Medical universal telescopic drill, control method and application thereof

Abstract

The present invention discloses a medical universal telescopic drill, a control method and an application thereof; the present invention relates to the technical field of dental surgical instruments; the present invention comprises a telescopic component 3 which is telescopically and universally adjusted in the main body component 2, the telescopic component 3 comprises a plurality of small free bodies 302 with dipole moments, the telescopic and universal adjustments can be performed separately or simultaneously; a discretized PID controller receives a target value tv (desired magnetic field strength and direction) and an actual feedback value fv (current magnetic field strength and direction) as inputs; an error E between the target value tv and the feedback value fv is calculated; the present invention realizes the telescopic and universal angle adjustment of the drill component through electromagnetic control, so that the doctor can position and operate more accurately during the operation, and by combining the discretized PID controller and the bidirectional Bi‑GRU recurrent neural network, the integral action and saturation value of the PID controller can be adjusted in real time to adapt to changes in the system state.

Description

Technical Field

[0001] The invention relates to the technical field of dental surgical instruments, in particular to the telescopic technology of a dental telescopic drill and a control method thereof, and in particular to a medical universal telescopic drill, a control method and an application thereof. Background Art

[0002] Medical dental drills are mainly used to clean teeth with caries, to drill pulp holes for teeth with pulpitis and apical periodontitis, and to prepare abutment teeth for toothless implants. In dental surgery, dental drills are used to drill away tooth decay and pulp for subsequent treatment and restoration. Common types of dental drills include round drills (also called diamond drills), horn drills, high-speed steel drills, and diamond twist drills. These drills are usually installed on handheld devices or dental robots.

[0003] Dental treatment requires extremely high precision, especially when performing pulp opening to remove decay, where even a small error may have a significant impact on the treatment effect. Therefore, doctors need to manually adjust to ensure that the distance and angle between the drill bit and the tooth are optimal. However, manual adjustment requires doctors to have rich experience and superb skills, otherwise it is difficult to complete accurately in a short time. For novice doctors, manual adjustment may increase the difficulty and risk of the operation. In this regard, many existing technologies attempt to solve this problem, such as:

[0004] (1) CN202220856371.X discloses a dental implant gingival ring cutting drill (publication date: 2023-02-14), which allows medical staff to directly see the depth to be cut through the scale lines on the drill bit, but this method still requires manual adjustment of the telescopic stroke and cannot adjust the angle;

[0005] (2) CN202320082757.4 discloses a retractable implant reamer (publication date: 2023-06-06), which adjusts the length of the reamer by adjusting the telescopic rod. Although this method solves the problem of manually adjusting the telescopic stroke, due to its mechanism design characteristics, the drill bit cannot be adjusted in angle, and manual adjustment of the drilling angle is still required, especially the angle adjustment during drilling operations;

[0006] (3) CN202110996744.3 discloses a dental implant device, a dental implant system and a dental implant robot (publication date: 2023-04-18), CN202110598496.7 discloses a dental tooth preparation mobile robot and control method (publication date: 2021-09-03), CN201710835480.7 discloses an automatic dental implant robot (publication date: 2018-01-19), CN202110998388.9 discloses an adjustment device, a robotic arm and a dental implant surgery robot (publication date: 2022-10- 18), CN202310283962.1 discloses a control method, system and storage medium for a dental surgical robot (publication date: 2023-06-30). The above-mentioned prior arts all adopt an optical positioning system, combined with a multi-axis robot (mechanical arm) equipped with a dental drill and its actuator, which not only solves the problem of manually adjusting the telescopic stroke, but also realizes semi-automatic adjustment of the drilling angle; however, the multi-axis robot is composed of multiple joints in series, each of which has its own range of motion and direction restrictions, and thus the motion path of the end effector (such as the drill) of the multi-axis robot will be affected by these joint restrictions. For example, when the drill tries to reach a certain angle of the tooth, the multi-axis robot needs to go through a series of complex movements to compensate for the restrictions of each joint, and there is a high probability that the end effector will have a cumulative error phenomenon (attained pose drift, also known as actual pose drift) in the final position. At the same time, each joint of the robotic arm has its own manufacturing tolerance and motion error. When multiple joints are combined to complete a complex movement, cumulative errors will also be caused, resulting in inaccurate final positions.

[0007] Therefore, if the drill bit itself can be telescopic and universally adjusted, it no longer relies on the joint movement of the robot arm to reach the desired angle, which can more directly control the direction of the drill bit and reduce the error caused by the limitation of the robot arm joint. Because the universal angle adjustment of the drill bit body is independent and does not need to be achieved through the movement of multiple joints, it can reduce the inaccuracy caused by the cumulative error of the joints.

[0008] In today's mechanics, the only mechanism that meets the above requirement of "the main body can be adjusted in both universal angles and telescopically" is the parallel mechanism. However, this mechanism requires multiple linear actuators to be assembled in parallel, which is not compatible with the volume of today's dental drills (the drill bits used for filling teeth are generally between 1-3mm in size, while the cylinder size of a single linear actuator, such as most servo electric cylinders, is significantly larger than 3mm). It is also not realistic to place such a large mechanism into the human body to perform complex surgical operations.

[0009] To this end, the present invention provides a medical universal telescopic drill, a control method and application thereof. Summary of the invention

[0010] In view of this, the present invention hopes to provide a medical universal telescopic drill, a control method and an application thereof, so as to solve or alleviate the technical problems existing in the prior art, that is, how to better realize the automatic telescopic adjustment and universal angle adjustment of the drill body without adopting a parallel mechanism; the technical solution of the present invention is implemented as follows:

[0011] First, medical universal telescopic drill bit:

[0012] The invention comprises a drill bit assembly 1 for drilling teeth, a body assembly 2 and:

[0013] A telescopic component 3 in the main body component 2 is telescopic and universally adjustable, the telescopic component 3 includes a plurality of small free bodies 302 with dipole moments, the telescopic and universally adjustable can be performed separately or simultaneously; the drill bit component 1 is mounted on the telescopic component 3;

[0014] A large free body 4 with a dipole moment is fixed in the main body component 2 and is used to generate a magnetic repulsion force with the small free body 302 to enable the telescopic component 3 to perform the telescopic and universal adjustment.

[0015] The telescopic assembly 3 includes a spherical shell 301; a plurality of small free bodies 302 are arranged in a circular array in the spherical shell 301 with the central axis of the large free body 4 as a reference; the drill assembly 1 is mounted at the front end of the spherical shell 301.

[0016] When in use, based on the electromagnetic interaction between the small free body 302 and the large free body 4 , the spherical shell 301 can be controlled to be extended and retracted in the main body component 2 .

[0017] In one embodiment, the telescopic assembly 3 further includes a plurality of electrical connectors 303, each of which is electrically connected to one of the small free bodies 302 via a wire;

[0018] When all the electrical connectors 303 supply power to all the small free bodies 302, the same or different magnetic moments are generated based on the magnitude of the current and voltage generated simultaneously, thereby generating a resultant force of the non-polar magnetic repulsion effect and controlling its magnitude;

[0019] Furthermore, when the drill bit assembly 1 needs to be telescopically adjusted, all the electrical connectors 303 are controlled to function as switches, and the same current and voltage are supplied to all the small free bodies 302 to generate magnetic moments, thereby making the multiple small free bodies 302 generate a uniform force, and then generate a magnetic repulsion force between the large free body 4, so as to control the telescopic assembly 3 as a whole and the drill bit assembly 1 matched therewith to extend;

[0020] When it is necessary to control the drill bit assembly 1 to retract, if direct current is used, the current direction of all the control electrical connectors 303 is reversed, such as switching the positive and negative poles, to directly change the direction of the magnetic moment, thereby converting the magnetic repulsion into magnetic attraction; if alternating current is used, the frequency and phase of the current can be adjusted, and the direction of the magnetic moment can also be directly changed.

[0021] In one embodiment, the resultant force is based on the same or different magnitudes of the currents passed through each of the small free bodies 302 to generate a uniform resultant force or a non-uniform resultant force;

[0022] When the uniform resultant force is generated, the spherical housing 301 and the drill assembly 1 perform the telescoping;

[0023] When the non-uniform resultant force is generated, the spherical shell 301 and the drill assembly 1 perform the telescopic and / or universal adjustment.

[0024] When it is necessary to control the universal angle adjustment of the drill bit assembly 1, the current output by the electrical connector 303 is controlled specifically according to the direction position to be adjusted, affecting one of the small free bodies 302 or the small free bodies 302 adjacent to it separately; thereby generating a non-uniform resultant force and forming the magnetic repulsion between the large free bodies 4.

[0025] That is, the interaction force between the small free body 302 and the large free body 4 will no longer be evenly distributed due to the influence of the magnetic repulsion, and the straight line trajectory will be controlled to move along the planned curved direction. In other words, the direction of the resultant force acting on the spherical shell 301 will no longer move completely along the original straight line movement direction, but will be offset along the direction of the small free body 302 that breaks the symmetry of the large free body 4. If preset control is applied, the universal angle adjustment of the drill bit assembly 1 can be achieved based on the above principle.

[0026] In one embodiment, the body component 2 includes a housing 201 as a large-sized medical universal telescopic drill, and the housing 201 is provided with arc-shaped slide grooves 202 in an annular array, and the spherical housing 301 of the telescopic component 3 is rollingly matched in the arc-shaped slide grooves 202. Furthermore, when the spherical housing 301 is linearly telescopic, universally angled, or both are performed simultaneously, trajectory constraints are performed based on the housing 201 and its arc-shaped slide grooves 202.

[0027] In one embodiment, a detection component 5 for detecting angle information of the drill head component 1 is also installed in the housing 201 .

[0028] The detection component 5 includes a ring 501 fixed in the housing 201, and the ring 501 is equipped with multiple strain gauges 502 in the form of a ring array; when the spherical shell 301 is universally adjusted, the current angle information of the spherical shell 301 can be obtained based on the multiple paddles 3011 carried thereon touching the corresponding strain gauges 502; and because the drill bit assembly 1 and the spherical shell 301 are relatively fixed, the current angle information of the spherical shell 301 is also obtained.

[0029] Second, the control method of the medical universal telescopic drill:

[0030] After locating and obtaining the position information of the tooth pulp opening through the optical positioning system, perform the following steps:

[0031] (1) Step S1, electrical signal calculation:

[0032] The discretized PID controller receives the target value tv (desired magnetic field strength and direction) and the actual feedback value fv (current magnetic field strength and direction) as input; calculates the error E between the target value tv and the feedback value fv, and calculates the preliminary control electrical signal e(t) according to the discretized PID control algorithm:

[0033]

[0034] Among them, E k represents the error between the target value tv and the feedback value fv at the kth sampling moment; K p is the proportionality coefficient, K i is the integration coefficient, K d is the differential coefficient. Represents the cumulative sum of errors from time k to the current time t. E prev Represents the error at the previous time step.

[0035] The method for calculating the error E between the target value tv and the feedback value fv is E=tv-fv.

[0036] (2) Step S2, activate the bidirectional Bi-GRU recurrent neural network:

[0037] The bidirectional Bi-GRU recurrent neural network receives historical error data D and error E as input, and dynamically calculates and updates the error threshold T1 and saturation value T2.

[0038] Whether to activate the integral action of the PID controller is determined according to the error threshold T1, and the accumulated step value CS of the integral term is adjusted according to the saturation value T2 to prevent integral saturation.

[0039] According to the preliminary control electrical signal e(t) of the discretized PID controller and the adjustment of the bidirectional Bi-GRU recurrent neural network, the final control electrical signal e(t)' is generated.

[0040] Among them, the bidirectional Bi-GRU recurrent neural network adopted by the present invention is different from its traditional form. Its core lies in the introduction of the Seq2Seq model to divide the input and output into two parts, the encoding layer and the decoding layer, so as to support the situation where the input and output sequences are of unequal length. At the same time, by taking advantage of the bidirectional Bi-GRU recurrent neural network model, the forward and reverse dependencies in the time series data are captured at the same time, and the accuracy and robustness of the model's dynamic calculation error threshold T1 and saturation value T2 are improved.

[0041] (2.1.1) Step S200, execute the encoder:

[0042] Receive an input sequence S = {D, E}, where D is the historical error data and E is the current error;

[0043] First initialize the forward and backward hidden states:

[0044] Then, through the bidirectional Bi-GRU network, the dependencies in the input sequence S are extracted from both the forward and backward directions to generate the hidden state sequence H. The forward and backward Bi-GRU update methods are:

[0045]

[0046] in, and are the forward and backward hidden state vectors respectively. and They are the forward and backward gated recurrent units respectively.

[0047] (2.1.2) Step S201, ECA attention mechanism (Encoder-Decoder Attention) processing:

[0048] The decoder considers all the hidden states of the encoder output when generating each output and performs a weighted sum of them to produce an intermediate vector C of fixed length i This vector C i The decoder is specific to the i-th time step and is based on the encoder hidden state sequence H and the decoder hidden state s at the previous time step i-1 Calculated.

[0049] Therefore, the hidden state sequence H output by the encoder is weighted and summed to generate the intermediate vector C:

[0050]

[0051] Among them, C i The intermediate vector C of the decoder at the i-th time step; h j is the jth encoder hidden state in the hidden state sequence H, α ij is the attention weight, which can be calculated by the softmax function or by the following calculation method:

[0052]

[0053] e ij is the energy function of the attention mechanism, calculated by a small neural network whose input is the encoder hidden state h j and the decoder’s hidden state s at the previous time step i-1 :

[0054] e ij =a(s i-1 ,h j );

[0055] Where a is a learnable function, such as a single-layer neural network, whose output is a real number representing s i-1 and h j The strength of the correlation between .

[0056] (2.1.3) Step S202, decoder processing:

[0057] Combine the intermediate vector C with the hidden state of the previous time step and And the prediction result y of the decoder at the previous time step t-1 , the prediction result y is gradually generated through the bidirectional Bi-GRU network t =[T1,T2]. The update equation of the decoder is:

[0058]

[0059] Where W and b are the learnable weight matrix and bias vector; [·;·] represents the vector concatenation operation, i.e. Indicates the final hidden state of the forward GRU and the final hidden state of the backward GRU Splice together; t is the output of the current time step, i.e., y t =[T1,T2];. and They are the forward and backward gated recurrent units respectively.

[0060] (2.1.4) Step S203, adjustment:

[0061] Whether to activate the integral action of the PID controller is determined according to the error threshold T1, and the cumulative step value CS of the integral term is adjusted according to the saturation value T2 to prevent integral saturation. The specific method is: based on the error threshold T1 and the saturation value T2, the following constraint terms are introduced into the discretized PID controller algorithm in step S1, and then the final control electrical signal e(t)' is updated.

[0062] 1) Constraint 1: Determine whether to activate the integral action based on the error threshold T1:

[0063]

[0064] Where IA is a flag or Boolean function used to indicate whether the integral action is activated.

[0065] When |e(t)|≥T1, the integral action is activated and IA is set to 1; otherwise, the integral action is not activated and IA is set to 0.

[0066] 2) Constraint 2: Adjust the cumulative step value CS of the integral term according to the saturation value T2. When the absolute value of the error signal is greater than or equal to the saturation value T2, the cumulative step value CS is set to a fixed value ΔCS min , to prevent integral saturation; otherwise, accumulation is performed according to the normal accumulation step value ΔCS:

[0067]

[0068] Among them, ΔCS and ΔCS min They are the normal cumulative step value and the smaller fixed value to prevent integral saturation. In this way, the cumulative step value of the integral term can be dynamically adjusted according to the size of the error signal to prevent integral saturation and improve the performance of the PID controller.

[0069] (3) Step S3, electrical signal output:

[0070] The final control electrical signal e(t)' is sent to the controller. The controller adjusts the output of the electrical connector 303 according to the received control electrical signal, thereby changing the periodic change of the magnetic field of all small free bodies 302 with dipole moments. By changing the periodic change of the magnetic field, the direction and size of the magnetic moment are controlled, thereby achieving the telescopic and universal angle adjustment of the drill assembly 1.

[0071] Thirdly, the application of medical universal telescopic drill:

[0072] The universal telescopic drill bit mentioned above can be widely used in dental surgical instruments, such as medical tooth filling drill bits or electric dental drill robots.

[0073] Compared with the prior art, the present invention has the following beneficial effects:

[0074] 1. High flexibility and controllability: The present invention realizes the telescopic and universal angle adjustment of the drill assembly through electromagnetic control, allowing doctors to more accurately position and operate during surgery, reducing surgical risks. Doctors can adjust the position and angle of the drill assembly in real time according to surgical needs, improving the accuracy of the surgery.

[0075] 2. Compact structure: The design of the main body component and the telescopic component of the present invention makes the overall structure compact, easy to operate and carry. The design of the arc-shaped slide groove provides a stable motion trajectory for the spherical shell, enhancing the rationality of the structure.

[0076] 3. Improve control accuracy and stability: By combining the discretized PID controller and the bidirectional Bi-GRU recurrent neural network, the present invention can more accurately calculate and control the alternating current supplied by the electrical connector 303 to all small free bodies 302 with dipole moments, thereby improving the control accuracy of the magnetic field and the stability of the system. Among them, by utilizing the dynamic computing capability of the bidirectional Bi-GRU recurrent neural network, the integral action and saturation value of the PID controller can be adjusted in real time to adapt to changes in the system state and avoid performance degradation caused by integral saturation.

[0077] 4. Enhance the adaptive ability of the system: The present invention introduces a bidirectional Bi-GRU recurrent neural network to make the automated control of the drill bit more adaptive, better able to cope with external interference and changes in system parameters, and maintain stable control performance. Moreover, by introducing the Seq2Seq model, the input and output can be divided into two parts: the encoding layer and the decoding layer, thereby supporting the situation where the input and output sequences are of unequal lengths. At the same time, by utilizing the advantages of the bidirectional Bi-GRU recurrent neural network model, it is possible to capture the forward and reverse dependencies in the time series data at the same time, and improve the accuracy and robustness of the model's dynamic calculation error threshold T1 and saturation value T2.

[0078] 5. Optimizing the control of the drill bit assembly: By precisely controlling the periodic changes of the magnetic field and the direction of the magnetic moment, the present invention can more accurately achieve the telescopic and universal angle adjustment of the drill bit assembly 1, thereby improving the efficiency and accuracy of the drilling operation. BRIEF DESCRIPTION OF THE DRAWINGS

[0079] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments or technical descriptions will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0080] Figure 1 It is a three-dimensional schematic diagram of the present invention;

[0081] Figure 2 This is a three-dimensional schematic diagram of a half-cut housing of the present invention;

[0082] Figure 3 It is a schematic diagram of the housing and the spherical shell of the present invention in a front view after half-sectioning;

[0083] Figure 4 It is a three-dimensional schematic diagram of the drill bit assembly and the telescopic assembly of the present invention;

[0084] Figure 5 It is a three-dimensional schematic diagram of the present invention cooperating with a dental filling drill bit;

[0085] Figure 6 It is a three-dimensional schematic diagram of the present invention cooperating with an electric dental drill robot;

[0086] Figure 7 It is a schematic diagram of the telescopic assembly of the present invention for telescopic and angle adjustment;

[0087] Figure 8 It is a schematic diagram of finite element deformation simulation of the drill bit assembly of the test example of the present invention under telescopic and angle adjustment;

[0088] Fig. 9 It is a schematic diagram of finite element stress simulation of the drill bit assembly of the test example of the present invention under telescopic and angle adjustment;

[0089] Fig.10 It is a flow chart of the control method of the present invention;

[0090] Fig.11 A schematic diagram of a bidirectional Bi-GRU recurrent neural network implementation flow of step S2 of the control method of the present invention;

[0091] Fig.12 It is a schematic diagram of MAPE comparison between the experimental group and the control group during the training process of Experimental Example 2 of the present invention;

[0092] Fig.13 It is a schematic diagram comparing the vibration effects of the experimental group and the control group of Experimental Example 2 of the present invention.

[0093] Description of the drawings: 1. Drill assembly; 101. Micro motor; 102. Dental drill; 2. Main body assembly; 201. Casing; 2011. Wire slot A; 2012. Wire slot B; 2013. Wire slot C; 2014. Wire slot D; 202. Arc slide; 203. Opening; 3. Telescopic assembly; 301. Spherical shell; 302. Small free body; 303. Electrical connector; 4. Large free body; 5. Detection assembly; 501. Ring; 502. Strain gauge; 6. Dental filling drill; 7. Electric dental drill robot. DETAILED DESCRIPTION

[0094] In order to make the above-mentioned purposes, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the accompanying drawings. In the following description, many specific details are set forth to facilitate a full understanding of the present invention. However, the present invention can be implemented in many other ways different from those described herein, and those skilled in the art can make similar improvements without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below;

[0095] It should be noted that the various embodiments in this specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments, and the same or similar parts between the various embodiments can be referred to each other. For the device disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the relevant parts can be referred to the method part description.

[0096] Explanation of relevant terms:

[0097] The small free body 302 with a dipole moment and the large free body 4 with a dipole moment: hereinafter referred to as the "small free body 302" and the "large free body 4", both of which are magnetically constrained floating bodies, and there is electromagnetic interaction between the two; wherein the magnetic moment of the small free body 302 will generate a resultant force of a non-polar magnetic repulsion effect, and generate magnetic repulsion or magnetic attraction between the small free body 302 and the large free body 4 with a dipole moment. The materials of the two can be selected as follows:

[0098] (1) Permanent magnets: Permanent magnetic materials such as neodymium iron boron (NdFeB), ferrite or aluminum nickel cobalt (AlNiCo). These materials are magnetic even when no electricity is applied, but their magnetic field can be enhanced when electricity is applied.

[0099] (2) Electromagnet: It consists of an iron core and a coil. The iron core can be soft iron or silicon steel, and the coil is a wire wound around the iron core. It generates a magnetic field when power is turned on and disappears when power is turned off.

[0100] Embodiment 1: Figures 1 to 7 As shown, the present embodiment discloses a medical universal telescopic drill. It comprises three main components: a drill assembly 1, a main body assembly 2 and a telescopic assembly 3. The telescopic assembly 3 is designed to realize the telescopic and universal adjustment functions in the main body assembly 2. The telescopic assembly 3 is composed of a spherical shell 301 and a plurality of small free bodies 302, wherein the small free bodies 302 are arranged in an annular array in the spherical shell 301 and are arranged based on the central axis of the large free body 4. The front end of the spherical shell 301 is responsible for carrying the drill assembly 1. In practical applications, by controlling the electromagnetic interaction between the small free bodies 302 and the large free body 4, precise control of the telescopic movement of the spherical shell 301 in the main body assembly 2 can be achieved.

[0101] Specifically, the drill assembly 1 includes a micro motor 101 and a dental drill 102 driven by the micro motor 101. The micro motor 101 is fixedly connected and matched with the spherical housing 301. The matching between the dental drill 102 and the output shaft of the micro motor 101 can adopt a conventional detachable structure.

[0102] Specifically: The core principle of this embodiment is to use electromagnetic interaction to achieve the telescopic and universal adjustment functions of the telescopic component 3. Specifically, the large free body 4 is fixed in the main body component 2 and generates a magnetic repulsion force with the small free body 302. This magnetic repulsion can be precisely controlled to achieve the telescopic and universal adjustment of the telescopic component 3. Since the small free body 302 is arranged in the spherical shell 301 in a ring array and is driven by the electromagnetic interaction with the large free body 4, this enables the spherical shell 301 to perform telescopic and universal adjustment movements in the main body component 2.

[0103] It can be understood that in the above scheme: this design of the medical universal telescopic drill achieves high precision and flexibility in dental surgery. By controlling the telescopic and universal adjustment of the telescopic assembly 3 through electromagnetic interaction, the doctor can more accurately control the position and angle of the drill assembly 1, thereby performing dental surgery more accurately. This design not only improves the accuracy of the surgery, but also reduces the doctor's need for manual adjustment during the surgery, reducing the difficulty and risk of the surgery. In addition, the design also allows the drill assembly 1 to perform complex movements during the surgery, further improving the treatment effect of the dental surgery.

[0104] In the technical solution provided in this embodiment, based on the utilization of the electromagnetic mutual force described above, it can be known that the core of this solution is how to control the small free body 302; for this purpose, please refer to Figure 2-4The telescopic assembly 3 further comprises a plurality of electrical connectors 303. Each electrical connector 303 is electrically connected to a small free body 302 through a wire, and the electrical connector 303 is externally connected to an external power supply and a power supply control system.

[0105] When all electrical connectors 303 supply power to all small free bodies 302 at the same time, the small free bodies 302 will generate the same or different magnetic moments based on the magnitude of the current and voltage generated at the same time. These magnetic moments then generate a resultant force without a polar magnetic repulsion effect, and the magnitude of the resultant force is controllable. When the drill bit assembly 1 needs to be telescopically adjusted, all electrical connectors 303 act as switches to supply the same current and voltage to all small free bodies 302 to generate magnetic moments. In this way, multiple small free bodies 302 will generate a uniform resultant force and generate magnetic repulsion with the large free body 4, thereby controlling the extension of the telescopic assembly 3 as a whole and the drill bit assembly 1 that cooperates with it.

[0106] Specifically, when the drill assembly 1 needs to be retracted, if direct current is used, the direction of the magnetic moment is directly changed by flipping the current direction of all control electrical connectors 303 (such as switching the positive and negative poles), thereby converting the magnetic repulsion into the magnetic attraction. If alternating current is used, the direction of the magnetic moment can be changed by adjusting the frequency and phase of the current.

[0107] Specifically: The principle of this embodiment is based on the basic principle of electromagnetism. By controlling the magnitude of the current and voltage, the magnitude and direction of the magnetic moment generated by the small free body 302 can be changed. When the direction of the magnetic moment generated by all the small free bodies 302 is the same, they will generate a uniform resultant force and generate magnetic repulsion or magnetic attraction with the large free body 4.

[0108] Furthermore, the magnetic repulsion or magnetic attraction can be used to control the telescopic movement of the telescopic component 3. By changing the direction of the current or adjusting the frequency and phase of the current, the direction of the magnetic moment can be changed, thereby controlling the extension or retraction of the telescopic component 3.

[0109] (1) When the small free body 302 is energized, it will generate a magnetic field, and the magnitude and direction of its magnetic moment are related to the current and voltage applied. According to the basic principles of electromagnetism, the current is the cause of the magnetic field, and the voltage determines the magnitude of the current. Therefore, by controlling the magnitude of the current and voltage, the intensity of the magnetic field generated by the small free body 302 can be changed, thereby changing the magnitude of its magnetic moment. At the same time, the direction of the current also determines the direction of the magnetic field, that is, the direction of the magnetic moment. Therefore, by controlling the direction of the current, the direction of the magnetic moment of the small free body 302 can be changed.

[0110] (2) When the directions of the magnetic moments generated by all small free bodies 302 are the same, the magnetic fields they generate will be superimposed in space to form a uniform resultant force field. The direction of this resultant force field is the same as the direction of the magnetic moments of all small free bodies. When this resultant force field meets the large free body 4, magnetic repulsion or magnetic attraction will be generated due to the interaction between the magnetic fields. Which specific force is generated depends on whether the directions of the magnetic moments of the small free body 302 and the large free body 4 are the same. If the directions are the same, magnetic attraction will be generated; if the directions are opposite, magnetic repulsion will be generated.

[0111] (3) Because magnetic repulsion or magnetic attraction is a non-contact force, it can act on an object without contact. In the medical universal telescopic drill bit of this embodiment, this force is used to control the telescopic movement of the telescopic component 3. When the magnetic moments generated by all the small free bodies 302 have the same direction, magnetic repulsion or magnetic attraction is generated between them and the large free body 4. This force can push or pull the telescopic component 3 to produce a telescopic movement. By controlling the magnitude and direction of the current and voltage, the magnitude and direction of the magnetic repulsion or magnetic attraction can be changed, thereby accurately controlling the telescopic movement of the telescopic component 3.

[0112] (4) The direction of the current determines the direction of the magnetic field, that is, the direction of the magnetic moment. Therefore, by changing the direction of the current, the direction of the magnetic moment of the small free body 302 can be directly changed. When using alternating current, the periodic change of the magnetic field can be changed by adjusting the frequency and phase of the current, thereby changing the direction of the magnetic moment. This change in the direction of the magnetic moment will cause a change in the magnetic repulsion or magnetic attraction, thereby controlling the extension or retraction of the telescopic component 3. Specifically, if the magnetic repulsion becomes a magnetic attraction, or the magnetic attraction becomes a magnetic repulsion, or the magnitude and direction of the force change, the motion state of the telescopic component 3 will change.

[0113] It can be understood that in the above scheme: this embodiment provides a high degree of controllability and flexibility of the medical universal telescopic drill. The doctor can accurately control the extension and retraction of the drill assembly 1 by controlling the magnitude and direction of the current and voltage. This control method not only improves the accuracy of the operation, but also allows the doctor to adjust the position and angle of the drill assembly 1 in real time as needed during the operation. In addition, this embodiment also provides a variety of current control strategies (such as direct current and alternating current) to adapt to different surgical needs and operating environments. This makes the medical universal telescopic drill have a wider application prospect and higher clinical value in dental surgery.

[0114] In the technical solution provided in this embodiment, the above solution is based on the current and voltage generated simultaneously when all the electrical connectors 303 simultaneously supply power to all the small free bodies 302, and the extension and retraction of the drill assembly 1 are realized based on the magnitude of the current and voltage generated simultaneously; however, in order to further realize its universal angle adjustment function, please continue to refer to Figure 2-4: The resultant force can be generated based on whether the magnitude of the current passed through each small free body 302 is the same or different. When all the small free bodies 302 are passed through the same magnitude of current, they will generate a uniform resultant force. This uniform resultant force will cause the spherical shell 301 and the drill bit assembly 1 to perform telescopic movement along a straight line trajectory. When the small free bodies 302 are passed through currents of different magnitudes, they will generate a non-uniform resultant force. This non-uniform resultant force can not only cause the spherical shell 301 and the drill bit assembly 1 to perform telescopic movement, but also cause them to perform universal adjustment movement. When it is necessary to control the drill bit assembly 1 to perform universal angle adjustment, the magnitude of the current output by the electrical connector 303 can be specifically controlled according to the direction position to be adjusted, so as to affect a small free body 302 or a small free body 302 adjacent thereto separately. In this way, a non-uniform resultant force can be generated to form a magnetic repulsion force between the large free bodies 4, thereby realizing the universal angle adjustment of the drill bit assembly 1.

[0115] Specifically: The principle of this embodiment is based on the principle of magnetic moment and magnetic repulsion in electromagnetism. When electric current is passed through the small free bodies 302, they will generate magnetic moments and generate magnetic repulsion between them and the large free body 4. By controlling the magnitude of the current passed through the small free bodies 302, the magnitude and direction of the magnetic moments they generate can be changed, thereby changing the magnitude and direction of the magnetic repulsion between them and the large free body 4. When all small free bodies 302 are passed through currents of the same magnitude, the magnetic moments they generate are in the same direction, forming a uniform resultant force, allowing the spherical shell 301 and the drill bit assembly 1 to perform telescopic movements along a straight line trajectory. When different magnitudes of current are passed through the small free bodies 302, the magnetic moments they generate are in different directions, forming a non-uniform resultant force, which can not only enable the spherical shell 301 and the drill bit assembly 1 to perform telescopic movements, but also enable them to perform universal adjustment movements. Among them:

[0116] (1) When the small free bodies 302 are supplied with currents of different magnitudes, the magnetic field strength generated by each small free body will be different, which results in different magnitudes of the magnetic moments generated by them. More importantly, due to the different magnitudes of the currents, the directions of the magnetic moments of some small free bodies will be slightly deflected and no longer completely consistent with the directions of the magnetic moments of other small free bodies. This inconsistency in the direction of the magnetic moments causes the resultant force they form in space to no longer be uniform. In other words, currents of different magnitudes break the originally uniform magnetic field distribution and form a non-uniform resultant force field.

[0117] (2) The formation of non-uniform resultant force means that the distribution of the resultant force field in space is no longer uniform, which causes the magnetic force acting on the telescopic component 3 to no longer be uniform. Due to the non-uniformity of the magnetic force, the telescopic component 3 (spherical shell 301) will be affected by a deflection torque, which will cause it to produce a universal adjustment movement. Specifically, the telescopic component 3 will deflect and adjust along the direction of the distribution of magnetic lines of force in the non-uniform resultant force field until a new equilibrium position is reached. By precisely controlling the magnitude and direction of the current passed through each small free body 302, precise control of the universal adjustment movement of the telescopic component 3 can be achieved. This control method enables the medical universal telescopic drill to perform arbitrary angle adjustment and position adjustment as needed during surgery, thereby improving the accuracy and flexibility of the surgery.

[0118] like Figure 7 As shown, it should be further pointed out that when the annular array of small free bodies 302 is arranged on the spherical shell 301, since the direction and size of each small free body are consistent, their respective dipole moments will strengthen each other. Dipole moment is a vector and has directionality. Therefore, when multiple dipole moments of the same direction and size exist at the same time, they will be added according to the principle of vector superposition to form a larger dipole moment. This larger dipole moment is the basis for forming a resultant force and is also the key to interacting with the large free body 4. Since the small free bodies 302 are evenly arranged on the spherical shell 301, the large dipole moments they form will be symmetrical. This means that the distribution of the resultant force in space is also symmetrical, providing a stable universal adjustment basis for the spherical shell 301. The magnetic field generated by the large free body 4 will interact with the total dipole moment composed of multiple small free bodies 302. By regulating the magnetic field size and charge of any small free body 302 (i.e., controlling the current parameters of the corresponding electrical connector 303), the intensity and direction of this interaction can be changed. This regulation is the key to achieving universal adjustment. When the adjustment of the magnetic field and the amount of charge causes the spherical shell 301 to be subjected to a resultant force of the non-polar magnetic repulsion effect along the housing 201, the resultant force will push the spherical shell 301 to move in the corresponding direction. Since the resultant force is formed based on the principle of vector superposition, it can achieve precise control of the universal adjustment movement of the spherical shell 301 by precisely controlling the magnetic field and the amount of charge of each small free body 302.

[0119] It should be further pointed out that in order to avoid uneven distribution of magnetic repulsion or excessive magnetic moment difference, in practical applications, only a single small free body 302 or other small free bodies 302 adjacent to the small free body 302 should be powered, and any two diagonally arranged small free bodies 302 should not be powered.

[0120] It should be pointed out that in practice, although the spherical shell 301 can also be moved or universally adjusted by using structures such as an external multi-axis servo electric cylinder or a gear system, compared with the large free body 4 and the small free body 302 forming a combined force and pushing the spherical shell 301 to move or universally adjust the angle, the bearings, guide rails, gears and other parts in the mechanical transmission system will wear out during long-term operation, and this wear will lead to a decrease in transmission accuracy, thereby generating cumulative errors. The gap between the transmission parts is also an important source of error, especially in a multi-axis system, where the accumulation of gaps between the axes will significantly affect the overall accuracy. However, by using the large free body 4 and the small free body 302 to directly act on the spherical shell 301 through magnetic force or electric field force, there is no need for a complex mechanical transmission structure, thereby reducing the accumulation of transmission errors, and by accurately controlling the parameters of the large free body and the small free body (such as distance, direction, strength, etc.), the high-precision movement and angle adjustment of the spherical shell can be achieved. This direct action method is less affected by changes in ambient temperature and humidity, and therefore has better environmental adaptability.

[0121] It can be understood that in the above scheme: this embodiment provides a high degree of flexibility and controllability of the medical universal telescopic drill. The doctor can accurately control the telescopic and universal angle adjustment of the drill assembly 1 by controlling the magnitude and direction of the current passed into the small free body 302. This control method not only improves the accuracy of the operation, but also allows the doctor to adjust the position and angle of the drill assembly 1 in real time as needed during the operation.

[0122] Furthermore, the electrical connector 303 is a key component for connecting the small free body 302 (or the large free body 4) with the power source. The following is an explanation of how to connect it with electricity through a wire:

[0123] (1) Wire connection: The electrical connector 303 is provided with a terminal or a socket for connecting a wire. One end of the wire is connected to the terminal or the socket of the electrical connector 303, and the other end is connected to a power source.

[0124] (2) Insulation treatment: To ensure safety, the connection between the wire and the electrical connector 303 should be insulated to prevent short circuit or electric shock. By controlling the switch of the power supply, the power on and off control of the small free body 302 and the large free body 4 can be achieved. The intensity and direction of the magnetic field can be controlled by adjusting the voltage and current parameters of the power supply.

[0125] Furthermore, in order to avoid the influence of magnetic field on human body, the housing 201 can be made of metal or magnetic-isolating material to form a Faraday cage; at the same time, a flexible protective cover should be installed on the opening 203, which is respectively connected to the opening 203 and the end face (non-output shaft) of the micro motor 101, and the protective cover should also be made of anti-magnetic material.

[0126] In the technical solution provided in this embodiment, the above solution realizes the telescopic function and universal angle adjustment function of the spherical shell 301; but in order to further restrict and make its running trajectory controllable, please refer to Figure 2-3 The main body component 2 is mainly composed of a housing 201, which forms the main structure of the entire medical universal telescopic drill. The housing 201 has an annular array of arc-shaped slide grooves 202, which are designed to achieve rolling cooperation with the spherical shell 301 of the telescopic component 3 (i.e., the arc-shaped slide grooves 202 cover the "spherical shell 301"). When the spherical shell 301 performs linear telescoping, universal angle adjustment, or both, its motion trajectory will be constrained and guided by the housing 201 and the arc-shaped slide grooves 202 inside it.

[0127] Specifically, the front of the housing 201 is provided with an opening 203 , whose diameter is larger than the spherical shell 301 ; when the spherical shell 301 moves to the limit point, the universal angle adjustment function thereof is not restricted due to the relatively large opening 203 .

[0128] Specifically: The principle of this embodiment is based on the rolling fit and trajectory constraint principle in mechanics. The design of the arc-shaped slide groove 202 provides a specific motion trajectory for the spherical shell 301, so that it can maintain a stable and precise motion state during the telescopic and universal adjustment process. When the spherical shell 301 rolls in the arc-shaped slide groove 202, its motion trajectory is constrained by the shape of the slide groove, thereby achieving the expected telescopic and universal adjustment movement.

[0129] Specifically: the housing 201 is also provided with a wire slot A2011, a wire slot B2012, a wire slot C2013 and a wire slot D2014, which are used to place and guide the wires of the micro motor 101, the strain gauge 502 described later, the electrical connector 303 (small free body 302) and the large free body 4 respectively;

[0130] It should be pointed out that the wires of the micro motor 101 and the electrical connector 303 should be of sufficient length to prevent the wires from being pulled when the spherical shell 301 is extended or universally adjusted; at the same time, the corresponding wire grooves should also be polished and slightly larger than the diameter of the wires that match them.

[0131] It can be understood that in the above scheme: this embodiment enhances the stability and accuracy of the medical universal telescopic drill. By constraining the motion trajectory of the spherical shell 301 through the housing 201 and the arc-shaped slide groove 202 inside it, it can be ensured that the drill assembly 1 maintains a stable motion state during the telescopic and universal adjustment process, thereby improving the accuracy and safety of the operation. In addition, this design also makes the medical universal telescopic drill more compact and reasonable in structure, and improves its overall performance and service life.

[0132] In the technical solution provided in this embodiment, please refer to Figure 2-3 : A detection component 5 for detecting the angle information of the drill bit assembly 1 is also installed in the housing 201. The detection component 5 is mainly composed of a collar 501 and a plurality of strain gauges 502. The collar 501 is fixed in the housing 201 and carries a plurality of strain gauges 502 in the form of a ring array. When the spherical shell 301 is universally adjusted in angle, the plurality of paddles 3011 carried thereon will touch the corresponding strain gauges 502, thereby triggering the strain gauges 502 to generate electrical signals. These electrical signals can be used to obtain the current angle information of the spherical shell 301. Since the drill bit assembly 1 and the spherical shell 301 are relatively fixed, the current angle information of the drill bit assembly 1 is also obtained.

[0133] It is understandable that the above scheme is based on the working principle of the strain gauge 502 and the mechanical touch trigger mechanism. The strain gauge is a micro sensor that can convert mechanical strain into an electrical signal, and is the preferred choice under the scene volume limitation of this embodiment. When the paddle 3011 on the spherical shell 301 touches the strain gauge 502, the strain gauge 502 will deform and generate corresponding electrical signals. These electrical signals are related to the position and strength of the touch, and can therefore be used to obtain the angle information of the spherical shell 301.

[0134] It should be noted that the strain gauge 502 can provide a hardware basis for obtaining the target value tv and the feedback value fv of the fourth embodiment. For example, after the current angle information of the spherical shell 301 is obtained using the current strain gauge 502, the difference between the current magnetic field strength and direction and the target value tv is calculated and converted based on the magnetic field strength and direction corresponding to the target value tv, so as to achieve the effect of obtaining the feedback value fv based on the angle information of the current strain gauge 502.

[0135] In the above technical solution provided in this embodiment, all the above electrical components can be controlled by a controller and intelligently driven using a PID controller algorithm.

[0136] Embodiment 2: Figure 5 As shown, this embodiment discloses a medical dental filling drill, comprising a dental filling drill 6 and the medical universal telescopic drill disclosed in the above-mentioned embodiment 1;

[0137] By connecting the medical universal telescopic drill to a traditional handheld dental filling drill 6 and installing a handle switch on the outside of the dental filling drill 6, the power supply and current direction of each small free body 302 can be controlled accordingly, so that the handheld dental filling drill 6 can be realized. Through the telescopic function and universal angle adjustment function of the drill assembly 1 of the medical universal telescopic drill, dental related surgical operations such as drilling, filling or implanting can be achieved.

[0138] Preferably, the handle switch can be a shaking handle switch, which can be shaken to a corresponding angle to activate the power supply of the corresponding small free body 302; at the same time, an independent switch can be installed externally or a specific mode can be set for the shaking handle switch to switch the current direction.

[0139] Embodiment 3: Figure 6 As shown, this embodiment discloses an automated tooth filling robot, including an electric dental drill robot 7 in the form of a multi-axis mechanical arm;

[0140] By connecting the medical universal telescopic drill to the electric dental drill robot 7 in the form of a traditional multi-axis robotic arm and using it as its end effector, and through the optical positioning system of the existing electric dental drill robot 7, combined with the calculation and control of the electrical signals of each small free body 302, dental-related surgical operations such as drilling, filling or implanting can be achieved.

[0141] Specifically, the calculation and control of the electrical signal of each small free body 302 is realized based on the control of an external or built-in controller of the electric dental drill robot 7 .

[0142] Embodiment 4: In Embodiment 1, it is pointed out that a PID controller can be used to control a medical universal telescopic drill. However, when the output of the PID controller reaches saturation, that is, when it reaches its maximum or minimum value, it is easy to produce integral saturation, which leads to a decrease in system performance; at the same time, the PID controller is prone to overshoot under integral saturation, that is, forming a static error in the position adjustment of the drill assembly 1. Therefore, if Figure 10-11 As shown, this embodiment discloses a control method for a medical universal telescopic drill, which can be carried out by an automated tooth filling robot as disclosed in the third embodiment as a hardware carrier. After locating and obtaining the position information of the tooth pulp opening through the optical positioning system, the following steps S1 to S3 are performed:

[0143] In this embodiment, regarding step S1, electrical signal calculation: the discretized PID controller receives two main input parameters: the target value tv (desired magnetic field strength m' and direction d', expressed as tv = [m', d']) and the actual feedback value fv (current magnetic field strength m and direction d, expressed as fv = [m, d]). These two parameters are the basis for the controller to perform subsequent calculations and controls.

[0144] Next, the controller calculates the error E between the target value tv and the feedback value fv. This step is the core of the PID control algorithm, because the error E will directly determine the control electrical signal e(t) output by the controller. The calculation method of the error E is E=tv-fv, which represents the difference between the target value and the current actual value. After obtaining the error E, the preliminary control electrical signal e(t) can be calculated according to the discretized PID control algorithm. This calculation process combines the three control terms of proportion, integration and differentiation:

[0145]

[0146] in, Represents the cumulative sum of errors from time k to the current time t. E prev Represents the error of the previous time step. K p is the proportionality coefficient, which determines the direct influence of the error E on the control electrical signal e(t); K i is the integral coefficient, which determines the impact of the accumulation of past errors on the control electrical signal; K d is the differential coefficient, which determines the effect of the error change rate on the control electrical signal, through (EE prev ) (the difference between the current error and the error of the previous time step). k Represents the error between the target value and the actual value at the kth sampling moment (or time step); k is an index or counter used to identify a specific point in the time series.

[0147] However, the above-mentioned discretized PID controller still has the phenomenon of system performance degradation caused by static error and integral saturation. To this end, the present embodiment aims to use a bidirectional Bi-GRU recurrent neural network to dynamically calculate and update the error threshold, and when the discretized PID controller error is reduced to a certain extent, activate the integral action of the PID controller to eliminate the static error. Because the bidirectional Bi-GRU recurrent neural network has the ability to process time series data, the error threshold can be dynamically adjusted according to historical error data, thereby more intelligently controlling the activation timing of the integral action. At the same time, the present embodiment also aims to use a bidirectional Bi-GRU recurrent neural network to dynamically calculate and adjust the cumulative step value of the integral term when the controller output reaches saturation to prevent the system performance degradation caused by integral saturation. Because the bidirectional Bi-GRU recurrent neural network can dynamically adjust the cumulative step value of the integral term according to the output state and historical data of the controller, thereby avoiding the occurrence of integral saturation.

[0148] To achieve this goal, step S2 is further introduced in this embodiment (see Fig.11 ), activate the bidirectional Bi-GRU recurrent neural network: receive historical error data D and error E as input, dynamically calculate and update the error threshold T1 and saturation value T2. Determine whether to activate the integral action of the PID controller according to the error threshold T1, and adjust the cumulative step value CS of the integral term according to the saturation value T2 to prevent integral saturation. Generate the final control electrical signal e(t)' based on the preliminary control electrical signal e(t) of the discretized PID controller and the adjustment of the bidirectional Bi-GRU recurrent neural network.

[0149] Among them, the above-mentioned bidirectional Bi-GRU recurrent neural network is different from its traditional form. Its core lies in the introduction of the Seq2Seq model to divide the input and output into two parts: the encoding layer (hereinafter referred to as the encoder) and the decoding layer (hereinafter referred to as the decoder), so as to support the situation where the input and output sequences are of unequal length. At the same time, the advantages of the bidirectional Bi-GRU recurrent neural network model are utilized to improve the accuracy of the dynamic calculation of the error threshold T1 and the saturation value T2. At the same time, the advantages of the bidirectional Bi-GRU recurrent neural network model are utilized to simultaneously capture the forward and reverse dependencies in the time series data. The neural network can more comprehensively understand the changing trend of the error, thereby more accurately calculating the error threshold T1 and the saturation value T2. The above-mentioned step S2 specifically includes the following steps S200 to S203.

[0150] Specifically, step S200 executes the encoder: receiving an input sequence S = {D, E}, where D is historical error data and E is the current error; wherein the historical error data D is a sequence that contains error data over a period of time in the past. These data are calculated by the controller in the previous time step and reflect the past performance of the system. The historical error data D can be a sequence of fixed length or a sequence of dynamic length. In the case of fixed length, D contains error data for the most recent N time steps, where N is a predefined constant. In the case of dynamic length, D contains all error data from a specific time point to the current time point.

[0151] For example, assume that the error is calculated every 1 second, and the error data of the last 5 time steps are selected as the historical error data D. Then, at a certain time t, D = {ε(t-5),ε(t-4),ε(t-3),ε(t-2),ε(t-1)}; where ε(ti) represents the error at i seconds before time t.

[0152] For example, if all error data from the start of system operation to the current time t is selected as historical error data D, and the system has been running for 10 seconds, then D = {ε(1),ε(2),ε(3),ε(4),ε(5),ε(6),ε(7),ε(8),ε(9),ε(10)}; where ε(i) represents the error at the i-th second of system operation.

[0153] Wherein, step S200 includes:

[0154] First initialize the forward and backward hidden states:

[0155] Then, through the bidirectional Bi-GRU network, the dependencies in the input sequence S are extracted from both the forward and backward directions to generate the hidden state sequence H. The forward and backward Bi-GRU update methods are:

[0156]

[0157] in, and are the forward and backward hidden state vectors respectively. and The bidirectional Bi-GRU network extracts the dependencies in the input sequence S from both the forward and backward directions.

[0158] The specific operation method is as follows: for each element s in the input sequence S t , do the following:

[0159] 1) Forward Bi-GRU update: calculate the update gate and reset gate These two gating signals determine how much information at the current time step should be retained or forgotten; at the same time, the candidate hidden state is calculated That is, the potential hidden state of the current time step, but it also needs to be filtered through the update gate. Finally, the forward hidden state of the current time step is updated by combining the hidden state of the previous time step with the candidate hidden state of the current time step

[0160]

[0161] 2) Backward Bi-GRU update: Similar to the forward update, but in the opposite direction. Calculate the backward update gate and reset gate and candidate hidden states Finally update the backward hidden state of the current time step

[0162] 3) The encoder finally outputs a hidden state sequence H: This sequence contains the forward and backward hidden states of each element in the input sequence S. These hidden states capture the dependencies in the input sequence and will be used for subsequent dynamic calculations and adjustment operations:

[0163]

[0164] In step S200, σ is a sigmoid activation function, which is used to map the input to the interval (0, 1); tanh is a hyperbolic tangent activation function, which is used to map the input to the interval (-1, 1); ⊙ is an element-wise multiplication operator, which indicates the multiplication of corresponding elements of two vectors or matrices; W z , W r and W h are three different weight matrices, which are the parameters that the model needs to learn; b z 、b r and b h is the bias vector, which is also the parameter that the model needs to learn; and Represents a vector concatenation operation, which takes the input vector s t and the hidden state vector or Splice into a longer vector;

[0165] is the activation value of the forward update gate at time step t, which is used to control the update degree of the forward hidden state;

[0166] is the activation value of the forward reset gate at time step t, which is used to control the degree of reset of the forward hidden state;

[0167] is the value of the forward candidate hidden state at time step t, and is the possible update value of the forward hidden state;

[0168] is the value of the forward hidden state at time step t, which is the result of the encoder forward propagation;

[0169] is the activation value of the backward update gate at time step t, which is used to control the update degree of the backward hidden state;

[0170] is the activation value of the backward reset gate at time step t, which is used to control the degree of reset of the backward hidden state;

[0171] is the value of the backward candidate hidden state at time step t, and is the possible update value of the backward hidden state;

[0172] is the value of the backward hidden state at time step t, which is the result of the encoder's backward propagation.

[0173] It can be understood that by extracting the dependencies in the input sequence through the bidirectional Bi-GRU network, the encoder can more accurately understand the changing trend of the error. This helps to dynamically calculate a more accurate error threshold, thereby improving control accuracy. When the controller output is close to saturation, dynamically adjusting the accumulated step value of the integral term using the bidirectional Bi-GRU network can prevent integral saturation. This helps to avoid system performance degradation and instability. The encoder provides rich information for dynamic calculation and adjustment by capturing the forward and reverse dependencies in the input sequence. This enables the control system to adapt to different working conditions and requirements more intelligently.

[0174] Specifically, step S201, ECA attention mechanism (Encoder-Decoder Attention) processing: The core principle of the ECA attention mechanism is to process the information interaction between the encoder and the decoder described later. The core of this mechanism is that when the decoder generates each output, it will consider all the hidden states of the encoder output and perform a weighted sum on them to generate an intermediate vector C of fixed length i . Because the intermediate vector C i Based on the encoder hidden state sequence H and the decoder hidden state s at the previous time step i-1 Calculated, so this vector is specific to the i-th time step of the decoder. Furthermore, the ECA attention mechanism allows the decoder to dynamically focus on different parts of the encoder output when generating each output. By calculating the strength of the correlation between the encoder hidden state and the decoder hidden state of the previous time step, a set of attention weights representing the importance of the encoder hidden state at the decoder's current time step is obtained. The execution logic of the ECA attention mechanism includes:

[0175] First, perform weighted summation on the hidden state sequence H output by the encoder to generate an intermediate vector C:

[0176]

[0177] Among them, C i The intermediate vector C of the decoder at the i-th time step; h j is the jth encoder hidden state in the hidden state sequence H, α ij is the attention weight, which indicates that at the i-th time step of the decoder, the encoder hidden state h j The importance of can be calculated by the softmax function, or by the following calculation method:

[0178]

[0179] e ijis the energy function of the attention mechanism, calculated by a small neural network whose input is the encoder hidden state h j and the decoder’s hidden state s at the previous time step i-1 :

[0180] e ij =a(s i-1 ,h j );

[0181] Where a is a learnable function, such as a single-layer neural network, whose output is a real number representing s i-1 and h j The strength of the correlation between .

[0182] Putting these together, the complete ECA attention mechanism process is expressed as:

[0183]

[0184] It can be understood that by dynamically paying attention to different parts of the encoder output, the decoder can more accurately understand the contextual information of the input sequence, thereby generating a more precise output. This helps to improve the control accuracy and stability of the PID controller. Since the ECA attention mechanism can dynamically adjust the degree of attention to the encoder hidden state, when the input sequence changes, the model can adapt to such changes more quickly and generate corresponding outputs. It helps to enhance the adaptability of the PID controller to complex environments. When the controller output reaches saturation, the ECA attention mechanism can dynamically calculate and adjust the accumulated step value of the integral term by dynamically adjusting the degree of attention to the encoder hidden state, which helps to prevent the degradation of system performance caused by integral saturation.

[0185] Specifically, step S202, decoder processing: combining the intermediate vector C and the hidden state of the previous time step and And the prediction result y of the decoder at the previous time step t-1 , the prediction result y is gradually generated through the bidirectional Bi-GRU network t =[T1,T2]. The update equation of the decoder is:

[0186]

[0187] Where W and b are the learnable weight matrix and bias vector; y t is the output of the current time step, i.e., y t =[T1,T2];[·;·] represents a vector concatenation operation. and They are forward and backward gated recurrent units respectively. The specific operation method is similar to step S200. The process is as follows:

[0188] 1) Calculate the input x of the current time step t :x t =[y t-1 ; C]; that is, input x t is the input of the current time step, which is composed of the prediction result yt-1 of the previous time step and the intermediate vector C;

[0189] 2) Calculate the hidden state h of the current time step through the Bi-GRU network t :

[0190] 2.1) Forward GRU calculation:

[0191]

[0192] 2.2) Backward GRU calculation:

[0193]

[0194] 2.3) Combine the forward and backward GRU to get the final hidden state:

[0195] Among them, x t is the input of the current time step. ht-1 and ht+1 are the hidden states of the previous time step and the next time step for the backward GRU, respectively. σ is the sigmoid activation function. tanh is the hyperbolic tangent activation function. ⊙ represents the multiplication between elements. W xz , W hz , W xr , W hr , W xh and W hh are six different weight matrices. z 、b r and b h is the bias vector. and They are the update gate, reset gate, candidate hidden state and final hidden state of the forward GRU. and They are the update gate, reset gate, candidate hidden state and final hidden state of the backward GRU. It means concatenating the final hidden states of the forward and backward GRU to form the final hidden state h t .

[0196] 3) Calculate the input o (of the output layer) t :o t =W o ht +b o ;

[0197] 4) Finally, apply the sigmoid activation function σ to get the predicted output

[0198] Calculate the error threshold T1 and saturation value T2:

[0199]

[0200] Among them, t1 and t2 are intermediate variables for calculating the error threshold T1 and the saturation value T2, which are obtained through linear transformation. That is, the hidden state ht is respectively transformed with the weight matrix W t1 and W t2 Multiply and add the bias vector b t1 and b t2 , and obtain intermediate variables t1 and t2. Activate t1 and t2 through activation function A (such as ReLU activation function) to obtain error threshold T1 and saturation value T2.

[0201] It can be understood that the decoder's processing is based on a bidirectional Bi-GRU network, which can consider both forward and backward information at the same time, thereby more accurately capturing the contextual information in the sequence. By combining the intermediate vector C and the state of the previous time step, the decoder can gradually generate prediction results while dynamically calculating the error threshold T1 and saturation value T2. Because the calculation of the error threshold T1 and saturation value T2 is based on the hidden state h of the current time step t , so they are dynamically calculated based on the input sequence and the state of the previous time step. This dynamic calculation method enables the PID controller to adjust its control strategy more flexibly to adapt to different control scenarios and input sequences.

[0202] At the same time, by dynamically calculating and updating the error threshold T1 and saturation value T2 through the bidirectional Bi-GRU network, the PID controller can track the target value more accurately and maintain stability in different control scenarios. This is because the bidirectional Bi-GRU network can capture the contextual information in the sequence, thereby generating more accurate prediction results and error thresholds. When the controller output reaches saturation, the bidirectional Bi-GRU network is used to dynamically calculate and adjust the cumulative step value of the integral term (here it is achieved by dynamically adjusting the saturation value T2), which can prevent the degradation of system performance caused by integral saturation. Because the dynamic calculation of the saturation value T2 enables the PID controller to adjust its control strategy when the output is saturated to avoid performance degradation caused by excessive accumulation of the integral term. Since the bidirectional Bi-GRU network can capture long-distance dependencies in the sequence, when the input sequence changes, the decoder can adapt to this change more quickly and generate corresponding prediction results and error thresholds. This enables the PID controller to better cope with complex and changing control environments.

[0203] Specifically, step S203, adjustment: judging whether to activate the integral action of the PID controller according to the error threshold T1, and adjusting the cumulative step value CS of the integral term according to the saturation value T2 to prevent integral saturation. The specific method is: introducing the following constraint term to the discretized PID controller algorithm in step S1 based on the error threshold T1 and the saturation value T2, and then updating the final control electrical signal e(t)'.

[0204] 1) Constraint 1: Determine whether to activate the integral action based on the error threshold T1:

[0205]

[0206] Among them, IA is a flag or Boolean function used to indicate whether the integral action is activated. When |e(t)|≥T1, the integral action is activated and IA is set to 1; otherwise, the integral action is not activated and IA is set to 0.

[0207] It is understandable that the integral action is used in the PID controller to eliminate static errors and improve control accuracy. However, in some cases, such as when the error is small or the system is close to stability, the integral action may not need to be activated to avoid system overshoot or oscillation caused by excessive integration. Therefore, by setting the error threshold T1, the activation of the integral action can be flexibly controlled to improve the stability and accuracy of the control. By dynamically activating or deactivating the integral action according to the error size, unnecessary integral accumulation can be avoided, the risk of system overshoot or oscillation can be reduced, and the stability and accuracy of the control can be improved.

[0208] 2) Constraint 2: Adjust the cumulative step value CS of the integral term according to the saturation value T2. When the absolute value of the error signal is greater than or equal to the saturation value T2, we set the cumulative step value CS to a fixed value ΔCS min , to prevent integral saturation; otherwise, accumulation is performed according to the normal accumulation step value ΔCS:

[0209]

[0210] Among them, ΔCS and ΔCS min They are the normal cumulative step value and the smaller fixed value to prevent integral saturation. In this way, the cumulative step value of the integral term can be dynamically adjusted according to the size of the error signal to prevent integral saturation and improve the performance of the PID controller.

[0211] It is understandable that integral saturation is a common problem in PID controllers. When the error exists for a long time and the integral action continues to accumulate, the integral term may reach saturation, resulting in the controller output being unable to increase or decrease further, thus affecting the control effect. By setting the saturation value T2 and a small fixed cumulative step value ΔCS min , when the error is large, the accumulation speed of the integral term can be limited to prevent integral saturation; when the error is small, the accumulation is performed according to the normal cumulative step value to maintain the sensitivity of the control, which can effectively prevent the occurrence of integral saturation and avoid the problem that the controller output reaches saturation and cannot be further adjusted. It helps to improve the performance and stability of the PID controller, especially when facing large errors or system dynamic changes, it can better control.

[0212] In this embodiment, regarding step S3, electrical signal output: the final control electrical signal e(t)' is sent to the controller. The controller adjusts the output of the electrical connector 303 according to the received control electrical signal, thereby changing the periodic change of the magnetic field of all small free bodies 302. By changing the periodic change of the magnetic field, the direction and size of the magnetic moment are controlled, thereby achieving the telescopic and universal angle adjustment of the drill assembly 1.

[0213] In this embodiment, the Python execution program of the above steps S203 to S3 is as follows:

[0214] import serial

[0215] importtime

[0216] #PID controller parameters

[0217] K_p=1.0#Proportional coefficient

[0218] K_i=0.1#Integral coefficient

[0219] K_d=0.05#differential coefficient

[0220] #Constraint Parameters

[0221] T1=5.0#Error threshold

[0222] T2=10.0#saturation value

[0223] Delta_CS_min=0.01#Small fixed value to prevent integral saturation

[0224] Delta_CS=0.1#Normal cumulative step value

[0225] # Initialize variables

[0226] e_prev=0.0#Last error

[0227] sum_E=0.0#The sum of errors

[0228] integral_active=False# Flag indicating whether the integral action is activated

[0229] CS=0.0#Accumulated step value of integral term

[0230] #Serial communication settings

[0231] ser=serial.Serial('COM1',9600,timeout=1)#Open the serial port and set the bit rate to 9600

[0232] time.sleep(2)#Wait for the serial port connection to stabilize

[0233] defcalculate_pid(e_t):

[0234] global sum_E,integral_active,CS,e_prev

[0235] #Constraint 1: Determine whether to activate the integral effect

[0236] ifabs(e_t)>=T1:

[0237] integral_active=True

[0238] else:

[0239] integral_active=False

[0240] #Constraint 2: Adjust the cumulative step value of the integral term

[0241] ifabs(e_t)>=T2 and integral_active:

[0242] CS=CS+Delta_CS_min

[0243] elifabs(e_t) <T2 and integral_active:

[0244] CS=CS+Delta_CS

[0245] elifnot integral_active:

[0246] CS=CS#Do not change the value of CS

[0247] #Calculate the output of the PID controller

[0248] sum_E+=CS*e_t# Update the cumulative sum of errors, where CS is used as the step value p_term=K_p*e_t

[0249] i_term=K_i*sum_E

[0250] d_term = K_d*(e_t-e_prev)

[0251] e_prev=e_t#Update the last error

[0252] #Calculate the final control electrical signal

[0253] control_signal=p_term+i_term+d_term

[0254] return control_signal

[0255] #Get the error

[0256] defget_error():

[0257] return-2.5#error value

[0258] #Main loop

[0259] try:

[0260] while True:

[0261] error=get_error()#Get the current error

[0262] control_signal = calculate_pid(error) # calculate PID control signal

[0263] #Send control signal to PLC controller

[0264] ser.write(str(control_signal).encode())#Send control signal

[0265] ser.write(b'\n')#Send a newline character as a sign of the end of the message

[0266] time.sleep(0.1)#wait for a while and then loop again

[0267] except KeyboardInterrupt:

[0268] ser.close()#Close the serial port connection

[0269] print("The program has stopped")

[0270] In the above program, the three parameters of the PID controller are first defined: the proportional coefficient K_p, the integral coefficient K_i, and the differential coefficient K_d. Then, it implements a calculate_pid function, which accepts the current error as input and calculates the control signal according to the PID formula. Inside the calculate_pid function, the program implements two constraints. The first constraint determines whether the integral action is activated based on the error threshold T1. The second constraint adjusts the accumulated step value CS of the integral term based on the saturation value T2 and whether the integral action is activated. The program uses the serial library to communicate with the PLC controller via the serial port. It first opens the serial port connection and sets the bit rate to 9600. Then, in the main loop, the program calculates the PID control signal and sends it to the PLC controller through the serial port. After each transmission, it waits for a period of time (0.1 seconds) to loop again. When the user presses a keyboard interrupt (such as Ctrl+C), the program closes the serial port connection and exits.

[0271] Embodiment 5: This embodiment discloses a training method for the bidirectional Bi-GRU recurrent neural network model of Embodiment 4, including:

[0272] P1. Collect data set D: D = [S1, S2, ..., S n ]; among them, S i is the historical data of the ith input sequence (as described in Example 1), and n is its number. Then the preprocessed data set is divided into the training set D train and the test set D test ;

[0273] P2. Set training parameters: including learning rate, batch size, number of iterations, and Adam optimizer.

[0274] P3. Training model: using training set D train Train the model and update the model parameters (weight matrix W) through the back propagation algorithm. z , W r , W h , W xz , W hz , W xr , W hr , W xh and W hh , and bias b z 、b r and b h ) to minimize the MAPE value between the predicted value and the actual value, and perform parameter tuning and model structure adjustment as needed.

[0275] P4, model evaluation and deployment: Repeat forward propagation, loss calculation and back propagation, and continuously iterate the training process until the stopping condition is met (the maximum number of iterations is reached or the loss function value is less than the threshold).

[0276] In this embodiment, the Python execution program of the bidirectional Bi-GRU recurrent neural network model described in Example 4 and the training method described in this embodiment is as follows:

[0277] import numpy as np

[0278] importtensorflow as tf

[0279] from tensorflow.keras.models import Model

[0280] from tensorflow.keras.layers import Input,GRU,Dense,Concatenate,TimeDistributed

[0281] from tensorflow.keras.optimizers importAdam

[0282] #Seq2Seq model structure - encoder

[0283] defencoder(input_dim,units):

[0284] inputs = Input(shape=(None, input_dim))

[0285] encoded = Bidirectional(GRU(units, return_sequences=True, return_state=True))(inputs)

[0286] encoder_outputs, forward_h, forward_c, backward_h, backward_c = encoded

[0287] state = Concatenate()([forward_h, backward_h])

[0288] return Model(inputs, [encoder_outputs, state])

[0289] # Seq2Seq model structure - Attention mechanism + Decoder

[0290] def decoder(units, output_dim):

[0291] state_input_h = Input(shape=(units,))

[0292] state_input_c = Input(shape=(units,))

[0293] decoder_inputs = Input(shape=(None, output_dim))

[0294] decoder_gru = GRU(units, return_sequences=True, return_state=True)

[0295] decoder_outputs, _, _ = decoder_gru(decoder_inputs, initial_state=[state_input_h, state_input_c])

[0296] decoder_dense = Dense(output_dim, activation='sigmoid')

[0297] decoder_outputs = decoder_dense(decoder_outputs)

[0298] return Model([decoder_inputs] + [state_input_h, state_input_c], [decoder_outputs])

[0299] def seq2seq_model(input_dim, units, output_dim):

[0300] encoder_model = encoder(input_dim, units)

[0301] decoder_model = decoder(units, output_dim)

[0302] # Encoder model inference

[0303] encoder_inputs = Input(shape=(None, input_dim))

[0304] encoder_outputs, state = encoder_model(encoder_inputs)

[0305] # Decoder model inference

[0306] decoder_state_input_h = Input(shape=(units,))

[0307] decoder_state_input_c = Input(shape=(units,))

[0308] decoder_outputs = decoder_model([encoder_outputs] +

[0309] [decoder_state_input_h, decoder_state_input_c])

[0310] # Seq2Seq model

[0311] model=Model([encoder_inputs]+[decoder_state_input_h,decoder_state_input_c],decoder_outputs)

[0312] return model

[0313] #Training-P1

[0314] #X and Y are preprocessed input and output data

[0315] from sklearn.model_selection import train_test_split

[0316] X_train,X_test,Y_train,Y_test=train_test_split(X,Y,test_size=0.2,random_state=42)

[0317] #Training-P2

[0318] learning_rate=0.001

[0319] batch_size=64

[0320] epochs=100

[0321] optimizer=Adam(learning_rate=learning_rate)

[0322] #Training-P3

[0323] model=seq2seq_model(input_dim=input_dim, units=units, output_dim=output_dim)

[0324] model.compile(optimizer=optimizer,

[0325] loss='mean_absolute_percentage_error')

[0326] model.fit([X_train,np.zeros((len(X_train),units)),np.zeros((len(X_train),units))],Y_train,batch_size=batch_size,epochs=epochs,validation_split=0.2)

[0327] #Training-P4

[0328] loss=model.evaluate([X_test,np.zeros((len(X_test),units)),np.zeros((len(X_test),units))],Y_test)

[0329] print(f"Test MAPE:{loss}")

[0330] #Model deployment (save model)

[0331] model.save('seq2seq_model.h5')

[0332] Embodiment 6: This embodiment further provides a method for implementing the technical solution of the small neural network involved in step S201 in embodiment 4:

[0333] In step S201, the energy function e of the attention mechanism ij is computed by a small neural network (here, a learnable function a) that measures the hidden state s of the decoder at the previous time step i-1 The strength of the correlation between and the encoder hidden state hj.

[0334] First, define the learnable function a as a single-layer neural network whose input is the hidden state s i-1 and the hidden state h j The concatenation of is a real number indicating the strength of the correlation between them. Then the function is:

[0335] a(s i-1 ,h j )=Wtanh(V[s i-1 ;h j ]+b);

[0336] Where W and V are weight matrices, b is the bias vector, [s i-1 ; hj] represents the hidden state s i-1 and the hidden state h jThe concatenated vector of , tanh is the activation function used to introduce nonlinearity.

[0337] Then use this function to calculate the energy function e ij :e ij =a(s i-1 ,h j );

[0338] This energy function represents the decoder hidden state s at the i-th time step i-1 and the encoder hidden state h j The higher this value is, the stronger the correlation between the two is, that is, the more important the encoder hidden state hj is to the output of the decoder at the current time step. Finally, the softmax function is used to calculate the attention weight α ij .

[0339] The principle of the above scheme is based on the attention mechanism, which allows the decoder to dynamically focus on different parts of the encoder output when generating each output. By calculating the correlation strength between the encoder hidden state and the decoder hidden state in the previous time step, a set of attention weights can be obtained, which represent the importance of the encoder hidden state at the decoder's current time step. These weights are then used to weight the encoder hidden state to obtain an intermediate vector of fixed length, which contains all the relevant information of the encoder output and is specific to the current time step of the decoder. In this way, the decoder can pay more attention to the encoder hidden state related to the current output, thereby improving control accuracy and stability.

[0340] Test Example 1: This example aims to verify the performance of the medical universal telescopic drill bit described in Example 1 under specific working conditions, especially the drilling effect of the dental drill and the stability of the system, by performing three-dimensional modeling and simulation analysis on the drill bit.

[0341] (I) Experimental materials and methods:

[0342] 1.1 3D Modeling:

[0343] Use Autodesk Inventor software to perform 3D modeling of the medical universal telescopic drill.

[0344] The dental drill 102 is given diamond material parameters, the Mohs hardness is set to 10 degrees, and the weight is set to 20g.

[0345] Autodesk Inventor software was used to build a tooth model similar in size to an adult tooth, with a length of 0.8-1.2 cm, a root length of 0.8-1.5 cm, and a width of 5-10 mm. Since there is no material configuration for enamel in the software, topaz with a Mohs hardness of 6.5 was used instead; at the same volume, the Mohs hardness (5-7 degrees) and mass (5 grams) of topaz and enamel are basically the same.

[0346] 1.2 Assembly and simulation settings:

[0347] In the assembly environment of Autodesk Inventor software, the finite element module provided by the software is used for simulation.

[0348] The rotation speed of the micro motor 101 is set to 360 rpm.

[0349] The total weight of the drill bit assembly 1, the spherical housing 301 and the small free body 302 is set to 86 g.

[0350] The small free body 302 and the large free body 4 are provided with an aluminum nickel cobalt material. Based on their volume restrictions, the software automatically configures the small free body 302 and the large free body 4 to have a magnetic repulsion of 1.74539N.

[0351] 1.3 Simulation process:

[0352] After the tooth model is fixed, the dental drill 102 of the medical universal telescopic drill is rotated to contact the surface of the tooth model. The dental drill 102 is set to move 0.5 mm and adjust the angle of 5° along its axis in one second to simulate the strain effect of the tooth model and the axial position change of the dental drill 102.

[0353] (II) Test results:

[0354] 2.1 Change of drill position:

[0355] like Figure 8 As shown in Table 1, the axial position variation of the dental drill 102 is only ±0.0000165016 mm, and this degree of position variation is basically negligible in practice. The variation of other positions is also basically less than this parameter, and the details can be seen in Table 1.

[0356]

[0357] Table 1. Summary chart of strain results generated by finite element simulation

[0358] 2.2 Effect of drilling holes in tooth model:

[0359] like Fig. 9As shown in the figure, holes are obviously drilled in the tooth model. Specific stress parameters can be found in Table 1.

[0360] (III) Conclusion:

[0361] This example verifies the excellent performance of the medical universal telescopic drill under specific working conditions. The position of the dental drill 102 changes very little during the drilling process, indicating that it has high stability and precision. At the same time, the tooth model was successfully drilled out of the hole, further proving the effectiveness and practicality of the drill. This experiment provides strong data support for subsequent research and development and application.

[0362] Experimental Example 2: This example aims to verify the effectiveness of the Seq2Seq model (as described in Example 4) combined with the bidirectional Bi-GRU recurrent neural network for intelligent adjustment of the discrete PID controller by comparing the performance of the experimental group and the control group in controlling the medical universal telescopic drill (as described in Example 1), so as to achieve precise control of the medical universal telescopic drill.

[0363] (I) Experimental and control group settings:

[0364] 1.1 Experimental group:

[0365] The Seq2Seq model combined with the bidirectional Bi-GRU recurrent neural network is used to intelligently adjust the discrete PID controller. The error threshold T1 and saturation value T2 are dynamically adjusted to achieve precise control of the medical universal telescopic drill.

[0366] 1.2 Control group:

[0367] The traditional bidirectional Bi-GRU recurrent neural network is used. Because it does not support unequal lengths of input and output sequences, it only performs the prediction tasks of the error threshold T1 and saturation value T2. The medical universal telescopic drill is adjusted in combination with a discrete PID controller.

[0368] (II) Experimental content:

[0369] 2.1 Tooth model preparation:

[0370] A common iron block with a length of 2 cm, a width of 2 cm, and a height of 3 cm was used to simulate human teeth (tooth model). A zinc alloy layer with a Mohs hardness of 5.5 and a thickness of 2 mm was set on the outer surface to simulate tooth enamel.

[0371] 2.2 Sensor placement:

[0372] A TP-2825 micro vibration sensor was placed inside the tooth model to record the amount of vibration.

[0373] 2.3 Opening operation:

[0374] Driven by the electric dental drill robot 7, both the experimental group and the control group performed pulp opening operations on the tooth model with the same stroke and the same universal angle adjustment. The vibration amounts of the experimental group and the control group during the pulp opening operation were recorded and compared. Because both the experimental group and the control group were controlled based on a discrete PID controller, when the output of the discrete PID controller could not be further increased or decreased, or when the output reached saturation and could not be further adjusted, it would cause unnecessary movements of the pulp opening of the medical universal telescopic drill, and thus generate more vibration. Based on the vibration amount, the control effects of the experimental group and the control group were compared.

[0375] (III) Experimental methods and conclusions:

[0376] 3.1 Model training:

[0377] The same data set D is used to train the models of the experimental group and the control group. The MAPE values ​​(Mean Absolute Percentage Error) of the experimental group and the control group during the training process are recorded and compared. Fig.12 As shown, it can be seen that among more than 4,000 training samples, the MAPE value of the experimental group during training is significantly smaller than that of the control group because it supports the case where the input and output sequences are of unequal lengths, indicating the advantages of the Seq2Seq model combined with the bidirectional Bi-GRU recurrent neural network in prediction and regulation.

[0378] 3.2 Opening operation and vibration recording:

[0379] After the training, the experimental group and the control group were subjected to pulp opening. The TP-2825 micro vibration sensor was used to record the vibration of the experimental group and the control group during the pulp opening process. Fig.13 As shown, it shows that the experimental group has higher accuracy and stability in controlling the medical universal telescopic drill.

[0380] All the above embodiments only express the implementation methods of the relevant practical applications of the present invention, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the invention patent. It should be pointed out that, for ordinary technicians in this field, several variations and improvements can be made without departing from the concept of the present invention, which all belong to the protection scope of the present invention. Therefore, the protection scope of the patent of the present invention shall be based on the attached claims.

[0381] For those skilled in the art, it can be further appreciated that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of the two. In order to clearly illustrate the interchangeability of hardware and software, the composition and steps of each example have been generally described in the above description according to function. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of the present invention.

[0382] At the same time, those skilled in the art can understand that all or part of the processes in all the above-mentioned embodiments can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media provided in this application and used in the embodiments may include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. As an illustration and not limitation, RAM is available in many forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (SSRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM).

Claims

1. A medical universal telescopic drill, comprising a drill assembly (1) for drilling a tooth pulp, characterized in that: It also includes a body component (2) and, a telescopic component (3) in the main body component (2) that is telescopic and universally adjustable, the telescopic component (3) comprising a plurality of small free bodies (302) having dipole moments, the telescopic and universally adjustable components being able to be performed separately or simultaneously; the drill bit component (1) being mounted on the telescopic component (3); A large free body (4) with a dipole moment fixed in the main body component (2) and used to generate a magnetic repulsive force with the small free body (302) so as to enable the telescopic component (3) to perform the telescopic and universal adjustment; All the small free bodies (302) with dipole moments generate the same or different magnetic moments based on the magnitude of the current, thereby generating a resultant force of the non-polar magnetic repulsion effect to form the magnetic repulsion; the telescopic component (3) includes a spherical shell (301); The plurality of small free bodies (302) are arranged in an annular array in the spherical shell (301) with the central axis of the large free body (4) as a reference; The front end of the spherical shell (301) is equipped with the drill assembly (1); the telescopic assembly (3) further comprises a plurality of electrical connectors (303), each of the electrical connectors (303) being electrically connected to one of the small free bodies (302); When all the electrical connectors (303) supply power to all the small free bodies (302), all the small free bodies (302) form the magnetic repulsion force between the large free body (4); the resultant force is a uniform resultant force or a non-uniform resultant force generated based on the same or different magnitudes of the currents passed through each of the small free bodies (302); When the uniform resultant force is generated, the spherical shell (301) and the drill bit assembly (1) perform the telescoping; When the non-uniform resultant force is generated, the spherical housing (301) and the drill bit assembly (1) perform the telescopic movement and / or the universal adjustment.

2. The universal telescopic drill bit according to claim 1, characterized in that: The body component (2) comprises a housing (201), an annular array of arc-shaped sliding grooves (202) are provided in the housing (201), and the spherical shell (301) is rollingly fitted in the arc-shaped sliding grooves (202).

3. The universal telescopic drill bit according to claim 2, characterized in that: The housing (201) is also provided with a detection component (5) for detecting angle information of the drill head component (1).

4. A method for controlling a universal telescopic drill according to any one of claims 1 to 3, comprising locating and obtaining position information of a tooth pulp opening by an optical positioning system, characterized in that: The control method comprises executing the following steps: S1, the discretized PID controller receives the target value tv as the desired magnetic field strength and direction, and the actual feedback value fv as the current magnetic field strength and direction as input, and calculates the preliminary control electrical signal e(t); S2, the bidirectional Bi-GRU recurrent neural network receives the historical error data D and the error E as input, dynamically calculates and updates the error threshold T1 and the saturation value T2; determines whether to activate the integral action of the PID controller according to the error threshold T1, and adjusts the accumulated step value CS of the integral term according to the saturation value T2; generates the final control electrical signal e(t)' according to the preliminary control electrical signal e(t) and the adjustment of the bidirectional Bi-GRU recurrent neural network; S3, sending the final control electrical signal e(t)' to the controller, controlling the periodic change of the magnetic field of all the small free bodies (302) and thus controlling the direction of the magnetic moment, thereby achieving the telescopic and universal angle adjustment of the drill assembly (1).

5. The control method according to claim 4, characterized in that: In S1, the discretized PID control algorithm is: Among them, E k represents the error between the target value tv and the feedback value fv at the kth sampling moment; K p is the proportionality coefficient, K i is the integration coefficient, K d is the differential coefficient; represents the cumulative sum of errors from time k to the current time t; E prev Represents the error of the previous time step; the method for calculating the error E between the target value tv and the feedback value fv is E=tv-fv.

6. The control method according to claim 5, characterized in that: In S2, the implementation method of the bidirectional Bi-GRU recurrent neural network includes: S200, executing the encoder: receiving an input sequence S = {D, E}, extracting the dependency relationship in the input sequence S from both the forward and backward directions through a bidirectional Bi-GRU network, and generating a hidden state sequence H; the updating methods of the forward and backward bidirectional Bi-GRU networks are respectively: in, and are the forward and backward hidden state vectors respectively; and They are the forward and backward gated recurrent units respectively; S201, ECA attention mechanism processing: weighted summing of the hidden state sequence H output by the encoder to generate an intermediate vector C: Among them, C i The intermediate vector C of the decoder at the i-th time step; h j is the jth encoder hidden state in the hidden state sequence H, α ij is the attention weight; S202, decoder processing: combining the intermediate vector C and the hidden state of the previous time step and hidden state And the prediction result y of the decoder at the previous time step t-1 , the prediction result y is generated through the bidirectional Bi-GRU network t =[T1,T2]: Where W and b are the weight matrix and bias vector; Indicates the final hidden state of the forward GRU and the final hidden state of the backward GRU Splice it together; and They are the forward and backward gated recurrent units respectively; In S2, the method for adjusting and generating the final control electrical signal e(t)' is: based on the error threshold T1 and the saturation value T2, a constraint term is introduced into the algorithm of the PID controller, and then the final control electrical signal e(t)' is updated.

7. Use of the universal telescopic drill bit according to any one of claims 1 to 4 in dental surgical instruments.

Images