Bone layer identification method for robot bone grinding

Abstract

The embodiment of the invention provides a bone layer identification method for robot bone grinding, and the method comprises the steps: collecting a force signal and / or a torque signal through a six-dimensional force sensor, collecting a vibration signal through an accelerometer, and obtaining a dual-mode signal data set under bone grinding operation for three speed modes; according to the collected force signals and / or torque signals, noise reduction processing is carried out to obtain force / torque signal characteristics; according to the collected vibration signals, vibration signal characteristics are obtained through modal decomposition; according to the signal features obtained in the three speed modes, a classification model is constructed, and bone layer decision hyperplanes in the three speed modes are generated; according to a speed mode used in actual bone grinding operation, a force signal and / or a torque signal and a vibration signal in the bone grinding process are / is collected, a signal feature array is solved, and according to the position relation between a plane function and the feature array in the actual speed mode, the bone tissue type of the position where a grinding tool is located at the moment corresponding to each set of data is obtained.

Description

[Technical field]

[0001] The invention relates to the field of orthopedic robot grinding operations, and in particular to a bone layer identification method based on dual-modal signal characteristics. [Background technology]

[0002] Orthopedic surgery poses severe challenges to doctors' professional skills and physical fitness due to its complexity, trauma, radiation exposure and time-consuming nature. The emergence of orthopedic robots will effectively improve the accuracy of surgical operations and shorten the duration of operations. However, the robot's perception of complex lesions will directly affect the robot's operation. In traditional surgery, doctors usually rely on CT images for preoperative planning and take images before and during surgery to understand the bone layer structure of the lesion area. There are some disadvantages in relying on medical imaging technology, including the potential harm of radiation generated by image shooting to the health of doctors and patients, and the non-real-time nature of images. In addition, frequent shooting during surgery will have an adverse effect on the continuity and duration of the operation. Orthopedic robots usually use sensor detection data to identify the bone layer, collect multiple signal data, and identify the tissue around the lesion area through processing and thresholding. Although more real-time data can be obtained by adjusting the sampling frequency of the sensor, most existing methods rely on single signal data. In the face of complex and dynamically changing surgical environments, interference with signal acquisition is difficult to avoid, which directly affects the accuracy of image data and the recognition results. The bone layer recognition technology based on multi-sensor perception fusion is an effective method to improve the recognition accuracy. It utilizes the redundancy, complementarity and real-time nature of multi-source information to achieve accurate recognition even in more complex situations. Therefore, it is a technical problem that needs to be solved urgently in this field for orthopedic robots to rely on multi-sensor information to achieve high-precision and stable recognition of bone layers. [Summary of the invention]

[0003] In view of this, the present invention provides a bone layer identification method for robot bone grinding, so as to determine the type of bone tissue where the grinding tool is located during the bone grinding operation.

[0004] The embodiment of the present invention provides a bone layer identification method, including a plane function generation process and an identification process:

[0005] The plane function generation process includes:

[0006] For the three speed modes, a six-dimensional force sensor is used to collect force signals and / or torque signals in each speed mode, and an accelerometer is used to collect vibration signals to obtain a dual-modal signal data set under the bone grinding operation; wherein each speed mode includes a drill speed and a feed speed;

[0007] According to the collected force signal and / or torque signal, noise reduction processing is performed to obtain the force / torque signal characteristics; wherein the filter used has preset filter parameters and is constructed by calculating the actual impulse response function using a window function;

[0008] According to the collected vibration signal, the vibration signal characteristics are obtained through modal decomposition;

[0009] According to the signal features obtained in the three speed modes, a classification model is constructed to generate the bone layer decision hyperplane in the three speed modes.

[0010] The identification process includes:

[0011] According to the speed mode used in the actual bone grinding operation, the grinding force and / or grinding torque signal and the vibration signal of the bone grinding process are collected to obtain the signal characteristics;

[0012] According to the plane function and the signal characteristics, the bone tissue type at the position of the grinding tool at the time corresponding to each group of signal characteristics is identified.

[0013] In the above method, for the three speed modes, a six-dimensional force sensor is used to collect force signals and / or torque signals under the bone grinding operation in each speed mode, and an accelerometer is used to collect vibration signals under the bone grinding operation to obtain a dual-modal signal data set under the bone grinding operation, including:

[0014] The speed mode includes a drill speed and a feed speed, based on which three speed modes are set for signal acquisition. The grinding operation tool is perpendicular to the surface of the bone material, and the grinding depth is set to 1mm. Grinding operations are performed on cancellous bone material, cortical bone material, and uncontacted bone material respectively. A six-dimensional force sensor is used to collect the force signal and / or torque signal F of the grinding drill at the end of the robotic arm, and an acceleration sensor is used to collect the vibration signal a of the robotic arm during the grinding process; by repeatedly grinding bone materials, the force signal and / or torque signal and vibration signal in the process are recorded, and time alignment is performed. After data annotation, a dual-modal signal data set under the bone grinding operation is generated.

[0015] In the above method, the noise reduction process is performed based on the collected force signal and / or torque signal to obtain the force / torque signal characteristics, including:

[0016] The force signal and / or torque signal F of the grinding drill bit at the end of the robot arm is characterized as follows:

[0017] F=[F x F y F z M x M y M z ]

[0018] Among them, [F x F y F z ] is the force in the xyz direction of the coordinate system, [M x M y M z ] is the moment around the coordinate axis xyz in three directions;

[0019] F x 、F y 、F z 、M x 、M y 、M z As the input of the filter, it is respectively brought into the low-pass filter to obtain the signal characteristics of the force signal and / or torque signal after noise reduction

[0020] In the above method, the step of obtaining vibration signal features by modal decomposition based on the collected vibration signal includes:

[0021] The vibration signal a collected by the robot arm during bone grinding operation is characterized as follows:

[0022] a=[a x a y a z ]

[0023] Among them, a x 、a y 、a z They are the vibration signals in the three-axis directions of the coordinate system xyz;

[0024] to a x 、a y 、a z Perform modal decomposition operations according to the following steps respectively, decompose into 8 modal components (IMF), and take the three modal components of IMF5, IMF6 and IMF7 as vibration signal features, denoted as

[0025] Step 1) Use Hilbert transform to obtain the analytical signal of each modal component, and define the variational problem with the minimum sum of modal estimation bandwidth:

[0026]

[0027] In the formula, u k (t) represents each modal component, where t represents a discrete time series, k is a single-sided spectrum, is the center frequency estimation operator; define {u i} represents the decomposed modal components, and {ui}={u 1 ,...,u k},{ω i} represents the center frequency of each modal component after decomposition, and {ω i}={ω 1 ,...,ω k};

[0028] Step 2) Introduce the Lagrangian operator λ(t) and the second-order penalty parameter α, and transform the variational problem in the above equation into an augmented Lagrangian expression as follows:

[0029]

[0030] In the formula, S(t) represents the sum of each mode, u i (t) represents the i-th modal component.

[0031] Step 3) Starting from parameter n = 1, update the parameter Lagrange multiplication operator λ n+1 , center frequency And by fixing the parameter λ n+1 , Update Parameters Combined with the Parseval Fourier isometric transform, we can get

[0032]

[0033] In the formula, They are all expressed as the sum of the modes in the frequency domain, the Lagrange coefficients, and the i-th modal component.

[0034] Step 4) Fix the parameters λ n+1 , update the parameters get And, fixed parameters Update parameter λ n+1 , we get n+1 ,as follows:

[0035]

[0036] Step 5) When the discrimination accuracy ε>0, Stop the iteration, otherwise, return to step 3).

[0037] In the above method, the classification model is constructed based on the signal features respectively obtained in the three speed modes to generate the bone layer decision hyperplane in the three speed modes, including:

[0038] A set of signal feature arrays obtained at time t in any speed mode is characterized as:

[0039]

[0040] The plane function parameter ω is calculated using the following formula: * 、b * :

[0041]

[0042] Among them, α j * represents the optimal Lagrangian coefficient α * A component of , satisfying 0<α j * <C, C represents the penalty parameter, x i Represents the input value of the i-th data point in the array, y i represents the output value of the algorithm for the i-th data point, x j Represents the input value of the jth data point in the array, y j Represents the output value of the algorithm for the jth data point;

[0043] f(x)=sign(ω * ·x+b * )

[0044] Among them, sign() represents the sign function, f(x) is the classification plane function, and the classification plane function f(x) is used to identify the category of bone material contacted by the bone grinding drill.

[0045] In the above method, the speed mode used in the actual bone grinding operation is used to collect the grinding force and / or grinding torque signal and the vibration signal of the bone grinding process to obtain the signal characteristics, and according to the positional relationship between the plane function and the characteristic array under the actual speed mode, the bone tissue type at the position of the grinding tool at the corresponding moment of each set of data is obtained, including:

[0046] Use the force signal and / or torque signal acquired during the bone grinding operation in actual speed mode to solve the signal feature array:

[0047] T test ={T(1),T(2),...}

[0048] According to the plane function and the signal features, the bone tissue type at the position of the grinding tool at the time corresponding to each group of signal features is identified, including: test Substitute the data T(t) of each set of signal features in (t) into the plane function under the corresponding speed mode and make two judgments:

[0049]

[0050] When y=1, the drill bit does not contact the bone layer; when y=0, the drill bit contacts the bone layer; when y=2, the drill bit is in the cortical bone layer; when y=3, the drill bit is in the cancellous bone layer, so as to obtain the bone tissue type at the position of the drill bit of the grinding tool.

[0051] In the technical solution of the embodiment of the present invention, the orthopedic robot can sense the position of the bone layer where the operating tool is located during the grinding process based on the information obtained by the sensor without the basis of medical images, which can provide protection for the safety of the bone grinding operation and ensure the effectiveness and safety of the operation.

Brief Description of the Drawings

[0052] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for use in the embodiments are briefly introduced below. It is obvious that the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creativity and labor.

[0053] Figure 1 It is a flow chart of a bone layer identification method for robot bone grinding provided by an embodiment of the present invention;

[0054] Figure 2 It is a technical route schematic diagram of a bone layer identification method for robot bone grinding provided by an embodiment of the present invention;

[0055] Figure 3 is a graph of force / torque signal processing during bone grinding collected by an embodiment of the present invention;

[0056] Figure 4 is a processing diagram of vibration signals collected during bone grinding according to an embodiment of the present invention;

[0057] Figure 5 4 is a diagram of bone layer identification results according to an embodiment of the present invention. [Specific embodiment]

[0058] In order to better understand the technical solution of the present invention, the embodiments of the present invention are described in detail below with reference to the accompanying drawings.

[0059] It should be clear that the described embodiments 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 creative work are within the scope of protection of the present invention.

[0060] The embodiment of the present invention provides a bone layer identification method based on dual-modal signal characteristics, please refer to Figure 1, which is a flow chart of a bone layer identification method based on dual-modal signal features provided by an embodiment of the present invention, such as Figure 1 As shown, the method comprises the following steps:

[0061] Step 101 , for three speed modes, a six-dimensional force sensor is used to collect force signals and / or torque signals in each speed mode, and an accelerometer is used to collect vibration signals, thereby constructing a dual-modal signal data set under the bone grinding operation.

[0062] The speed mode includes a drill speed and a feed speed, thus three speed modes are set for signal acquisition. The specific setting parameters of the three speed modes are: drill speed of 20000r / min, feed speed of 0.133mm / s, drill speed of 20000r / min, feed speed of 0.167mm / s, and drill speed of 16000r / min, feed speed of 0.167mm / s.

[0063] The grinding operation tool is perpendicular to the surface of the bone material, and the grinding depth is set to 1mm. Grinding operations are performed on cancellous bone material, cortical bone material, and uncontacted bone material respectively. A six-dimensional force sensor is used to collect the force signal and / or torque signal F of the grinding drill at the end of the robotic arm, and an acceleration sensor is used to collect the vibration signal a of the robotic arm during the grinding process;

[0064] Carry out data collection for bone grinding operations. By repeatedly grinding bone materials, record the force signal and / or torque signal and vibration signal during the process, and perform time alignment. After data annotation, generate a dual-modal signal data set under the bone grinding operation.

[0065] Step 102: Perform noise reduction processing based on the collected force signal and / or torque signal to obtain force signal characteristics.

[0066] The force signal and / or torque signal F of the grinding drill bit at the end of the robot arm is characterized as follows:

[0067] F=[F x F y F z M x M y M z ]

[0068] In the above formula, [F x F y F z ] is the force in the xyz direction of the coordinate system, [M x M y M z ]Torques in three directions around the coordinate axes xyz;

[0069] To design a filter to achieve signal noise reduction, first retain the preset filter parameters, and then use the window function to calculate the actual impulse response function, where the ideal filter frequency response H d (e jω )for:

[0070]

[0071] In the formula, ω c is the cut-off frequency, α represents the delay parameter of the filter;

[0072] And, the unit impulse response h is calculated d (n):

[0073]

[0074] Among them, n is a sequence variable;

[0075] The impulse response function intercepted by the window function is:

[0076]

[0077] w(n) is the selected window function, N is the sequence length;

[0078] The impulse response H(e) of the actual filter is obtained by calculating the convolution jω ):

[0079] H(e jω )=H d (e jω )×W(e jω )

[0080] In the formula, H d (e jω ) is the ideal filter frequency response, W(e jω ) is the frequency response of the window function.

[0081] Through inverse Fourier transform, the impulse response in the time domain is obtained, and the filter is constructed based on the impulse response:

[0082] H(t)=H d (t)*W(t)

[0083] Where H(t) is the impulse response of the ideal filter in the time domain, and W(t) is the frequency response of the window function in the time domain.

[0084] The experiment collects force / torque signals, such as Figure 3 (a), characterized by F x 、F y 、F z 、M x 、My 、M z , input into the constructed low-pass filter, thereby achieving signal noise reduction, satisfying the following expression:

[0085] y[n]=a 0 x[n]+a 1 x[n-1]+a 2 x[n-2]+...a m x[nm]

[0086] In the above formula, x[n] represents the current input signal sample, y[n] represents the current output signal sample, and a m represents the impact response coefficient of the filter, m represents the order of the filter, and the force / torque signal with high-frequency noise filtered out is taken as the signal feature, which is recorded as The filtered signal is Figure 3 (b) as shown.

[0087] Step 103, obtaining vibration signal features through modal decomposition based on the collected vibration signal.

[0088] The vibration signal a of the acquisition robot arm during bone grinding operation is characterized as follows:

[0089] a=[a x a y a z ]

[0090] In the above formula, a x 、a y 、a z They are the vibration signals in the three axes of the coordinate system xyz. x 、a y 、a z The modal decomposition is performed according to the following steps respectively, and it is decomposed into 8 modal components (IMF), and the three modal components of IMF5, IMF6 and IMF7 are used as vibration signal features, denoted as like Figure 4 As shown,.

[0091] ① Use Hilbert transform to obtain the analytical signal of each modal component, and use the following formula to solve the single-sided spectrum k:

[0092]

[0093] In the formula, u k (t) represents each modal component.

[0094] ②Multiply the obtained analytical signals of each modal component by the estimated center frequency operator to modulate each modal spectrum to the corresponding baseband, as shown in the formula:

[0095]

[0096] In the formula, is the operator for estimating the center frequency.

[0097] ③ Calculate the 2-norm square of the demodulated signal gradient and define the variational problem with the minimum sum of the modal estimation bandwidth, as shown in the following formula:

[0098]

[0099] In the formula, {u i} represents the decomposed modal components, and {u i}={u 1 ,...,u k},{ω i} represents the center frequency of each mode after decomposition, and {ω i}={ω 1 ,...,ω k}.

[0100] ④ By introducing the Lagrangian operator λ(t) and the second-order penalty parameter α, the variational problem defined in the above formula is transformed into an augmented Lagrangian expression as follows:

[0101]

[0102] ⑤ From parameter n = 1, use the alternating direction multiplier method to update the parameters Lagrange multiplication operator λ n+1 , center frequency As follows:

[0103]

[0104] ⑥ By fixing the parameter λ n+1 , Update Parameters Solution

[0105] Parameters Perform frequency domain transformation and use Parseval Fourier isometric transform to obtain the following formula:

[0106]

[0107] Replace the first polynomial in the above formula with ω-ω i Replace ω, transform it into an integral in a non-negative frequency interval, and take the derivative to get The solution is as follows:

[0108]

[0109] ⑦Fixed parameters λ n+1 ,renew Available Similarly, fixed parameters Update λ n+1 , we can get λ n+1 , as follows:

[0110]

[0111] ⑧When the discrimination accuracy ε>0, Stop the iteration, otherwise return to step ⑥.

[0112] Step 104: construct a classification model based on the signal features obtained in the three speed modes respectively, and generate bone layer decision hyperplanes in the three speed modes.

[0113] A set of signal feature arrays obtained at time t in any speed mode is characterized as:

[0114]

[0115] Each set of signal features corresponds to a label y∈{0,1,2}, which indicates the bone layer (uncontacted, cortical bone or cancellous bone) to which the signal feature belongs.

[0116] The support vector machine (SVM) is used to optimize the parameters w and b of the plane function under a certain speed mode, that is, to find an optimal decision hyperplane for classification.

[0117] SVM calculates the optimal segmentation hyperplane by maximizing the interval of the decision hyperplane, where the objective function used for optimization is:

[0118]

[0119] Among them, w k is the normal vector of the hyperplane; b k is the bias term, ξ i is a slack variable that allows a certain degree of classification error; C is a penalty parameter that controls the tolerance of the error; k represents the speed mode.

[0120] To solve the optimization problem, SVM introduces the Lagrange multiplier method to solve the optimal model parameters. After obtaining the optimal Lagrange coefficient, the plane function parameters are calculated using the following formula:

[0121]

[0122] Among them, α j * represents the optimal Lagrangian coefficient α *A component of , satisfying 0<α j * <C, C represents the penalty parameter; Indicates w k 、b k Optimal solution; x i Represents the input value of the i-th data point in the array, y i represents the output value of the algorithm for the i-th data point, x j Represents the input value of the jth data point in the array, y j Represents the output value of the algorithm for the jth data point;

[0123] The parameters obtained by calculation and Constructing classification hyperplane function This function performs bone layer classification prediction for a set of signal features in the kth velocity mode.

[0124] Support vector machines (SVMs) use symbolic functions to make classification decisions:

[0125]

[0126] Among them, sign() represents the sign function; is the classification plane function under the kth speed mode; y k Indicates the category of bone material contacted by the bone grinding drill as judged by the classification model.

[0127] In order to judge the three categories of samples, two classifiers are constructed to judge whether the grinding tool contacts the bone material and whether the contacted bone material is cortical bone or cancellous bone, thereby realizing the perception of the position of the grinding tool. The final bone layer classification effect is as follows: Figure 5 shown.

[0128] Step 105, for the speed mode used in the actual bone grinding operation, collect the grinding force and / or grinding torque signal and vibration signal of the bone grinding process to obtain the signal characteristics, and obtain the bone tissue type at the position of the grinding tool at the corresponding moment of each set of data based on the positional relationship between the plane function and the characteristic array under the actual speed mode.

[0129] Use the force signal and / or torque signal acquired during the bone grinding operation in actual speed mode to solve the signal feature array:

[0130] T test ={T(1),T(2),...}

[0131] According to the plane function and the signal features, the bone tissue type at the position of the grinding tool at the time corresponding to each group of signal features is identified, including: testSubstitute the data T(t) of each set of signal features in (t) into the plane function under the corresponding speed mode and make two judgments:

[0132]

[0133] When y=1, the drill bit does not contact the bone layer; when y=0, the drill bit contacts the bone layer; when y=2, the drill bit is in the cortical bone layer; when y=3, the drill bit is in the cancellous bone layer, so as to obtain the bone tissue type at the position of the drill bit of the grinding tool.

[0134] According to the above method provided by the embodiment of the present invention, a bone layer identification method based on two signal features of a robot in an orthopedic grinding operation is simulated.

[0135] This embodiment uses a UR10 six-degree-of-freedom robot arm and an M212HD high-speed grinding drill to perform bone grinding operations. The robot system uses an FT-300S six-dimensional force sensor and an accelerometer included in an MTi-300 inertial measurement unit to collect grinding force / torque signals and vibration signals.

[0136] In the embodiment of the present invention, the actual drill speed is 20000r / min and the feed speed is 0.133mm / s. Figure 2 As shown in the flow chart, by collecting grinding force / torque signal and vibration signal, the force / torque signal is subjected to noise reduction processing to obtain the force / torque signal characteristics, refer to Figure 3 As shown in the figure, a) is the original signal collected, b) is the force / torque signal after noise reduction, and the vibration signal is decomposed into 8 layers by modal decomposition to obtain the vibration signal characteristics. Figure 4 As shown, a) is the vibration signal in the x direction, b) is the vibration signal in the y direction, and c) is the vibration signal in the z direction. The left side of each figure is the decomposed time domain image, and the right side is the corresponding spectrum diagram. According to the obtained signal characteristics, the plane function under this speed mode is solved. Then, with this speed mode, the bone grinding operation is carried out, and the force / torque signal and vibration signal are collected to extract the signal characteristics. The obtained signal feature array is substituted into the obtained plane function to obtain the positional relationship between the data point and the plane function, and then the bone layer at this time is judged. The horizontal axis is time T and the vertical axis is the y value. Figure 5 .

[0137] The technical solution of the embodiment of the present invention has the following beneficial effects:

[0138] (1) In the technical solution of the embodiment of the present invention, force / torque signals and vibration signals are obtained during the bone grinding operation under three speed modes, and the force / torque signals are subjected to noise reduction processing, and the vibration signals are subjected to modal decomposition to obtain force / torque signal features and vibration signal feature quantities;

[0139] (2) The present invention constructs a classification model, solves the classification plane function under the three speed modes according to the signal feature quantities obtained under the three speed modes, and judges the position of the bone layer where the bone grinding drill is located at this moment according to the feature array of the force / torque signal and the vibration signal collected by the actual drill bit during the grinding process under the speed mode, combined with the position relationship between the plane function and the feature array under the corresponding actual speed mode, so as to realize the identification of the bone layer.

[0140] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

[0141] The contents not described in detail in the specification of the present invention belong to the common knowledge of those skilled in the art.

Claims

1. A method for identifying bone layers in robot bone grinding, characterized in that: The method includes a plane function generation process and a recognition process: The plane function generation process includes: For the three speed modes, a six-dimensional force sensor is used to collect force signals and / or torque signals in each speed mode, and an accelerometer is used to collect vibration signals to obtain a dual-modal signal data set under the bone grinding operation; wherein each speed mode includes a drill speed and a feed speed; According to the collected force signal and / or torque signal, noise reduction processing is performed to obtain the force / torque signal characteristics; wherein the filter used has preset filter parameters and is constructed by calculating the actual impulse response function using a window function; According to the collected vibration signal, the vibration signal characteristics are obtained through modal decomposition; According to the signal features obtained in the three speed modes, a classification model is constructed to generate the bone layer decision hyperplane in the three speed modes. The identification process includes: According to the speed mode used in the actual bone grinding operation, the grinding force signal and / or torque signal and vibration signal of the bone grinding process are collected to obtain signal characteristics; According to the bone layer decision hyperplane and the signal features, the bone tissue type at the position of the grinding tool at the corresponding moment of each group of signal features is identified.

2. The method according to claim 1, characterized in that For the three speed modes, a six-dimensional force sensor is used to collect force signals and / or torque signals, and an accelerometer is used to collect vibration signals to obtain a dual-modal signal data set under the bone grinding operation, including: The speed mode includes a drill speed and a feed speed, based on which three speed modes are set for signal collection; the grinding operation tool is perpendicular to the surface of the bone material, and the grinding depth is set to 1mm. Grinding operations are carried out on cancellous bone material, cortical bone material, and uncontacted bone material respectively. A six-dimensional force sensor is used to collect the force signal and / or torque signal F exerted on the grinding drill at the end of the robotic arm, and an acceleration sensor is used to collect the vibration signal a of the robotic arm during the grinding process; by repeatedly grinding bone materials, the force signal and / or torque signal and vibration signal in the process are recorded, and time alignment is performed. After data annotation, a dual-modal signal data set under bone grinding operation is generated.

3. The method according to claim 1, characterized in that According to the collected force signal and / or torque signal, noise reduction processing is performed to obtain force / torque signal characteristics, including: The force signal and / or torque signal F of the grinding drill bit at the end of the robot arm is characterized as follows: F=[F x F y F z M x M y M z ] Among them, [F x F y F z ] is the force in the xyz direction of the coordinate system, [M x M y M z ] is the moment around the coordinate axis xyz in three directions; F x 、F y 、F z 、M x 、M y 、M z As the input of the filter, they are respectively brought into the low-pass filter to obtain the signal characteristics of the force signal and / or torque signal after noise reduction 4. The method according to claim 1, characterized in that: Based on the collected vibration signal, the vibration signal characteristics are obtained through modal decomposition, including: The vibration signal a collected by the robot arm during bone grinding operation is characterized as follows: a=[a x a y a z ] Among them, a x 、a y 、a z They are the vibration signals in the three-axis directions of the coordinate system xyz; to a x 、a y 、a z The modal decomposition operation is performed according to the following steps respectively, and it is decomposed into 8 modal components (IMF), and the three modal components of IMF5, IMF6 and IMF7 are taken as the vibration signal features, which are recorded as Step 1) Use Hilbert transform to obtain the analytical signal of each modal component, and define the variational problem with the minimum sum of modal estimation bandwidth: In the formula, u k (t) represents each modal component, where t represents a discrete time series, k is a single-sided spectrum, is the center frequency estimation operator; define {u i } represents the decomposed modal components, and {u i }={u1,...,u k },{ω i } represents the center frequency of each modal component after decomposition, and {ω i }={ω1,...,ω k }; Step 2) Introduce the Lagrangian operator λ(t) and the second-order penalty parameter α, and transform the variational problem in the above equation into an augmented Lagrangian expression as follows: In the formula, S(t) represents the sum of each mode, u i (t) represents the i-th modal component; Step 3) Starting from parameter n = 1, update the parameter Lagrange multiplication operator λ n+1 , center frequency And by fixing the parameter λ n+1 , Update Parameters Combined with the Parseval Fourier isometric transform, we can get In the formula, They are all expressed as the sum of the modes in the frequency domain, the Lagrange coefficients, and the i-th modal component. Step 4) Fix the parameters λ n+1 , update the parameters get And, fixed parameters Update parameter λ n+1 , we get n+1 ,as follows: Step 5) When the discrimination accuracy ε>0, Stop the iteration, otherwise, return to step 3).

5. The method according to claim 1, characterized in that The method of constructing a classification model based on the signal features respectively obtained in the three speed modes and generating bone layer decision hyperplanes in the three speed modes includes: A set of signal feature arrays obtained at time t in any speed mode is characterized as: The plane function parameter ω is calculated using the following formula: * 、b * : Among them, α j * represents the optimal Lagrangian coefficient α * A component of , satisfying 0<α j * <C, C represents the penalty parameter; x i Represents the input value of the i-th data point in the array, y i represents the output value of the algorithm for the i-th data point, x j Represents the input value of the jth data point in the array, y j Represents the output value of the algorithm's jth data point; ω * is the optimal weight vector, which represents the optimal weight obtained through training or optimization process; b * It is the optimal bias term and determines the location of the classification boundary. f(x)=sign(ω * ·x+b * ) Among them, sign() represents the sign function, f(x) is the classification plane function, and the classification plane function f(x) is used to identify the category of bone material contacted by the bone grinding drill.

6. The method according to claim 1, characterized in that The method collects grinding force and / or grinding torque signals and vibration signals during the bone grinding process according to the speed mode used in the actual bone grinding operation, obtains signal characteristics, and obtains the bone tissue type at the position of the grinding tool at the corresponding moment of each set of data according to the positional relationship between the plane function and the characteristic array under the actual speed mode, including: Use the force signal and / or torque signal acquired during the bone grinding operation in actual speed mode to solve the signal feature array: T test ={T(1),T(2),...} According to the plane function and the signal features, the bone tissue type at the position of the grinding tool at the time corresponding to each group of signal features is identified, including: test Substitute the data T(t) of each set of signal features in (t) into the plane function under the corresponding speed mode and make two judgments: When y=1, the drill bit does not contact the bone layer; when y=0, the drill bit contacts the bone layer; when y=2, the drill bit is in the cortical bone layer; when y=3, the drill bit is in the cancellous bone layer, so as to obtain the bone tissue type at the position of the drill bit of the grinding tool.

Images