possible vibrations and resonances); the process parameters correspond to a contact force, a bonding time, an ultrasonic amplitude; and the quality corresponds to an evaluation criterion that takes into account a connection cross-section, which has a significant influence on the load-bearing capacity of the connection, and deformation of the bond wire.
[0070] Another example of a manufacturing process is laser welding. In laser welding, a focused laser beam is used to locally melt metal. The molten metal firmly bonds two joining parts. The laser beam is either absorbed (absorbed) or reflected by the metal. Only the absorbed power is available for melting.
[0071] With reference to the above definitions, the measured variables can be a surface absorbency, a surface roughness, contamination and the like, the state variables can be information about a workpiece material (e.g. copper / aluminum / alloy type), a component geometry, a type of laser welding machine, a type of laser optics, the process parameters can be a laser wavelength, a focus diameter, a laser power, a feed rate and the like, and the quality can be an evaluation measure that takes into account a mechanical load capacity of the seam, a flatness of the seam, a welding depth, a process stability (amount and size of weld spatter, pores or holes) and the like. A semiconductor production process can be given as another example of a manufacturing process. In semiconductor manufacturing, many individual process steps are linked together.There are a multitude of relevant variables, although the concrete relationships are not always immediately apparent. In one process step, the wafer (the semiconductor component) is provided with a coating. With regard to the above definitions, the measured variables can be a wafer condition (e.g. defect density, inline measurements from previous processes, e.g. measurement of layer thicknesses) and the like, the state variables can be information about the condition of the chambers (number of operating hours) and the like, the process parameters can be an aging temperature, O2 quantity, etchant quantity, performance of the CVD (chemical vapor deposition), PVD (physical vapor deposition) or spin coating process and the like, and the quality can be an evaluation measure that takes into account layer thickness, edge steepness of the coating after etching, dimensional accuracy after resist processing and the like.
Claims
Claims 1. Computer-implemented method for providing a process parameter model (11) for parameterizing one or more process steps (2) of a manufacturing process (1) for producing a component (B), comprising the following steps: Providing (S3) a quality model (12) for determining a quality (G), wherein the quality model (12) is designed to specify the quality (G) of the resulting component (B) directly or by means of a predefined quality function (13), depending on one or more predefined measurement variables (M) and / or one or more predefined state variables (Z), each of which specifies a property of a pre-product or intermediate product of the component (B) to be manufactured and / or of a manufacturing device (3) for carrying out a process step (2) and / or at least one environmental condition, and on one or more process parameters (P) that control a corresponding of the process steps (2); Training (S4) of a data-based process parameter model (11) to output one or more process parameters (P) depending on one or more sensor-acquired measured variables (M) and / or on one or more predefined state variables (Z) by optimizing the quality (G).
2. Method according to claim 1, wherein the process parameters (P) comprise a constant control variable for a process step (2), a time profile of a control variable for a process step (2), a control parameter of a control for a process step (2) and / or a setpoint control variable for a control for a process step (2).
3. A method according to claim 1 or 2, wherein the quality model (12) comprises a physical model, a heuristic, or a data-based model and is configured to adapt to an initial situation defined by the at least one measured variable (M) and / or at least one state variable (Z) and at least one process parameter (P) that characterizes the execution of the manufacturing process (1) is specified, to evaluate the properties of the manufactured component (B) and / or the costs of the manufacturing process (1), whereby the quality (G) is determined using the quality function (13) with respect to the properties of the manufactured component (B) and / or the costs of the manufacturing process (1).A method according to claim 1 or 2, wherein the quality model (12) comprises a data-based model, wherein training data sets are determined for training the quality model (12), the training data sets being determined by training data points, which are determined by varying values of one or more measured variables (M) and / or one or more state variables (Z) and varying values of the process parameters (P) within predefined permissible value ranges, with each assigned quality as a label, wherein the quality values (G) are each derived from at least one property of the manufactured component (B) and / or costs of the manufacturing process (1) using a quality function (13), wherein the quality model (12) is trained with the training data sets.A method according to claim 1 or 2, wherein the quality model (12) comprises a data-based model, wherein training data sets are determined for training the quality model (12), the training data sets being defined by training data points that are determined by varying values of one or more measured variables (M) and / or one or more state variables (Z) and varying values of the process parameters (P) within predefined permissible value ranges, each with an associated at least one property of the manufactured component (B) and / or costs of the manufacturing process (1) as labels, from which the quality (G) can be determined using a quality function (13), wherein the quality model (12) is trained with the training data sets. A method according to any one of claims 3 to 5, wherein the properties of the manufactured component are geometric dimensions including dimensional tolerances, and a surface finish of the manufactured component. The method comprises the component, an electrical property of the manufactured component, and the robustness of the manufactured component, wherein the costs of the manufacturing process include tool wear, the duration of the manufacturing process (1), the energy expenditure of the manufacturing process (1), and material expenditure. A method according to any one of claims 1 to 6, wherein training data points are used by varying one or more measured variables (M) and / or one or more predetermined state variables (Z) within their respective value ranges to train the data-based process parameter model (11) depending on a loss function determined by the quality (G) resulting from the application of the quality model (12).A method according to any one of claims 1 to 6, wherein the process parameter model (11) is used prior to the manufacture of a component (B) to parameterize one or more process steps (2), in particular depending on at least one sensor-detected measured variable (M) of a property or state of one or more intermediate products, a property or state of one or more manufacturing devices (3) for the one or more process steps (2), and / or one or more environmental conditions, using the trained process parameter model (11). A device for carrying out one of the methods according to any one of claims 1 to 8. A computer program product comprising instructions which, when the program is executed by at least one data processing device, cause it to execute the steps of the method according to any one of claims 1 to 8.Machine-readable storage medium comprising instructions which, when executed by at least one data processing device, cause it to execute the steps of the method according to any one of claims 1 to 8.
