A smart upper limb motor rehabilitation device and method

Abstract

This invention discloses an intelligent upper limb motor rehabilitation device and its method, belonging to the field of rehabilitation device design. The device includes a motor training module and a functional electrical stimulation module. The motor training module includes a permanent magnet synchronous motor and its drive and transmission systems. The drive system can perform torque control, speed control, and position control on the permanent magnet synchronous motor. The transmission system is a worm gear structure, achieving the effects of deceleration and torque increase. This upper limb motor rehabilitation device has multiple training modes, which can exercise upper limb muscle strength, improve joint range of motion, and evaluate the torque applied by the user, assess muscle strength, and detect spasticity through a method based on a fully connected neural network. Furthermore, combined with the functional electrical stimulation module, it can electrically stimulate and activate relevant muscles based on the user's force application range. Compared to large rehabilitation training devices in hospitals, this device has a compact structure, rich functions, and is more suitable for everyday home use, reducing the workload of doctors and improving rehabilitation efficiency.

Description

Technical Field

[0001] This invention belongs to the field of rehabilitation device design, specifically relating to an intelligent upper limb motor rehabilitation device and its method. Background Technology

[0002] Upper motor neuron damage caused by diseases such as stroke can lead to hemiplegia and even spasticity, severely impacting daily life. Rehabilitation medicine has proven that repetitive upper limb movements can improve the degree of hemiplegia, joint range of motion, and muscle activity. Traditional manual rehabilitation training relies primarily on the doctor's clinical experience to assess the patient's level of impairment and then perform repetitive joint exercises accordingly. The disadvantages are low efficiency and high cost.

[0003] To reduce the burden on doctors and improve rehabilitation efficiency, rehabilitation assistive devices are used in rehabilitation training. Currently, the mainstream devices are divided into end-effector traction rehabilitation devices and exoskeleton rehabilitation devices. The former has a simpler structure and is easier to control, but it cannot effectively stabilize the patient's joints, resulting in less targeted training; the latter can completely wrap around the patient's arm, allowing for independent training of each joint, but its overall structure is complex and generally larger. In general, both types of assistive devices suffer from high levels of specialization, high cost, and inconvenience, typically limiting their use to rehabilitation hospitals and failing to meet the needs of my country's large and complex rehabilitation population. Designing portable, miniaturized, and intelligent upper limb motor rehabilitation devices is of great significance for addressing the needs of home-based rehabilitation training. Summary of the Invention

[0004] The purpose of this invention is to overcome the deficiencies in the prior art and provide an intelligent upper limb motor rehabilitation device and method. This invention includes a motor training module and a functional electrical stimulation module. The motor training module includes a permanent magnet synchronous motor and its drive and transmission systems. The drive system can perform torque control, speed control, and position control on the permanent magnet synchronous motor. The transmission system is a worm gear structure, achieving the effects of deceleration and torque increase. This upper limb motor rehabilitation device has multiple training modes, which can exercise upper limb muscle strength, improve joint range of motion, and evaluate the torque applied by the user, assess muscle strength, and detect spasticity through a method based on a fully connected neural network. Furthermore, combined with the functional electrical stimulation module, it can electrically stimulate and activate relevant muscles based on the user's force application range.

[0005] The specific technical solution adopted in this invention is as follows:

[0006] In a first aspect, the present invention provides an intelligent upper limb motor rehabilitation device, comprising a motor training module and a functional electrical stimulation module;

[0007] The exercise training module is located in an internal cavity consisting of an instrument shell and an instrument base, and includes a permanent magnet synchronous motor, a motor transmission module, and a motor drive module; the instrument base integrates a power module, which provides operating voltage for the entire exercise training module and the functional electrical stimulation module.

[0008] The motor drive module includes a turbine and a worm gear. The output shaft of the permanent magnet synchronous motor is connected to one end of the worm gear, and the other end of the worm gear is connected to the turbine gear. The transmission output shaft of the turbine is located on its central axis, and the two ends are respectively connected to a first grip and a second grip located outside the internal cavity.

[0009] The motor drive module is connected to the permanent magnet synchronous motor via a drive board fixing device, and includes an ARM main control module, a motor gate driver module, a magnetic encoder module, a dissipation circuit module, and an ADC sampling circuit module. The ARM main control module processes the acquired data and sends control commands to the motor gate driver module. The motor gate driver module controls the permanent magnet synchronous motor by controlling the switching frequency of the MOSFET. The magnetic encoder module detects the rotation of the magnet fixed on the drive board fixing device to obtain the position of the permanent magnet synchronous motor and communicates with the ARM main control module via SPI. The dissipation circuit module is externally connected to a dissipation resistor to release the energy input to the device during passive drive in a timely manner. The ADC sampling module measures the three-phase current value of the permanent magnet synchronous motor through the sampling resistor.

[0010] The functional electrical stimulation module is connected to the power module via a power data cable and includes an upper arm fixation shell, a first upper arm muscle stimulation electrode, a second upper arm muscle stimulation motor, a deltoid muscle stimulation electrode, and a supraspinatus muscle stimulation electrode. The upper arm fixation shell is made of flexible material and can be fixedly connected to the user's upper arm. The inner wall of the upper arm fixation shell is provided with the first upper arm muscle stimulation electrode and the second upper arm muscle stimulation motor, and the deltoid muscle stimulation electrode and the supraspinatus muscle stimulation electrode are connected to the outside via wires.

[0011] Preferably, the output end housing of the permanent magnet synchronous motor is connected to one end of the motor drive mounting base, and the other end of the motor drive mounting base is connected to a first drive mounting side plate, a second drive mounting side plate, and a drive mounting top plate; the first drive mounting side plate, the second drive mounting side plate, and the drive mounting top plate are respectively connected to the instrument housing and together form the upper cover structure of the worm; the bottom of the drive mounting top plate is fixed with a first bearing seat and a second bearing seat at intervals along the worm axial direction, and the worm is respectively mounted on the first bearing seat and the second bearing seat through worm bearings; the turbine is provided with a first turbine mounting side plate and a second turbine mounting side plate on both sides, and the turbine is fixed to the second turbine coupling through a first turbine coupling and precisely meshes with the worm.

[0012] Preferably, one side of the upper arm fixing shell is provided with a first hook and loop fastener with barbed ends and a second hook and loop fastener with barbed ends, and the other side is provided with a first hook and loop fastener with rounded ends and a second hook and loop fastener with rounded ends; the upper arm fixing shell can be fixedly connected to the user's upper arm through the first hook and loop fastener with barbed ends, the first hook and loop fastener with rounded ends, the second hook and loop fastener with barbed ends and the second hook and loop fastener with rounded ends.

[0013] Preferably, the ARM main control module, motor gate driver module, magnetic encoder module, dissipation circuit module and ADC sampling circuit module are integrated on the same circuit board. The circuit board is kept centered with the permanent magnet synchronous motor by the drive board fixing device, so that the magnetic encoder module and the permanent magnet synchronous motor can ensure the optimal measurement posture.

[0014] Secondly, the present invention provides a control method for a permanent magnet synchronous motor of an intelligent upper limb motor rehabilitation device as described in any of the first aspects, as follows:

[0015] Ignoring copper losses and changes in inductor energy storage, the permanent magnet synchronous motor can be modeled as follows:

[0016] The electromagnetic power is as follows:

[0017]

[0018] Relationship between electrical angle and mechanical angle:

[0019] ω e =n p ω m

[0020] The electromagnetic torque is calculated as follows:

[0021]

[0022] Rotor motion equations:

[0023]

[0024]

[0025] Where P e Represents electromagnetic power, ω e Represents electric angular velocity, ω m Represents mechanical angular velocity. Represents mechanical angular acceleration, Ψ f I represents the flux linkage of the rotor permanent magnet. q I represents the quadrature-axis current. d L represents the direct-axis current. q L represents quadrature axis inductance. d Indicates direct-axis inductance, n p T represents the number of pole pairs in a permanent magnet synchronous motor.e T represents electromagnetic torque. l J represents the load torque, B represents the rotor inertia of the permanent magnet synchronous motor, and G represents the velocity viscosity coefficient. motor (s) represents the transfer function of the permanent magnet synchronous motor, and s represents the complex frequency of the transfer function;

[0026] Based on the modeling of the permanent magnet synchronous motor, three basic control methods—torque control, speed control, and position control—were implemented. Based on these three basic control methods, two training modes—active and passive—were established. In active mode, the user actively exerts force to rotate the rehabilitation device, and the device can adjust the resistance of rotation to adapt to different training needs. In passive mode, the rehabilitation device rotates the user's upper limbs at a certain speed, and the speed can be smoothly adjusted.

[0027] Thirdly, the present invention provides a control method for the dissipation circuit module of the intelligent upper limb motor rehabilitation device as described in any of the first aspects, as follows:

[0028] The input energy is converted into heat energy and dissipated by the dissipation resistor. The duty cycle of the dissipation circuit PWM is controlled according to the current bus current and power supply voltage to control the duration of current passing through the dissipation resistor.

[0029] Based on the relationship between the bus power and the quadrature-axis and direct-axis power of a permanent magnet synchronous motor:

[0030]

[0031] Among them, U q For quadrature axis voltage, I q For quadrature-axis current, U d For direct-axis voltage, I d For direct-axis current, U bus For the bus voltage, I bus Let be the bus current; after rearranging, we can obtain the formula for calculating the bus current:

[0032]

[0033] If the bus current is reverse-input to the power supply, it is called regenerative current. To protect the power supply, regenerative current is considered harmful, meaning the maximum regenerative current is zero. The power supply voltage can be directly obtained through ADC sampling, and the duty cycle setting can be calculated as follows:

[0034]

[0035] if duty > 1 : duty = 1

[0036] else if duty < 0: duty = 0

[0037] Where "duty" represents the duty cycle, which is affected by both bus current and power supply voltage. Whether there is inverted input current or excessively high power supply voltage, the duty cycle will increase; R brake U is the resistance value of the dissipation resistor. end U start To enable the relevant voltage parameters of the dissipation circuit, i.e., the power supply voltage reaches U start The dissipation circuit is activated at time U to achieve end When the dissipation circuit is fully activated, the PWM duty cycle is set to 1.

[0038] Fourthly, the present invention provides a torque self-sensing method based on a fully connected neural network for an intelligent upper limb motor rehabilitation device as described in any of the first aspects, as follows:

[0039] A five-layer fully connected neural network is constructed, consisting of one input layer, three hidden layers, and one output layer. The input layer includes five values: current velocity, previous velocity, current position error, previous position error, and current quadrature-axis current. The position error represents the difference between the actual position and the target position in position control mode. The three hidden layers are identical, each containing 12 neurons. The activation function for each neuron is the sigmoid activation function, as shown below:

[0040]

[0041] The output layer contains a single value representing the perceived torque, and no longer contains an activation function;

[0042] During the use of this rehabilitation device, real-time position error, speed, and quadrature-axis current values ​​of the internal permanent magnet synchronous motor, as well as torque values ​​recorded by the external torque sensor, were collected. The data were then processed, divided into training, validation, and test sets, and normalized. The training set was used to train a fully connected neural network with the above structure, using the following cross-entropy as the loss function:

[0043]

[0044] Where J represents the optimization objective of the loss function, M represents the total number of samples in the training set, n represents the current sample index, and v m y represents the output value of the m-th sample through the neural network. m Let J represent the actual external torque value of the m-th sample; since J is a non-convex function, the optimization method chosen is the conjugate gradient method or the quasi-Newton method.

[0045] Based on the aforementioned external torque estimation method, the occurrence of spasticity in the user can be monitored during training. If a user experiences spasticity during rehabilitation training, both the acceleration detected by the permanent magnet synchronous motor and the estimated external torque will generate a large peak value, then rapidly decrease to a slightly lower value and remain there for a period of time before decreasing to 0. This corresponds to the sudden increase in muscle tone when spasticity occurs, which hinders the original stretching activity. The muscle tone is then maintained for a period of time, and finally disappears after the muscles relax.

[0046] Acceleration and torque acceleration are used as the criteria for judging whether spasm has occurred. Thresholds are set for acceleration and torque acceleration. Once the actual values ​​of both exceed the threshold, spasm is considered to have occurred, and the training process is stopped immediately.

[0047] Fifthly, the present invention provides a method for using the functional electrical stimulation module based on the force application zone of the intelligent upper limb motor rehabilitation device as described in any of the first aspects, as follows:

[0048] When the user continuously rotates this rehabilitation device, the force exertion of each upper limb exhibits certain periodic characteristics. The rotation of the training device can be divided into a left-hand force exertion zone and a right-hand force exertion zone. The left-hand force exertion zone is the angle range in which the device is rotated primarily by the left hand, and the right-hand force exertion zone is the angle range in which the device is rotated primarily by the right hand. It is assumed that the force exertion zones of the user's left and right hands should be equal in size, i.e., both 180°. For hemiplegic users, the healthy hand compensates for the force exertion zone of the affected hand. First, the force exertion zone of the healthy hand is assessed, and its complementary zone is the force exertion zone of the affected hand. Let the force exertion zone of the healthy hand be (start, start+180°), where start represents the starting angle of the healthy hand force exertion zone. A certain rotational resistance is set, allowing the user to actively, continuously, and uniformly rotate the prototype using only the healthy hand. By enumerating start values ​​within the range of 0–360°, the average power of the healthy hand force exertion zone at each value is calculated. The start value that maximizes the average power is the force exertion zone assessment result.

[0049] The functional electrical stimulation module is fixed to the user's upper arm via an upper arm fixation shell. The electrical stimulation sites are the supraspinatus, deltoid, biceps brachii, and triceps brachii muscles on the hemiplegic side. Electrical stimulation is applied to the supraspinatus, deltoid, and triceps brachii muscles in the affected side's force application zone, while stimulation is applied to the biceps brachii muscle in the unaffected side's force application zone. This is done alternately, with the stimulation frequency set to a low frequency of 50Hz. The stimulation intensity is adjusted in real time based on the muscle strength sensed by the rehabilitation device, calculated using the following formula:

[0050]

[0051] Where S is the current stimulation intensity, in mA; S maxThe current is the maximum stimulation current value, which should be adjusted according to the patient's comfort level and should not exceed 100mA; T user The external torque value is estimated using a fully connected neural network; T max The peak torque applied to the current user.

[0052] Compared with the prior art, the present invention has the following advantages:

[0053] This invention presents a compact, feature-rich, and intelligent upper limb rehabilitation device. This device offers multiple training modes to strengthen upper limb muscles, improve joint range of motion, and assess the torque applied by the user, muscle strength, and spasticity using a fully connected neural network-based method. Furthermore, it incorporates a functional electrical stimulation module to activate relevant muscles based on the user's exertion zone. Compared to large-scale rehabilitation devices used in hospitals, this device is more suitable for everyday home use, reducing the workload of doctors and improving rehabilitation efficiency. Attached Figure Description

[0054] Figure 1 This is a diagram of the internal structure of the upper limb motor rehabilitation device provided in this embodiment;

[0055] Figure 2 This is a structural diagram of the worm gear transmission part of the upper limb motor rehabilitation device provided in this embodiment;

[0056] Figure 3 This is a schematic diagram of the upper limb motor rehabilitation device and functional electrical stimulation module provided in this embodiment;

[0057] Figure 4 This is a schematic diagram of the printed circuit board of the driver for the upper limb motor rehabilitation device provided in this embodiment;

[0058] Figure 5 This is a schematic diagram of the driver circuit for the upper limb motor rehabilitation device provided in this embodiment;

[0059] Figure 6 This is a schematic diagram of the workflow of the upper limb motor rehabilitation device provided in this embodiment;

[0060] Figure 7 This is a schematic diagram of the dissipation circuit module of the upper limb motor rehabilitation device provided in this embodiment;

[0061] Figure 8 This is a schematic diagram of the torque estimation neural network for the upper limb motor rehabilitation device provided in this embodiment;

[0062] Figure 9 This is a schematic diagram illustrating the force application range determination of the upper limb motor rehabilitation device provided in this embodiment;

[0063] The attached figures are labeled as follows: 1. Permanent magnet synchronous motor; 2. Drive plate fixing device; 3. Motor drive module; 4. Dissipative resistor; 5. Instrument housing; 6. Instrument base; 7. Motor transmission fixing seat; 8. First transmission fixing side plate; 9. First turbine fixing side plate; 10. First bearing seat; 11. Transmission output shaft; 12. First grip; 13. Transmission fixing top plate; 14. Worm; 15. Turbine; 16. First turbine coupling; 17. Second turbine coupling; 18. Second grip; 19. Second bearing seat; 20. Worm bearing; 21. Second transmission fixing side plate; 22. Power data cable; 23. First upper arm muscle stimulation electrode; 24. Second upper arm muscle stimulation electrode; 25. First hook and loop fastener round bristle end; 26. Second hook and loop fastener round bristle end; 27. First hook and loop fastener barb end; 28. Second hook and loop fastener barb end; 29. ​​Upper arm fixing housing; 30. Deltoid muscle stimulation electrode; 31. Supraspinatus muscle stimulation electrode. Detailed Implementation

[0064] The present invention will be further described and illustrated below with reference to the accompanying drawings and specific embodiments. The technical features of each embodiment of the present invention can be combined accordingly, provided that there is no mutual conflict.

[0065] like Figures 1-3 As shown, this invention provides an intelligent upper limb motor rehabilitation device, which mainly includes a motor training module and a functional electrical stimulation module.

[0066] In this invention, the device housing 5 and the device base 6 together form a relatively enclosed shell structure, with a power module integrated inside the device base 6. The power module provides operating voltage for the entire exercise training module and the functional electrical stimulation module. The exercise training module is entirely located within the internal cavity of this housing, and mainly includes a permanent magnet synchronous motor 1, a motor transmission module, and a motor drive module 3.

[0067] In this invention, the motor drive module includes a turbine 15, a worm gear 14, and surrounding connecting and fixing devices. The output shaft of the permanent magnet synchronous motor 1 is drive-connected to one end of the worm gear 14, and the other end of the worm gear 14 is meshed with the turbine 15. The drive output shaft 11 of the turbine 15 is located on its central axis, and its two ends are respectively connected to a first grip 12 and a second grip 19 located outside the internal cavity.

[0068] In this embodiment, the connection and fixing device is specifically as follows: the output end housing of the permanent magnet synchronous motor 1 is connected to one end of the motor transmission fixing base 7, and the other end of the motor transmission fixing base 7 is connected to a first transmission fixing side plate 8, a second transmission fixing side plate 22, and a transmission fixing top plate 13. The first transmission fixing side plate 8, the second transmission fixing side plate 22, and the transmission fixing top plate 13 are respectively connected to the instrument housing 5. The first transmission fixing side plate 8 and the second transmission fixing side plate 22 are fixed at intervals to the bottom of the transmission fixing top plate 13, and the three together constitute the upper cover structure of the worm gear 14.

[0069] In this invention, a first bearing seat 10 and a second bearing seat 20 are fixed at intervals along the axial direction of the worm 14 at the bottom of the transmission fixed top plate 13. The worm 14 is mounted on the first bearing seat 10 and the second bearing seat 20 respectively through worm bearings 21. A first turbine fixed side plate 9 and a second turbine fixed side plate 17 are respectively provided on both sides of the turbine 15. The turbine 15 is fixed to the second turbine coupling 18 through a first turbine coupling 16 and a second turbine coupling 18 and precisely meshes with the worm 14.

[0070] In this invention, such as Figure 4 As shown, the motor drive module 3 is connected to the permanent magnet synchronous motor 1 via the drive board fixing device 2, and includes an ARM main control module, a motor gate driver module, a magnetic encoder module, a dissipation circuit module, and an ADC sampling circuit module. In this invention, the ARM main control module processes the acquired data and sends control commands to the motor gate driver module. The motor gate driver module controls the permanent magnet synchronous motor 1 by controlling the switching frequency of the MOSFET. The magnetic encoder module detects the rotation of the magnet fixed on the drive board fixing device 2 to obtain the position of the permanent magnet synchronous motor 1 and communicates with the ARM main control module via SPI. The dissipation circuit module is externally connected to a dissipation resistor 4 to promptly release the energy input to the device during passive drive. The ADC sampling module measures the three-phase current value of the permanent magnet synchronous motor 1 through the sampling resistor.

[0071] In this embodiment, the ARM main control module, motor gate driver module, magnetic encoder module, dissipation circuit module and ADC sampling circuit module are preferably integrated on the same circuit board. The circuit board is kept centered with the permanent magnet synchronous motor 1 by the drive board fixing device 2, so that the magnetic encoder module and the permanent magnet synchronous motor 1 can ensure the optimal measurement posture.

[0072] In this invention, the functional electrical stimulation module is connected to the power module via a power data cable 23, and includes an upper arm fixation shell 30, a first upper arm muscle stimulation electrode 24, a second upper arm muscle stimulation motor 25, a deltoid muscle stimulation electrode 31, and a supraspinatus muscle stimulation electrode 32. The upper arm fixation shell 30 is made of flexible material and can be fixedly connected to the user's upper arm. The inner wall of the upper arm fixation shell 30 is provided with the first upper arm muscle stimulation electrode 24 and the second upper arm muscle stimulation motor 25, and the deltoid muscle stimulation electrode 31 and the supraspinatus muscle stimulation electrode 32 are connected externally via wires.

[0073] In this embodiment, the upper arm fixing housing 30 has a first hook and loop fastener barbed end 28 and a second hook and loop fastener barbed end 29 on one side, and a first hook and loop fastener rounded end 26 and a second hook and loop fastener rounded end 27 on the other side. The upper arm fixing housing 30 can be fixedly connected to the user's upper arm through the first hook and loop fastener barbed end 28, the first hook and loop fastener rounded end 26, the second hook and loop fastener barbed end 29, and the second hook and loop fastener rounded end 27.

[0074] In practical applications, if the changes in copper losses and inductor energy storage are ignored, the permanent magnet synchronous motor 1 can be modeled as follows:

[0075] The electromagnetic power is as follows:

[0076]

[0077] Relationship between electrical angle and mechanical angle:

[0078] ω e =n p ω m

[0079] The electromagnetic torque is calculated as follows:

[0080]

[0081] Rotor motion equations:

[0082]

[0083]

[0084] Where P e Represents electromagnetic power, ω e Represents electric angular velocity, ω m Represents mechanical angular velocity. Represents mechanical angular acceleration, Ψ f I represents the flux linkage of the rotor permanent magnet. q I represents the quadrature-axis current. d L represents the direct-axis current. q L represents quadrature axis inductance. d Indicates direct-axis inductance, n pT represents the number of pole pairs in a permanent magnet synchronous motor. e T represents electromagnetic torque. l J represents the load torque, B represents the rotor inertia of the permanent magnet synchronous motor I, and G represents the velocity viscosity coefficient. motor (s) represents the transfer function of permanent magnet synchronous motor 1, and s represents the complex frequency of the transfer function;

[0085] Specifically, such as Figure 5 As shown, the control algorithm for the permanent magnet synchronous motor 1 in this embodiment is a vector control algorithm, and the control input includes the target quadrature-axis current. With the target direct-axis current The control strategy adopted is to make Setting it to 0 means making the stator rotating magnetic field space vector orthogonal to the rotor magnetic field space vector; only control is needed. Size. This can be achieved through three control methods: torque control, speed control, and position control. The value of .

[0086] It has been specified and The value still needs the actual I. q with I d The value serves as the feedback quantity for the current loop. Let I be the three-phase current value of the permanent magnet synchronous motor. a I b I c The current values ​​I of the two phases are obtained by sampling the two currents of the gate driver. a I b Then, the current value of the third phase is obtained using Kirchhoff's laws:

[0087] I c =-I a -I b

[0088] The three-phase current values ​​are simplified into two mutually orthogonal current values ​​I after Clark's equal power conversion. α ,I β :

[0089] I α =I a

[0090]

[0091] Based on the current electrical angle θ of the motor, the current value, which is sinusoidally transformed in a fixed coordinate system, is converted into a linearly transformed current value in a rotating coordinate system through Park transformation. A proportional-integral-derivative controller is then used for control.

[0092] I q =I α cosθ+Iβ sinθ

[0093] I d =I α sinθ+I β cosθ

[0094] Actual I q with I d The error value is obtained by subtracting the target value from the current value. The control voltage U is obtained through a proportional-integral controller in the current loop. q with U d The control voltage U in the stationary coordinate system is obtained by inverse Park transformation. α with U β :

[0095] U α =U d cosθ-U q sinθ

[0096] U β =U a sinθ+U q cosθ

[0097] According to U α with U β The duty cycle of the three-phase PWM of the gate driver is obtained by analyzing the sector of space vector pulse width modulation, which in turn controls the energization of each phase of the motor to make the motor rotate.

[0098] Based on vector control, the motor is subjected to three basic control modes: torque control, speed control, and position control. In torque control mode, a target torque value needs to be input, based on the fundamental formula for motor electromagnetic torque:

[0099]

[0100] Motor output torque and quadrature axis current I q Proportional to the actual speed, a proportional controller is used. In speed control mode, the speed error is obtained by subtracting the target speed value from the measured speed value, and then the control input (I) is obtained through a proportional-integral controller. q To eliminate static error, an integral term is added. In position control mode, the target position value is input, and a position closed loop is added on top of speed control. This closed loop uses a proportional controller to control the position.

[0101] like Figure 6As shown, the training modes in this embodiment include active and passive modes. In active mode, the user actively exerts force to rotate the rehabilitation device, and the device can adjust the resistance of rotation to adapt to different training needs. Torque control is used in this mode, with the permanent magnet synchronous motor 1 outputting torque opposite to the movement. In passive mode, the rehabilitation device rotates the user's upper limb at a certain speed, which can be smoothly adjusted using speed control. During training, functional electrical stimulation can be provided to assist rehabilitation therapy based on the force application range determination method, and the training duration can be set. Data such as training duration, number of rotations, average speed, muscle strength, and symmetry in both active and passive modes are recorded and displayed after training is completed.

[0102] like Figure 7 As shown, this embodiment provides a control method for the dissipation circuit module of the aforementioned intelligent upper limb motor rehabilitation device. The target application scenario for the motor is rehabilitation training and assessment, thus involving a large number of passive driving situations. Furthermore, patients with upper limb motor disorders may experience spasticity, generating a reaction force on the motor. When using a regulated power supply, if this energy input to the motor is not dissipated, it will cause the power supply voltage to rise, thereby damaging the power supply. Therefore, it is necessary to convert or dissipate this portion of energy input to the motor to prevent the power supply voltage from rising.

[0103] This embodiment dissipates the input energy as heat through a dissipation resistor. The duty cycle of the PWM circuit in the dissipation circuit is controlled based on the current bus current and power supply voltage to regulate the duration of current flow through the dissipation resistor. The relationship between the bus power and the quadrature-axis and direct-axis power of the permanent magnet synchronous motor 1 is based on the following formula:

[0104]

[0105] U q For quadrature axis voltage, I q For quadrature-axis current, U d For direct-axis voltage, I d For direct-axis current, U bus For the bus voltage, I bus Let be the bus current. Rearranging the terms, we can obtain the formula for calculating the bus current:

[0106]

[0107] If the bus current is reverse-input to the power supply, it is called regenerative current. To protect the power supply, regenerative current is considered harmful, meaning the maximum regenerative current is zero. The power supply voltage can be directly obtained through ADC sampling, and the duty cycle setting can be calculated as follows:

[0108]

[0109] if duty > 1 : duty = 1

[0110] else if duty < 0: duty = 0

[0111] Where "duty" refers to the duty cycle, which is affected by both bus current and power supply voltage. Whether there is reverse-inverting input current or excessively high power supply voltage, the duty cycle will increase. R brake U is the resistance value of the dissipation resistor. end U start To enable the relevant voltage parameters of the dissipation circuit, i.e., the power supply voltage reaches U start The dissipation circuit is activated at time U to achieve end When the dissipation circuit is fully activated, the PWM duty cycle is set to 1.

[0112] like Figure 8 As shown, this embodiment provides a torque self-sensing method based on a fully connected neural network for the rehabilitation machinery of the present invention. In order to assess the muscle strength and muscle power of the user of this intelligent upper limb motor rehabilitation machinery, and to detect and relieve spasticity in a timely manner, it is necessary to sense the torque applied by the user to the device in real time. The torque applied by the user to the rehabilitation machinery is the load of the motor. Based on the actual parameters of the permanent magnet synchronous motor 1, the real-time torque applied by the user can be derived as follows:

[0113]

[0114] Where T user K represents the torque applied by the user. T ω represents the torque constant of the permanent magnet synchronous motor. m I represents mechanical angular velocity. q Let f represent the quadrature-axis current, J represent the dynamic friction force, J represent the rotor inertia of permanent magnet synchronous motor 1, and B represent the velocity viscosity coefficient. Accurate torque estimation using this modeling method requires precise identification of relevant parameters of permanent magnet synchronous motor 1, and modeling the friction term f is extremely difficult. This invention proposes a torque self-sensing method for permanent magnet synchronous motors based on a fully connected neural network, transforming this problem from model-driven to data-driven. It can calculate the external torque of rehabilitation equipment in real time by collecting relevant data from the permanent magnet synchronous motor.

[0115] A five-layer fully connected neural network is constructed, consisting of one input layer, three hidden layers, and one output layer. The input layer includes five values: current velocity, previous velocity, current position error, previous position error, and current quadrature-axis current. The position error represents the difference between the actual position and the target position in position control mode. The three hidden layers are identical, each containing 12 neurons. Each neuron uses the sigmoid activation function, as shown below:

[0116]

[0117] The output layer contains a single value representing the perceived torque and no longer contains an activation function.

[0118] During the use of this rehabilitation device, real-time position error, speed, and quadrature-axis current values ​​of the internal permanent magnet synchronous motor 1, as well as torque values ​​recorded by the external torque sensor, were collected. The data were processed, divided into training, validation, and test sets, and normalized. The fully connected neural network with the above structure was trained using the training set, employing the following cross-entropy as the loss function:

[0119]

[0120] Where J represents the optimization objective of the loss function, M represents the total number of samples in the training set, m represents the current sample index, and v m y represents the output value of the m-th sample through the neural network. m Let J represent the actual external torque value of the m-th sample. Since J is a non-convex function, the conjugate gradient method or the quasi-Newton method is chosen for optimization.

[0121] Based on the aforementioned external torque estimation method, the occurrence of spasticity in the user can be monitored during training. If a spasticity occurs during rehabilitation training, both the acceleration detected by the permanent magnet synchronous motor 1 and the estimated external torque will produce a large peak value, then rapidly decrease to a slightly lower value and remain there for a period of time before decreasing to 0. This corresponds to the sudden increase in muscle tone when a spasticity occurs, which hinders the original stretching activity, followed by a period of muscle tone maintenance, and finally the disappearance of muscle tone after muscle relaxation.

[0122] Acceleration and torque acceleration are used as the criteria for judging whether spasm has occurred. Thresholds are set for acceleration and torque acceleration. Once the actual values ​​of both exceed the threshold, spasm is considered to have occurred, and the training process is stopped immediately.

[0123] like Figure 9 As shown, this embodiment provides a method for using a functional electrical stimulation module based on force exertion zones. During training, the symmetry of force exertion in both upper limbs of the user is assessed and displayed in real time to provide a reference for adjusting the training method. Based on the symmetry information obtained from the assessment, the functional electrical stimulation module is controlled to stimulate the movement of the currently exerting muscle group, activate muscle motor ability, and improve neuromuscular function.

[0124] When a user continuously rotates this rehabilitation device, the force exertion of each upper limb exhibits a certain periodicity. The rotation of the training device can be divided into a left-hand force exertion zone and a right-hand force exertion zone. The left-hand force exertion zone is the angular range in which the device is primarily rotated by the left hand, and the same applies to the right-hand force exertion zone. It is assumed that the force exertion zones of the user's left and right hands should be equal in size, i.e., both 180°. For hemiplegic users, the healthy hand compensates for the force exertion zone of the affected hand. First, the force exertion zone of the healthy hand is assessed, and its complementary zone is the force exertion zone of the affected hand. Let the force exertion zone of the healthy hand be (start, start+180°), where start represents the starting angle of the healthy hand force exertion zone. A certain rotational resistance is set, allowing the user to actively, continuously, and uniformly rotate the prototype using only the healthy hand. By enumerating start values ​​within the range of 0–360°, the average power of the healthy hand force exertion zone for each value is calculated. The start value that maximizes the average power is the force exertion zone assessment result.

[0125] The functional electrical stimulation module is fixed to the user's upper arm via the upper arm fixing shell 30 and hook and loop fasteners. The electrical stimulation sites are the supraspinatus, deltoid, biceps brachii, and triceps brachii muscles on the hemiplegic side. Electrical stimulation is applied to the supraspinatus, deltoid, and triceps brachii muscles in the affected side's force application zone; stimulation is applied to the biceps brachii muscle in the unaffected side's force application zone, alternating between the two. The stimulation frequency is specified as a low frequency of 50Hz, and the stimulation intensity is adjusted in real-time based on the muscle strength sensed by the rehabilitation device. The calculation formula is as follows:

[0126]

[0127] Where S is the current stimulation intensity, in mA; S max The current is the maximum stimulation current value, which should be adjusted according to the patient's comfort level and should not exceed 100mA; T user The external torque value is estimated using a fully connected neural network; T max The peak torque applied to the current user.

[0128] This invention provides a rehabilitation device with multiple training modes, capable of strengthening upper limb muscles and improving joint range of motion. It also employs a fully connected neural network-based method to assess the torque applied by the user, evaluate muscle strength, and detect spasticity. Furthermore, a functional electrical stimulation module activates relevant muscles by electrical stimulation based on the user's exertion zone. Compared to large rehabilitation training devices used in hospitals, this device is compact, feature-rich, and more suitable for everyday home use, reducing the workload of doctors and improving rehabilitation efficiency.

[0129] The embodiments described above are merely preferred embodiments of the present invention and are not intended to limit the invention. Those skilled in the art can make various changes and modifications without departing from the spirit and scope of the invention. Therefore, all technical solutions obtained through equivalent substitution or transformation fall within the protection scope of the present invention.

Claims

1. An intelligent upper limb motor rehabilitation device, characterized in that, It includes a sports training module and a functional electrical stimulation module; The exercise training module is located in an internal cavity consisting of an instrument shell (5) and an instrument base (6), and includes a permanent magnet synchronous motor (1), a motor transmission module and a motor drive module (3); the instrument base (6) integrates a power supply module, which is used to provide working voltage for the entire exercise training module and the functional electrical stimulation module; The motor drive module includes a turbine (15) and a worm (14). The output shaft of the permanent magnet synchronous motor (1) is connected to one end of the worm (14), and the other end of the worm (14) is connected to the turbine (15). The transmission output shaft (11) of the turbine (15) is located on its central axis, and the two ends are respectively connected to a first grip (12) and a second grip (19) located outside the internal cavity. The motor drive module (3) is connected to the permanent magnet synchronous motor (1) through the drive board fixing device (2), and includes an ARM main control module, a motor gate driver module, a magnetic encoder module, a dissipation circuit module and an ADC sampling circuit module; the ARM main control module is used to process the acquired data and send control commands to the motor gate driver module; the motor gate driver module is used to control the permanent magnet synchronous motor (1) by controlling the switching frequency of the MOS transistor; the magnetic encoder module is used to detect the rotation of the magnet fixed on the drive board fixing device (2) to obtain the position of the permanent magnet synchronous motor (1), and communicates with the ARM main control module through SPI; the dissipation circuit module is connected to an external dissipation resistor (4) to release the energy input to the device during passive drive in a timely manner; the ADC sampling module is used to measure the three-phase current value of the permanent magnet synchronous motor (1) through the sampling resistor; The functional electrical stimulation module is connected to the power module via a power data cable (23) and includes an upper arm fixing shell (30), a first upper arm muscle stimulation electrode (24), a second upper arm muscle stimulation motor (25), a deltoid muscle stimulation electrode (31), and a supraspinatus muscle stimulation electrode (32). The upper arm fixing shell (30) is made of flexible material and can be fixedly connected to the user's upper arm. The inner wall of the upper arm fixing shell (30) is provided with the first upper arm muscle stimulation electrode (24) and the second upper arm muscle stimulation motor (25), and the deltoid muscle stimulation electrode (31) and the supraspinatus muscle stimulation electrode (32) are connected to the outside via wires.

2. The intelligent upper limb motor rehabilitation device according to claim 1, characterized in that, The output end housing of the permanent magnet synchronous motor (1) is connected to one end of the motor transmission fixing seat (7). The other end of the motor transmission fixing seat (7) is connected to the first transmission fixing side plate (8), the second transmission fixing side plate (22), and the transmission fixing top plate (13). The first transmission fixing side plate (8), the second transmission fixing side plate (22), and the transmission fixing top plate (13) are respectively connected to the instrument housing (5) and together form the upper cover structure of the worm (14). The bottom of the transmission fixing top plate (13) is fixed with the first bearing seat (10) and the second bearing seat (20) at intervals along the axial direction of the worm (14). The worm (14) is installed on the first bearing seat (10) and the second bearing seat (20) respectively through the worm bearing (21). The turbine (15) is provided with the first turbine fixing side plate (9) and the second turbine fixing side plate (17) on both sides respectively. The turbine (15) is fixed with the first turbine coupling (16) and the second turbine coupling (18) and precisely meshes with the worm (14).

3. The intelligent upper limb motor rehabilitation device according to claim 1, characterized in that, The upper arm fixing shell (30) has a first hook and loop fastener barbed end (28) and a second hook and loop fastener barbed end (29) on one side, and a first hook and loop fastener rounded end (26) and a second hook and loop fastener rounded end (27) on the other side; the upper arm fixing shell (30) can be fixedly connected to the user's upper arm through the first hook and loop fastener barbed end (28), the first hook and loop fastener rounded end (26), the second hook and loop fastener barbed end (29), and the second hook and loop fastener rounded end (27).

4. The intelligent upper limb motor rehabilitation device according to claim 1, characterized in that, The ARM main control module, motor gate driver module, magnetic encoder module, dissipation circuit module and ADC sampling circuit module are integrated on the same circuit board. The circuit board is kept aligned with the permanent magnet synchronous motor (1) by the drive board fixing device (2), so that the magnetic encoder module and the permanent magnet synchronous motor (1) can be in the best measurement position.

5. A smart upper limb motor rehabilitation device according to any one of claims 1 to 4, characterized in that, The control method of the permanent magnet synchronous motor (1) is as follows: Ignoring copper losses and changes in inductor energy storage, the permanent magnet synchronous motor (1) can be modeled as follows: The electromagnetic power is as follows: ； Relationship between electrical angle and mechanical angle: ； The electromagnetic torque is calculated as follows: ； Rotor motion equations: ； ； in Indicates electromagnetic power. Represents electric angular velocity. Represents mechanical angular velocity. Represents mechanical angular acceleration. Indicates the magnetic flux linkage of the rotor permanent magnet. Indicates quadrature-axis current. Represents direct-axis current. Indicates quadrature axis inductance. Indicates direct-axis inductance. This indicates the number of pole pairs in a permanent magnet synchronous motor (1). Indicates electromagnetic torque. Indicates load torque. This represents the rotor inertia of a permanent magnet synchronous motor (1). Represents the velocity viscosity coefficient. The transfer function of the permanent magnet synchronous motor (1) is represented. Represents the complex frequency of the transfer function; Based on the modeling of the permanent magnet synchronous motor (1), the three basic control methods of torque control, speed control and position control are completed. Based on the three basic control methods, two training modes, active mode and passive mode, are completed. In active mode, the user actively exerts force to drive the rehabilitation device to rotate. The rehabilitation device can adjust the resistance of rotation to adapt to different training needs. In passive mode, the rehabilitation device drives the user's upper limb to rotate at a certain speed. The speed can be smoothly adjusted.

6. A smart upper limb motor rehabilitation device according to any one of claims 1 to 4, characterized in that, The control method for the dissipation circuit module is as follows: The input energy is converted into heat energy and dissipated by the dissipation resistor (4). The duty cycle of the dissipation circuit PWM is controlled according to the current bus current and power supply voltage to control the duration of current passing through the dissipation resistor (4). According to the relationship between the bus power and the quadrature-axis direct-axis power of a permanent magnet synchronous motor (1): ； in, It is the quadrature axis voltage. For quadrature axis current, It is the direct-axis voltage. For direct-axis current, This is the bus voltage. Let be the bus current; after rearranging, we can obtain the formula for calculating the bus current: ； If the bus current is reverse-input to the power supply, it is called regenerative current. To protect the power supply, regenerative current is considered harmful, meaning the maximum regenerative current is zero. The power supply voltage can be directly obtained through ADC sampling, and the duty cycle setting can be calculated as follows: ； ； ； in The duty cycle is affected by both bus current and power supply voltage. Whether there is current from the reverse input power supply or the power supply voltage is too high, the duty cycle will increase. This is the resistance value of the dissipation resistor; To enable the relevant voltage parameters of the dissipation circuit, i.e., the power supply voltage reaches... The dissipation circuit is activated at the specified time to achieve [the desired effect]. When the dissipation circuit is fully activated, the PWM duty cycle is set to 1.

7. The intelligent upper limb motor rehabilitation device according to any one of claims 1 to 4, characterized in that, The torque self-sensing method based on a fully connected neural network is as follows: A five-layer fully connected neural network is constructed, consisting of one input layer, three hidden layers, and one output layer. The input layer includes five values: current velocity, previous velocity, current position error, previous position error, and current quadrature-axis current. The position error represents the difference between the actual position and the target position in position control mode. The three hidden layers are identical, each containing 12 neurons. The activation function for each neuron is the sigmoid activation function, as shown below: ； The output layer contains a single value representing the perceived torque, and no longer contains an activation function; Collect the real-time position error value, speed value, cross-axis current value of the internal permanent magnet synchronous motor (1) and the torque value recorded by the external torque sensor during the use of this rehabilitation device; organize the data, divide it into training set, validation set and test set, and normalize it; use the training set to train the fully connected neural network with the above structure, and use the following cross-entropy as the loss function: ； in The loss function represents the optimization objective. This represents the total number of samples in the training set. Indicates the current sample number. Indicates the first Each sample is processed by the output value of the neural network. Indicates the first The actual external torque value of each sample; due to It is a non-convex function, and the optimization method should be either the conjugate gradient method or the quasi-Newton method. Based on the external torque estimation method described above, the occurrence of spasms in users can be monitored during training. If a user experiences a spasm during rehabilitation training, the acceleration detected by the permanent magnet synchronous motor (1) and the estimated external torque will both generate a large peak value, then rapidly drop to a slightly lower value and remain for a period of time, and then drop to 0; this corresponds to the sudden increase in muscle tension when a spasm occurs, which hinders the original stretching activity, and then the muscle tension remains for a period of time, and finally disappears after the muscles relax. Acceleration and torque acceleration are used as the criteria for judging whether spasm has occurred. Thresholds are set for acceleration and torque acceleration. Once the actual values ​​of both exceed the threshold, spasm is considered to have occurred, and the training process is stopped immediately.

8. A smart upper limb motor rehabilitation device according to any one of claims 1 to 4, characterized in that, The specific usage method of the functional electrical stimulation module based on the force application zone is as follows: When the user continuously rotates this rehabilitation device, the force exertion of each upper limb exhibits a certain periodicity. One rotation of the training device can be divided into a left-hand force exertion zone and a right-hand force exertion zone. The left-hand force exertion zone is the angular range in which the device is rotated primarily by the left hand, and the right-hand force exertion zone is the angular range in which the device is rotated primarily by the right hand. It is assumed that the force exertion zones of the user's left and right hands should be equal in size, i.e., both 180°. For hemiplegic users, the healthy hand compensates for the force exertion zone of the affected hand. First, assess the force exertion zone of the healthy hand of the hemiplegic user; the complementary zone is the force exertion zone of the affected hand. Let the force exertion zone of the healthy hand be... °), Indicates the starting angle of the force application range of the unaffected hand; sets a certain rotational resistance, allowing the user to actively, continuously, and uniformly rotate the prototype using only the unaffected hand, through... Enumeration within range Calculate the average power of the healthy hand during the force application range for each value, to maximize the average power. The value represents the assessment result of the power exertion range; The functional electrical stimulation module is fixed to the user's upper arm via an upper arm fixing shell (30). The electrical stimulation sites are the supraspinatus, deltoid, biceps brachii, and triceps brachii on the hemiplegic side. Electrical stimulation is applied to the supraspinatus, deltoid, and triceps brachii in the force exertion zone on the affected side, and to the biceps brachii in the force exertion zone on the healthy side. The stimulation is performed alternately, with the stimulation frequency specified as a low frequency of 50Hz. The stimulation intensity is adjusted in real time according to the muscle strength sensed by this rehabilitation device. The calculation formula is as follows: ； in The intensity of the electrical stimulation is expressed in units of... ; This is the current maximum stimulation current value, which should be adjusted according to the patient's comfort level and should not exceed [the specified value]. ; The external torque value is estimated using a fully connected neural network. The peak torque applied to the current user.

Images