Method and device for performing a screwing-in process by means of artificial intelligence methods
By controlling the tightening process of self-tapping screws using a data-based parametric model, the problem of controlling the preload during screw insertion was solved, achieving precise preload control and optimization of screw design.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- ROBERT BOSCH GMBH
- Filing Date
- 2021-09-16
- Publication Date
- 2026-07-03
Smart Images
Figure CN114193141B_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to a method for performing a mechanical screwing process, and more particularly to a method for performing a self-tapping screwing process. Technical Background
[0002] In the case of self-tapping screws, the screw is screwed into the material of the mating partner, particularly a material with pre-drilled holes. Upon initial screwing, the self-tapping screw cuts mating threads into the material of the mating partner. Typically, the self-tapping screw is loosened again after the first tightening, and then finally tightened to a predetermined torque. During the first tightening, the screw head presses against the material of the mating partner with its jaws, while the torque is limited to a maximum torque during the second tightening.
[0003] Physical models known from existing technologies, such as Senevilatne, LD et al., “Theoretical modelling of self-tapping screw fastening process,” Proceedings of the Institution of Mechanical Engineers, Part C: Journal of Mechanical Engineering Science, pp. 135-154, are used to determine the tightening torque allowed to be used when finally tightening the self-tapping screw. However, the formulas described there cannot be applied to altered conditions, such as different materials. Summary of the Invention
[0004] According to the present invention, a method for performing a screwing-in process according to claim 1 and an apparatus according to the parallel independent claims are provided.
[0005] Other designs are described in the dependent claims.
[0006] According to the first aspect, a computer-implemented method is provided for performing a screwing process, particularly a self-tapping screwing process, using a pre-given preload, comprising the following steps:
[0007] - Screw the screw into the workpiece, especially until the pre-given first maximum tightening torque is reached;
[0008] - Plot the change in tightening torque during the tightening process;
[0009] -The second maximum turning torque is determined, in particular, by means of at least one trained data-based parametric model, based on the turning torque variation process and a pre-given preload.
[0010] - Screw the screw in using a screwing torque limited to the maximum screwing torque.
[0011] Specifically, the characteristics of the turning torque variation process can be determined based on the turning torque variation process, wherein at least one data-based parameter model is trained to provide turning process parameters based on the turning torque variation process characteristics, wherein the maximum turning torque is determined based on the at least one turning process parameter and a pre-given preload.
[0012] During the self-tapping process, the self-tapping screw is first screwed into a pre-drilled hole in the material, whereby the screw cuts its own mating threads. The screw is then loosened approximately one-quarter to one full turn and finally tightened to a pre-given maximum torque. Once the screw jaws abut against the mating partner, establishing a preload between the screw head and the mating partner, the applied tightening torque increases dramatically.
[0013] The preload acting between the screw head and the mating partner is not directly measurable because sensors suitable for measurement in mass production are difficult or impossible to implement. Furthermore, due to manufacturing tolerances of the screw, pre-drilled hole, and workpiece material, calibrating the preload while keeping the parameters constant during tightening is challenging, as such calibration involves significant static uncertainty. However, the greater the maximum tightening torque set during screwing, the greater the expected value of the axial preload achieved.
[0014] One way to ensure a specific minimum preload for a screw-in connection is to adequately determine the magnitude of the tightening torque applied when finally tightening the screw, such that even within a range of statistical uncertainty, the minimum preload remains higher than the required minimum preload. This approach is only suitable if the desired minimum preload is on the order of the screw's load limit, and the variation in the set preload is correspondingly small. Therefore, self-tapping screws used in manufacturing are typically designed such that the minimum preload to be set is sufficiently far from the load limit. Consequently, in many cases, screws are oversized or a large number of screws are used, each absorbing a small preload, negatively impacting product weight and cost.
[0015] At least one tightening process parameter can define a pre-given physical tightening model, which is designed to output the maximum tightening torque based on a pre-given preload. Therefore, the above method specifies the use of a physical tightening model to determine the maximum tightening torque, particularly for the second tightening, by means of one or more data-based probabilistic parameter models, which can be constructed, for example, as regression models, particularly Gaussian process models. To this end, a tightening torque curve is first plotted during one or more preceding tightening stages, performed in the particularly second tightening stage before the final tightening of the screw. This tightening torque curve includes the variation of the torque applied at the screw's rotation angle or the variation of the torque applied during the tightening time, particularly until the end of the first tightening.
[0016] Features of the tightening torque curve are extracted, and one or more tightening model parameters of the physical tightening model are determined from one or more probabilistic parameter models. The tightening model parameters determined in this way are used in the physical tightening model to calculate the maximum tightening torque required for the second tightening under a pre-given desired preload.
[0017] Using a data-based probabilistic parametric model allows for the evaluation of a large series of measurements of the tightening torque during one or more previous tightening stages, in order to determine the tightening model parameters that can be used for the second tightening stage by means of a suitable physical model that can be applied to the previous tightening stages.
[0018] According to another aspect, a computer-implemented method is provided for training at least one data-based parametric model to provide at least one twisting model parameter for a physical twisting model, wherein the physical twisting model describes the relationship between preload and twisting torque based on at least one twisting model parameter, comprising the following steps:
[0019] - Screw multiple screws into the workpiece, especially until the pre-given first maximum tightening torque is reached;
[0020] - Plot the change process of the twisting torque;
[0021] - For each twisting torque variation process, in particular by fitting or compensation calculation, at least one twisting model parameter is determined by at least one other physical twisting model, the physical twisting model being defined by at least one twisting model parameter, in order to obtain a corresponding training dataset, the training dataset assigning the twisting torque variation process to the determined at least one twisting model parameter;
[0022] - Train at least one data-based parametric model based on the corresponding training dataset.
[0023] Furthermore, the at least one data-based parametric model can be trained to assign the characteristics of the twisting torque variation process to the twisting model parameters, wherein the characteristics of the twisting torque variation process are determined based on the twisting torque variation process.
[0024] To train a data-based parametric model, the twisting model parameters are determined by fitting or compensating the twisting torque curves to the physical twisting model applicable to the corresponding twisting stage, based on the twisting torque variation process in each previous twisting stage. Subsequently, twisting torque variation process features are extracted from the twisting torque variation process in the previous twisting stages. These features may include, for example, the absolute value of the twisting torque at one or more specific moments or rotation angles, the gradient of the twisting torque at one or more specific moments or rotation angles, and the integral value of the twisting torque within a specific time window or rotation angle range. The parametric model is trained based on a training dataset, which includes the twisting torque variation process features and one or more assigned twisting model parameters. Individual parametric models can be trained separately for multiple twisting model parameters.
[0025] As additional input variables for the data-based parametric model, corresponding tightening process stages from which the characteristics of the tightening torque variation process can be extracted can be considered. These include, for example, the preparation stage where the screw tip is inserted into the hole and the thread is not yet engaged; the screwing-in stage where the thread engages with the inner wall of the hole and cuts out the internal thread; the first tightening stage where the screw head rests on the mating partner and applies preload; the reverse rotation stage during which the screw is unscrewed again; and the second screwing-in and second tightening stages. Therefore, the tightening torque variation process is divided temporally so that regions of the tightening torque curve can be assigned to the corresponding tightening process stages.
[0026] In this way, even if the screw and mating partner configurations are different, customized process guidance can be performed in a controlled manner during multi-stage screwing.
[0027] However, the above method is not limited to a two-stage screwing process, but can also be applied to a screwing process with only one tightening stage. Attached Figure Description
[0028] The embodiments are explained in more detail below with reference to the accompanying drawings. Wherein:
[0029] Figure 1 A schematic diagram of a self-tapping screw system for screwing self-tapping screws into a workpiece is shown;
[0030] Figure 2 A graph is shown to illustrate the variation of the screwing torque during different screwing stages of the screwing process into the workpiece;
[0031] Figure 3 It shows the illustration in Figure 1 A flowchart of the method for performing the self-tapping screwing process in a screwing system; and
[0032] Figure 4 A flowchart illustrating a method for training at least one parametric model to determine the parameters of the twisting process in a physical twisting model is shown. Detailed Implementation
[0033] Figure 1 A schematic diagram of a screwing system 1, as part of a product manufacturing system, is shown. The screwing system 1 has a screwing device 2 that grips a screw 3 in a manner known per se and screws the screw into a workpiece 4 by applying a screwing torque. Preferably, the workpiece 4 has a pre-drilled hole 5, the diameter of which is smaller than the diameter of the threaded portion of the screw 3. Specifically, the diameter of the pre-drilled hole can be larger than the diameter of a screw without a threaded flange, but smaller than the diameter of a screw including the threaded flange.
[0034] Screw 3 is typically screwed in according to a so-called self-tapping screwing process, which is executed by control unit 6 in a corresponding manner by operating screwing device 2.
[0035] The self-tapping screwing process takes place in multiple screwing stages, which are in Figure 2 The diagram is illustrated exemplarily. The tightening process illustrated herein is performed in two stages, with a first tightening and a final second tightening. However, this method can also be used without limitation for a one-stage tightening process. The changes in tightening torque T and preload F over time are shown in... Figure 2 As shown in the image.
[0036] The screwing stage includes: a preparation stage e, in which the screw tip 33 is inserted into the hole 5 and the thread 34 is not yet fully engaged; a screwing-in stage a, in which the thread 34 engages with the inner wall of the hole 5 and cuts out the internal thread; a first tightening stage t, in which the screw 3 is tightened with a predetermined (first) maximum screwing torque, such that the screw head 31 abuts against the workpiece 4 with its jaws 32 and applies a preload; a reverse rotation stage r, during which the screw 3 is screwed out one-quarter to one full turn; a second screwing-in stage a*; and a second tightening stage t*, in which the screw 3 is tightened with a predetermined maximum screwing torque.
[0037] It can be seen that during the first tightening stage, once the lower jaw 32 of the screw head 31 abuts against the workpiece 4, the preload increases. Simultaneously, the applied tightening torque T also increases sharply, limited by the maximum tightening torque T pre-given for the first tightening. During the reverse rotation stage, the screw 3 rotates, the preload F decreases to zero again, and the applied tightening torque T becomes negative.
[0038] In this embodiment, the two-stage process of tightening screw 3 is used to more accurately determine the screwing model parameters, which are used to determine the screwing torque to be applied for the defined preload, so that the preload between workpiece 3 and screw head 31 can be set more accurately.
[0039] The various stages of the twisting process correspond to different functions, where the overall twisting torque T is defined as...
[0040]
[0041] Each tightening stage i in the tightening process e, a, t, r, a*, t* is defined by a starting angle and a ending angle. In the reverse rotation stage r, the ending angle is smaller than the starting angle because screw 3 rotates in the opposite direction here. It can be seen that although the underlying tightening models differ, the preload F and rotation angle ϕ are determined by the same parameters. , Definition. Tightening process parameters. These are model parameters that can vary for each tightening process, such as the diameter of the pre-drilled hole 5 or the energy required to rotate screw 3 from rotation angle 41 to rotation angle 42. These are parameters that do not change within a series of rotations and therefore their values must be determined only once for a series. For example, model parameters. This may include the relevant material properties of screw 3 and / or the workpiece.
[0042] Physical twisting model T t* This should be used for the second tightening to set a specified preload F when the preload cannot be measured on the production line. The maximum required tightening torque T is calculated using a tightening model for this purpose. max Then tighten screw 3 accordingly until the tightening torque T (in Nm) is reached when the screw stops rotating.
[0043] In principle, this method can also be used for a one-stage tightening process, where only the tightening stages e, a, and t are considered. The tightening stage here represents the final tightening of the screw. Here, the required tightening model parameters are detected by observations in the preparation and screwing-in stages and are used to determine the tightening torque T to be applied in the (first) final tightening stage.
[0044] exist Figure 3 The flowchart illustrates the process that can be implemented as software or hardware in the screwing system 1. The method specifies that the screw is screwed into the workpiece so that the screw head 31 presses against the workpiece 4 with a defined preload F (in N).
[0045] In step S1, the desired preload force F is described accordingly.
[0046] Then, the screwing process begins in step S2. For this purpose, screw 3 rotates at a predetermined speed or a pre-given speed variation and is placed on the pre-drilled hole 5 with a defined pre-given pressure. In engagement phase e, the thread of screw 3 is not yet fully engaged in the pre-drilled hole 5 and seeks an engagement point to cut into the material of workpiece 4. Then, screw 3 is screwed in during screwing phase a. Here, the thread of screw 3 cuts into the wall of the pre-drilled hole 5 and thus forms a thread.
[0047] The subsequent first tightening phase t involves the process from placing the screw head on the workpiece and continues until the first predetermined maximum tightening torque T is reached. max Up to this point. The first pre-defined maximum turning torque T max The maximum tightening torque selected corresponds to a safe distance from the following tightening torque, at which the screw head may or will definitely tear. Additionally, the first predetermined maximum tightening torque should be greater than the second predetermined maximum tightening torque T used for the subsequent second tightening. max Therefore, thread cutting will not occur again when tightening a second time.
[0048] The change in tightening torque T was detected during the tightening process.
[0049] In step S3, the characteristics of the turning torque change process are determined from the turning torque change process of one or more turning stages e, a, t. The characteristics of the turning torque change process include, for example, determining the absolute value of the turning torque at one or more specific moments, the gradient of the turning torque at one or more specific moments, and the integral value of the turning torque within a specific time window during the turning stages e, a, t.
[0050] In step S4, at least one screwing model parameter is now determined from the screwing torque characteristics using at least one data-based parametric model. The screw-tightening model parameters correspond to the parameters of the physical screw-tightening model and represent variable variables that may change during the screw-tightening process, such as the diameter of the pre-drilled hole or the angle of rotation of the screw. φ 1. Start rotating to rotation angle φ 2. Energy required.
[0051] If the corresponding screwing model parameters are determined, the screw 3 is further screwed in during the reverse rotation phase r and the second screwing-in phase a* in step S5.
[0052] To determine the maximum tightening torque T for the second tightening max Using at least one screwing model parameter previously determined from a data-based parametric model. To determine the physical tightening model T for the second tightening t* .
[0053] Then, in step S6, during the second tightening phase, the corresponding maximum tightening torque T is used for the second tightening. max Tighten screw 3.
[0054] For example, the screwing model can correspond to the model in the publication mentioned at the beginning, as shown below.
[0055]
[0056] Where p corresponds to the screw pitch, μ1 corresponds to the coefficient of friction between the screw and the hole, and D... s D corresponds to the maximum diameter of the screw. h For the diameter of the hole, μ2 corresponds to the coefficient of friction between the screw head and the workpiece, and D... sh D corresponds to the diameter of the screw head. n This corresponds to the diameter of the unthreaded screw. These variables correspond to the screwing model parameters. .
[0057] The preparation stage e typically corresponds to three turns, during which the threads in the tapered inlet region of screw 3 engage. Then, the tightening stage a begins. Each tightening stage e, a, and t can be identified from the change in tightening torque. Here, the transition between preparation stage e and tightening stage a can be determined by increasing the tightening torque by a gradient higher than a pre-defined first tightening torque gradient threshold. The transition from tightening stage a to the first tightening stage t can be determined by correspondingly increasing the applied tightening torque by a gradient higher than a pre-defined second tightening torque gradient threshold.
[0058] Parametric models are data-based probabilistic models, particularly regression models, such as Gaussian process models. This is used to analyze the twisting torque curve. The parameters of the screwing model are evaluated by the prediction function in the screwing torque feature set. One of them, output the Gaussian process. Expected value μ and variance Here C N Let the covariance matrix be denoted by . Given, where x n and x m It is a combination of features already incorporated into the parametric model. The variable β represents the precision (reciprocal of the variance) of the normal distribution, which represents experimental reproducibility with the same feature set. δ nm It is the Kronecker symbol.
[0059] Scalar c is usually composed of Given. Vector t contains vectors for each feature set x. iThe corresponding result. Vector k contains the values of the kernel function, which encodes the features x. n The result of the combination is related to feature x m The results contain information about the level of influence. Here, a large value represents a high level of influence. If the value is 0, there is no influence. To predict the expected value μ and variance σ in the above formula, we need to calculate the expected value μ and variance σ from all feature sets x. i (i=1...N) and the feature set to be studied x N+1 Calculate k. For specific cases, an exponential kernel is preferred as the kernel function to use.
[0060]
[0061] The exponential kernel has optional hyperparameters Θ0 and Θ1. Θ1 is related to the feature set x. n and x m The distance between the function values at time Θ1 has a decisive effect, because for large Θ1 values, the function tends to zero.
[0062] From this model, we can obtain each twisting torque curve and the parameters used for the corresponding twisting model. Twisting torque characteristics x N+1 Determine the expected value E(v) N+1 )=µ(x N+1 ) and variance Var(v N+1 )=σ²(x N+1 ).
[0063] The training of the parametric model can be based on... Figure 4 The flowchart illustrates the method execution.
[0064] Therefore, in step S11, the turning torque change process is determined by the turning torque change process before the second tightening stage, especially the turning torque change process during the engagement stage e, the first screwing-in stage a, and the first tightening stage t.
[0065] In step S12, the twisting torque change process is fitted to the physical model on which the corresponding twisting stage is based, one by one, in order to determine at least a portion of the twisting model parameters. These screwing model parameters are required for the physical model of the second tightening stage.
[0066] This allows the variation of the twisting torque to be assigned to the physical models of different twisting stages, for example:
[0067] In step S13, a torque change process feature is assigned to each torque change process. The torque change process feature can correspond to the gradient, integral, and absolute value on the selected angle segment, respectively.
[0068] In step S14, a separate parameter model is trained for each determined screwing model parameter so as to assign the screwing torque variation process characteristics of the screwing process to the corresponding screwing model parameter.
[0069] Therefore, the training of the parameter model can be performed based on the characteristics of the twisting torque change process and the twisting model parameters, which are generated based on the basic physical model of each twisting stage.
Claims
1. A computer-implemented method for performing the screwing process using a pre-given preload F, comprising the following steps: - Screw the screw (3) into the workpiece (4) until the first maximum screwing torque is reached; - Plot the change in tightening torque during the tightening process; - Using at least one trained data-based parametric model, the second maximum turning torque T is determined based on the turning torque variation process and a pre-given preload F. max ; -Use is limited to the second maximum turning torque T max Tightening torque T is used to screw the screw in (3); The characteristics of the turning torque change process are determined based on the turning torque change process, wherein at least one data-based parameter model is trained to provide turning process parameters based on the turning torque change process characteristics, and the second maximum turning torque T is determined based on the at least one turning process parameter and a pre-given preload F. max .
2. The method according to claim 1, wherein the screwing process is a self-tapping screwing process.
3. The method according to claim 1, wherein one or more of the following variables are determined from the twisting torque change process as characteristics of the twisting torque change process: the absolute value of the twisting torque at one or more specific moments or rotation angles, the gradient of the twisting torque at one or more specific moments or rotation angles, and the integral value of the twisting torque within a specific time window or a specific rotation angle range.
4. The method according to any one of claims 1 to 3, wherein the at least one data-based parametric model comprises a regression model.
5. The method according to claim 4, wherein the regression model is a Gaussian process model.
6. The method according to any one of claims 1 to 3, wherein the at least one twisting process parameter defines a pre-given physical twisting model, the physical twisting model being designed to output the second maximum twisting torque T according to a pre-given preload F. max .
7. A computer-implemented method for training at least one data-based parametric model to provide at least one twisting model parameter for a physical twisting model, wherein the physical twisting model describes the relationship between preload and twisting torque based on at least one twisting model parameter, comprising the following steps: - Screw multiple screws into the workpiece until the pre-given first maximum tightening torque is reached; - Plot the change process of the twisting torque; - For each twisting torque change process, at least one twisting model parameter is determined by at least one other physical twisting model, the physical twisting model being defined by at least one twisting model parameter, in order to obtain a corresponding training dataset, the training dataset assigning the twisting torque change process to the determined at least one twisting model parameter; - Train at least one data-based parametric model based on the corresponding training dataset; - Use at least one data-based parametric model during the screwing process.
8. The method of claim 7, wherein at least one twisting model parameter is determined by fitting or compensation calculation.
9. The method of claim 7 or 8, wherein the at least one data-based parametric model is trained to assign a twisting torque variation process feature to the twisting model parameters, wherein the twisting torque variation process feature is determined based on the twisting torque variation process.
10. An apparatus for performing the method according to any one of claims 1 to 9.
11. A computer program product, comprising instructions that, when executed by a computer, cause the computer to perform the steps of the method according to any one of claims 1 to 9.
12. A machine-readable storage medium comprising instructions that, when executed by a computer, cause the computer to perform the steps of the method according to any one of claims 1 to 9.