Intelligent mechanical arm bone grinding mechanical model construction method and system

Abstract

The invention discloses an intelligent mechanical arm bone grinding mechanical model construction method and system, and the method comprises the steps: S1, collecting bone grinding experiment data under the conditions of different bone mineral densities, main shaft rotating speeds and feeding speeds through a design experiment, carrying out the preprocessing of the bone grinding experiment data, and obtaining a bone grinding experiment model; a mechanical arm bone grinding data set containing grinding parameters and corresponding grinding force is formed, and the grinding parameters comprise bone density, the main shaft rotating speed of a spherical grinding drill on the intelligent mechanical arm and the feeding speed; and S2, on the basis of the intelligent mechanical arm bone grinding data set, a bone grinding mechanical model is constructed through a radial basis function neural network RBNN. According to the method, the complex nonlinear relation between the grinding parameters and the grinding force can be effectively captured, and powerful support is provided for optimization and control of the robot bone grinding process.

Description

Technical Field

[0001] The present invention relates to the field of bone grinding mechanical model construction, and in particular to a method and system for constructing a bone grinding mechanical model of an intelligent robotic arm. Background Art

[0002] Bone grinding is a common procedure in modern orthopedic surgery. With the development of orthopedic robotic technology, surgical robots have shown great advantages in bone grinding surgery. However, the current robot grinding accuracy is insufficient and the bone grinding force cannot be accurately controlled. If the grinding force is too large, it will cause mechanical damage to the bone tissue. Therefore, it is crucial to construct a mechanical model of the grinding force based on grinding parameters such as spindle speed and feed rate. At present, the research on the construction of mechanical models for bone grinding is mainly divided into two aspects: theoretical analysis and data-driven modeling. Theoretical analysis focuses on the cutting edge modeling and mechanical behavior analysis of spherical drill burrs, which makes the mathematical representation of the grinding force too complex and difficult to apply to subsequent grinding parameter optimization. Data-driven modeling focuses on the collection of experimental data and the construction of mathematical representation models. Since the grinding process is a complex multi-physics field coupling problem, it is difficult to construct a high-quality data set for bone grinding. In addition, there is a complex nonlinear mapping relationship between grinding parameters and grinding force, which makes it difficult for empirical models to accurately represent the mapping relationship between grinding parameters and grinding force. With the development of artificial intelligence technology, neural networks have shown significant advantages in fitting nonlinear relationships. Their multi-layer structure and nonlinear activation function can effectively capture complex input-output mappings, and show great application potential in the field of bone grinding mechanical model construction. Summary of the Invention

[0003] The purpose of the present invention is to provide a method and system for constructing a bone grinding mechanics model of an intelligent robotic arm, aiming to solve the problem of constructing a bone grinding mechanics model.

[0004] The present invention provides a method for constructing a mechanical model of bone grinding of an intelligent robotic arm, comprising:

[0005] S1. Collect bone grinding experimental data under different bone density, spindle speed, and feed speed conditions through designed experiments, and preprocess the bone grinding experimental data to form a robotic arm bone grinding dataset containing grinding parameters and corresponding grinding forces. The grinding parameters include: bone density, spindle speed of the spherical drill on the intelligent robotic arm, and feed speed;

[0006] S2. Based on the intelligent robotic arm bone grinding dataset, a bone grinding mechanical model was constructed using the radial basis function neural network (RBNN).

[0007] The present invention also provides an intelligent robotic arm bone grinding mechanical model construction system, comprising:

[0008] The dataset module is used to collect bone grinding experimental data under different bone density, spindle speed, and feed speed conditions through designed experiments, and preprocess the bone grinding experimental data to form a robotic arm bone grinding dataset containing grinding parameters and corresponding grinding forces. The grinding parameters include: bone density, spindle speed of the spherical drill on the intelligent robotic arm, and feed speed;

[0009] A construction module is used to construct a bone grinding mechanical model based on the intelligent robotic arm bone grinding dataset using a radial basis function neural network (RBNN).

[0010] By adopting the embodiment of the present invention, the method can effectively capture the complex nonlinear relationship between grinding parameters and grinding force, and provide strong support for the optimization and control of the robot bone grinding process.

[0011] The above description is only an overview of the technical solution of the present invention. In order to more clearly understand the technical means of the present invention, it is implemented in accordance with the contents of the specification, and in order to make the above and other purposes, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are specifically listed below. BRIEF DESCRIPTION OF THE DRAWINGS

[0012] In order to more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the specific embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0013] Figure 1 is a flow chart of a method for constructing a mechanical model of bone grinding of an intelligent robotic arm according to an embodiment of the present invention;

[0014] Figure 2 Schematic diagram of the grinding path of the method for constructing a mechanical model of bone grinding of an intelligent robotic arm according to an embodiment of the present invention;

[0015] Figure 3 Schematic diagram of the grinding force components in the X, Y, and Z directions of the method for constructing a mechanical model of bone grinding of an intelligent robotic arm according to an embodiment of the present invention;

[0016] Figure 4 The bone density provided by the present invention is 480kg / m 3 Schematic diagram of the influence of spindle speed and feed rate on grinding force under the condition of ;

[0017] Figure 5 Schematic diagram of the effects of spindle speed and bone density on grinding force under a feed rate of 2 mm / s provided in an example of the present invention;

[0018] Figure 6 Schematic diagram showing the effects of feed rate and bone density on grinding force when the spindle speed is 4000 rpm provided in an example of the present invention;

[0019] Figure 7 Schematic diagram of the grinding force resultant signal, filtering and smoothing signal of the method for constructing a mechanical model of bone grinding of an intelligent robotic arm provided by an example of the present invention;

[0020] Figure 8 Schematic diagram of the radial basis function neural network structure of the method for constructing the mechanical model of bone grinding of an intelligent robotic arm provided by an example of the present invention;

[0021] Figure 9 This is a schematic diagram of the fitting of the training set after the neural network training of the intelligent robotic arm bone grinding mechanical model construction method provided by the example of the present invention is completed;

[0022] Figure 10 This is a schematic diagram of the fitting of the test set after the neural network training of the method for constructing the mechanical model of bone grinding of an intelligent robotic arm provided by an example of the present invention is completed;

[0023] Figure 11 Schematic diagram of a system for constructing a mechanical model of bone grinding of an intelligent robotic arm according to an embodiment of the present invention.

[0024] Description of reference numerals:

[0025] 1110: Dataset module; 1120: Construction module. DETAILED DESCRIPTION

[0026] The following will clearly and completely describe the technical solutions of the present invention in conjunction with the embodiments. Obviously, the embodiments described are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0027] Method Example

[0028] According to an embodiment of the present invention, a method for constructing a mechanical model of bone grinding of an intelligent robotic arm is provided. Figure 1 FIG. 1 is a flow chart of a method for constructing a mechanical model of bone grinding of an intelligent robotic arm according to an embodiment of the present invention. Figure 1 As shown, specifically including:

[0029] S1. Collect bone grinding experimental data under different bone density, spindle speed, and feed speed conditions through designed experiments, and preprocess the bone grinding experimental data to form a robotic arm bone grinding dataset containing grinding parameters and corresponding grinding forces. The grinding parameters include: bone density, spindle speed of the spherical drill on the intelligent robotic arm, and feed speed;

[0030] The S1 specifically includes: adding a force sensor to the intelligent robotic arm, arranging and combining the three grinding parameters of bone density, spherical drill spindle speed and feed speed to obtain a grinding parameter group, performing intelligent robotic arm milling based on the parameters in the grinding parameter group, recording milling data of different bone densities at multiple sampling time points through the force sensor, obtaining a resultant force signal corresponding to the bone density based on the milling data of different bone densities, processing the resultant force signal, calculating the average value of the resultant force for each bone density at multiple sampling time points, and constructing a bone grinding data set including the average value of the resultant force of a certain bone density, the corresponding bone density, the spindle speed corresponding to the average value, and the feed speed corresponding to the average value.

[0031] In this embodiment, the specific implementation process of S1 is as follows:

[0032] Based on the UR10 robot arm with a repeatability of 0.1mm, a six-axis force sensor is installed, and a spherical drill with a diameter of 5mm, 14 blades, and a helix angle of 35 degrees is selected. 3 , 800kg / m 3 、1640kg / m 3 The solid rigid polyurethane foam of SAWBONES with three densities simulated cancellous bone, dense cancellous bone and cortical bone for bone grinding experiments. All combinations of grinding parameters were arranged into an experimental matrix as shown in Table 1, where the bone density was 480 kg / m 3 , 800kg / m 3 、1640kg / m 3 , the spindle speed range is 3000~5000rpm, and the feed speed range is 1.5~5mm / s. Based on this, the grinding experiment was carried out. In each experimental run, the grinding path diagram of the intelligent robot arm bone grinding mechanical model construction method is shown in Figure 2 As shown, the robot moves the milling unit along a straight path at a fixed linear speed perpendicular to the surface of the bone block, maintaining the grinding depth at 0.5 mm during the process, and records the six-dimensional force sensor data as the raw signal.

[0033] Table 1 Full factorial experimental design matrix

[0034]

[0035] like Figure 3As shown, the grinding force components F in the X, Y, and Z directions of the intelligent robotic arm bone grinding mechanical model construction method X 、F Y and F Z Schematic diagram of;

[0036] Taking Experiment 1 as an example, the bone density is 480kg / m 3 , feed speed 1.5mm / s, spindle speed 3000,

[0037] Based on k sampling time points, obtain the resultant force F k for:

[0038]

[0039] F X 、F Y and F Z The force sensor records the milling data of X, Y and Z axes of a certain bone density milling at time k respectively;

[0040] In this embodiment, the specific implementation method of processing the resultant force signal in S1 is as follows:

[0041] like Figure 7 As shown in FIG, the resultant force signal is input into a Butterworth low-pass filter for denoising, and the denoised resultant force signal is smoothed using a moving average. After smoothing, the average value of the resultant force for each bone density is calculated.

[0042] The resultant force signal is input into the Butterworth low-pass filter for denoising, which specifically includes:

[0043] Formula 2 is used for denoising;

[0044]

[0045] Among them, F[k] is the resultant force signal, y[k] is the filtered signal, reverse is the signal sequence inversion, which is used for zero phase shift filtering, that is, eliminating phase distortion through bidirectional filtering, a n with b n As the filter coefficient, there are cutoff frequency and sampling frequency, z -n is the unit delay operator, N is the filter order, and H(z) transforms the time domain difference equation of the discrete time system into a rational function form in the complex frequency domain through Z transform, where z is the core complex variable of Z transform, which is used to describe the complex frequency domain characteristics of the discrete system and is expressed as z=e sT , s is the complex frequency of continuous time, and T is the sampling period.

[0046] The moving average is used to smooth the denoised resultant force signal. Specifically, the denoised resultant force signal is smoothed using Formula 3;

[0047]

[0048] Where x is the denoised force signal, y smooth is the smoothed signal, M is the moving average window size, k is the index of the current sampling point, and m is the offset within the window.

[0049] Calculate y at k sampling times smooth [k] The average value A of the smoothed signal. The intelligent robotic arm bone grinding dataset is represented as (A, 480, 3000, 1.5).

[0050] In the embodiment of the present invention, the effects of the spindle speed, feed rate, and bone density of the spherical grinding drill on the average value of the resultant force are analyzed to verify the rationality of the experimental data collected. It is also clarified that there is a nonlinear mapping relationship between the spindle speed, feed rate, bone density, and grinding force, which guides the subsequent neural network structure design and parameter optimization. The specific process is as follows:

[0051] Analyze the robotic bone grinding process and calibrate the X, Y, and Z axes of the robot's end-point spherical drill. The X axis is perpendicular to the grinding bone surface, the Y axis is the feed direction, and the Z axis is perpendicular to the X and Y axes.

[0052] Abstract and simplified bone grinding mechanical behavior process, the spherical drill feed speed f along the Y axis, and the spherical drill spindle speed n are constructed. Based on the constructed robotic bone grinding dataset, a statistical analysis is conducted on the influence of the two controllable parameters, feed speed and spindle speed, on the grinding force during robotic grinding of artificial bone blocks with different densities. The grinding force here is the average value of the resultant force in the embodiments of the present invention. The results show that the grinding force decreases with the increase of the robot spindle speed, increases with the increase of the feed speed, and increases with the increase of bone density.

[0053] Figure 4 The bone density provided by the present invention is 480kg / m 3 Schematic diagram of the influence of spindle speed and feed speed on grinding force under the condition of Figure 4 As shown, bone density remains at 480 kg / m 3 , the grinding force decreases with the increase of spindle speed and increases with the increase of feed speed.

[0054] Figure 5 FIG. 1 is a schematic diagram showing the influence of spindle speed and bone density on grinding force under the condition of feed rate of 2 mm / s provided by the embodiment of the present invention. Figure 5 As shown,

[0055] The feed rate was kept at 2 mm / s, and the grinding force increased with the increase of bone density and decreased with the increase of spindle speed.

[0056] Figure 6 Schematic diagram of the effect of feed rate and bone density on grinding force when the spindle speed is 4000 rpm provided in an example of the present invention;

[0057] The spindle speed is kept at 4000 rpm. The grinding force increases with the increase of bone density and the grinding force increases with the increase of feed rate. Figure 6 shown.

[0058] S2. Based on the intelligent robotic arm bone grinding dataset, a bone grinding mechanical model was constructed using the radial basis function neural network (RBNN).

[0059] The S2 specifically includes: constructing a bone grinding mechanical model of the average value of the three grinding parameters of bone density, spindle speed and feed speed and the resultant force through a radial basis function neural network based on the intelligent robotic arm bone grinding data set.

[0060] Figure 8 Schematic diagram of the radial basis function neural network structure of the intelligent robot arm bone grinding mechanical model construction method provided by the example of the present invention. Figure 8 As shown;

[0061] In the embodiment of the present invention, the radial basis function neural network consists of three layers: input layer, hidden layer and output layer;

[0062] A certain bone density, the spindle speed corresponding to the average value, and the feed speed corresponding to the average value are used as training samples x i Enter the input layer;

[0063] Training sample x i Represented as x i ={x i1 , x i2 , x i3}, the total training samples are expressed as X = {x1, x2, x3, ....., x m}, 1≤i≤m, m represents the number of training samples;

[0064] The hidden layer implements nonlinear transformation through radial basis function, and the i-th training sample x in the input layer i The information output of the neuron input to the jth node of the hidden layer is calculated as:

[0065]

[0066] Among them, c j and σ j They are respectively represented as the cluster center and standard deviation of the j-th Gaussian activation function, ||x i -c j|| is the input sample x i With center c j The Euclidean distance,

[0067] Use the K-means clustering algorithm to determine the cluster center c of the j-th Gaussian activation function in the hidden layer j , expressed as:

[0068]

[0069] Among them, S j is the sample set belonging to the jth cluster, |S j | is the number of samples actually assigned to the jth cluster, which is used to locally calculate the cluster center;

[0070] The output layer includes neurons with linear activation function, which transforms the hidden layer neuron information φ j (x i ) is weighted and summarized as follows:

[0071]

[0072] Among them, x i is the i-th training sample, y k is the output value of the kth output neuron, w jk is the weight from the jth neuron in the hidden layer to the kth neuron in the output layer, b k is the bias of the kth neuron in the output layer;

[0073] Formula 6 is a bone grinding mechanical model constructed by constructing the average value of the three grinding parameters, bone density, spindle speed, and feed rate, and the resultant force.

[0074] Figure 9 FIG. 1 is a schematic diagram showing the fitting of the training set after the neural network training of the intelligent robotic arm bone grinding mechanical model construction method provided by the embodiment of the present invention is completed. Figure 9 As shown, its R is 0.99928, close to the ideal value of 1, indicating that the model fits the training data set well;

[0075] Figure 10 : is a schematic diagram of the fitting of the test set after the neural network training of the intelligent robotic arm bone grinding mechanical model construction method provided by the example of the present invention is completed, as shown in FIG. Figure 10 As shown in Figure 2, its R is 0.99937, which is close to the ideal value of 1, indicating that the model has excellent predictive performance.

[0076] In the embodiment of the present invention, the mean absolute error, root mean square error, determination coefficient and mean absolute percentage error are used to evaluate the grinding mechanics model.

[0077] The calculation method of MAPE is as follows:

[0078]

[0079] The calculation method of RMSE is as follows:

[0080]

[0081] The calculation method of MAE is as follows:

[0082]

[0083] The calculation method of R2 is as follows:

[0084]

[0085] in, is the bone grinding force predicted by the neural network, y k is the true bone grinding force, and N is the number of test sets.

[0086] The model evaluation results are shown in Table 2, where RMSE is 0.28277, MAE is 0.22299, MAPE is 1.4731%, and R 2 The determination coefficient R is 0.99873. 2 The model's prediction performance is good, with a value of 0.99873, close to 1. The RMSE and MAE are small, demonstrating the model's high prediction accuracy. The MAPE is also low, demonstrating the model's high prediction stability. In summary, the constructed model is effective in grinding force prediction and possesses high credibility and practical value.

[0087] Table 2 RMSE, MAE, MAPE and R of grinding force model 2

[0088] RMSE MAE MAPE (%) <![CDATA[R 2 ]]> F 0.28277 0.22299 1.4731 0.99873

[0089] The technical solution provided by the present invention can realize automatic evaluation of bone grinding force based on the three grinding parameters of grinding bone density, feed speed and spindle speed, which is of great significance for determining the safety boundary of robot bone grinding and optimizing grinding parameters.

[0090] System Example

[0091] According to an embodiment of the present invention, a system for constructing a mechanical model of bone grinding of an intelligent robotic arm is provided. Figure 11 FIG is a schematic diagram of a system for constructing a mechanical model of bone grinding of an intelligent robotic arm according to an embodiment of the present invention. Figure 11 As shown, specifically including:

[0092] The dataset module is used to collect bone grinding experimental data under different bone density, spindle speed, and feed speed conditions through designed experiments, and preprocess the bone grinding experimental data to form a robotic arm bone grinding dataset containing grinding parameters and corresponding grinding forces. The grinding parameters include: bone density, spindle speed of the spherical drill on the intelligent robotic arm, and feed speed;

[0093] A construction module is used to construct a bone grinding mechanical model based on the intelligent robotic arm bone grinding dataset using a radial basis function neural network (RBNN).

[0094] The embodiment of the present invention is a system embodiment corresponding to the above-mentioned method embodiment. The specific operations of each module can be understood by referring to the description of the method embodiment, which will not be repeated here.

[0095] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the above embodiments, or replace some or all of the technical features therein with equivalents. These modifications or replacements of the technical solutions of the embodiments of the present invention do not cause the essence of the corresponding technical solutions to deviate from the scope of this solution.

Claims

1. A method for constructing a mechanical model of bone grinding of an intelligent robotic arm, characterized in that: include, S1. Collect bone grinding experimental data under different bone density, spindle speed, and feed speed conditions through designed experiments, and preprocess the bone grinding experimental data to form a robotic arm bone grinding dataset containing grinding parameters and corresponding grinding forces. The grinding parameters include: bone density, spindle speed of the spherical drill on the intelligent robotic arm, and feed speed; S2. Based on the intelligent robotic arm bone grinding dataset, a bone grinding mechanical model was constructed using the radial basis function neural network (RBNN).

2. The method according to claim 1, characterized in that The S1 specifically includes: adding a force sensor to the intelligent robotic arm, arranging and combining the three grinding parameters of bone density, spindle speed of the spherical drill on the intelligent robotic arm, and feed speed to obtain a grinding parameter group, performing intelligent robotic arm milling based on the parameters in the grinding parameter group, recording milling data of different bone densities at multiple sampling time points through the force sensor, obtaining a resultant force signal corresponding to the bone density based on the milling data of different bone densities, processing the resultant force signal, calculating the average value of the resultant force for each bone density at multiple sampling time points, combining the average value of the resultant force with the corresponding three grinding parameters into a data pair, and forming a structured robotic bone grinding data set.

3. The method according to claim 2, characterized in that The method of obtaining the resultant force signal corresponding to the bone density based on the milling data of different bone densities specifically includes: Formula 1 is used to calculate the resultant force signal, k is the sampling time point, F X 、F Y and F Z The force sensor records the milling data of the X, Y, and Z axes of a certain bone density milling at time k.

4. The method according to claim 3, characterized in that The processing of the resultant force signal specifically includes: The resultant force signal is input into a Butterworth low-pass filter for denoising, and the denoised resultant force signal is smoothed using a moving average.

5. The method according to claim 4, characterized in that The step of inputting the resultant force signal into a Butterworth low-pass filter for denoising specifically includes: Formula 2 is used for denoising, and formula 2 is as follows: y[k]=reverse(H(z)(reverse(H(z)F[k]))), Among them, F[k] is the resultant force signal, y[k] is the filtered signal, reverse is the signal sequence inversion, which is used for zero phase shift filtering, that is, eliminating phase distortion through bidirectional filtering, a n with b n is the filter coefficient, with cutoff frequency and sampling frequency, z -n is the unit delay operator, N is the filter order, and H(z) transforms the time domain difference equation of the discrete time system into a rational function form in the complex frequency domain through Z transform, where z is the core complex variable of Z transform, which is used to describe the complex frequency domain characteristics of the discrete system and is expressed as z=e sT , s is the complex frequency of continuous time, and T is the sampling period.

6. The method according to claim 4, characterized in that The moving average is used to smooth the denoised force signal, specifically including: Formula 3 is used to smooth the denoised resultant force signal. Formula 3 is as follows: Where x is the denoised force signal, y smooth is the smoothed signal, M is the moving average window size, and m is the offset within the window.

7. The method according to claim 6, characterized in that The S2 specifically includes: The radial basis function neural network consists of three layers: input layer, hidden layer and output layer; Input a certain bone density, the spindle speed corresponding to the average value, and the feed speed corresponding to the average value as training samples xi into the input layer; Training sample x i Represented as x i ={x i1 , x i2 , x i3 }, the total training samples are expressed as X = {x1, x2, x3, ....., x m }, 1≤i≤m, m represents the number of training samples; The hidden layer implements nonlinear transformation through radial basis function, and the i-th training sample x in the input layer i The information output of the neuron input to the jth node of the hidden layer is calculated as: Among them, c j and σ j They are respectively represented as the cluster center and standard deviation of the j-th Gaussian activation function, ||x i -c j || is the input sample x i With center c j The Euclidean distance of Use the K-means clustering algorithm to determine the cluster center c of the j-th Gaussian activation function in the hidden layer j , expressed as: Among them S j is the sample set belonging to the jth cluster, |S j | is the number of samples actually assigned to the jth cluster, which is used to locally calculate the cluster center; The output layer includes neurons with linear activation function, which transforms the hidden layer neuron information φ j (x i ) is weighted and summarized as follows: Among them, x i is the i-th training sample, y k is the grinding force output by the kth output neuron, w jk is the weight from the jth neuron in the hidden layer to the kth neuron in the output layer, b k is the bias of the k-th neuron in the output layer; The optimization algorithm is used to adjust the parameters of the radial basis function neural network so that the average error between the output grinding force and the corresponding input resultant force is minimized. Formula 6 is a bone grinding mechanical model constructed by constructing the average value of the three grinding parameters, bone density, spindle speed, and feed rate, and the resultant force.

8. The method according to any one of claims 1 to 7, further comprising evaluating the bone grinding mechanical model using mean absolute error, root mean square error, coefficient of determination, and mean absolute percentage error.

9. An intelligent robotic arm bone grinding mechanical model construction system, characterized by: include, The dataset module is used to collect bone grinding experimental data under different bone density, spindle speed, and feed speed conditions through designed experiments, and preprocess the bone grinding experimental data to form a robotic arm bone grinding dataset containing grinding parameters and corresponding grinding forces. The grinding parameters include: bone density, spindle speed of the spherical drill on the intelligent robotic arm, and feed speed; A construction module is used to construct a bone grinding mechanical model based on the intelligent robotic arm bone grinding dataset using a radial basis function neural network (RBNN).

Images