A control method and system for a multi-air-bag scoliosis intelligent brace based on LS-DYNA simulation driving

Abstract

This invention discloses a control method and system for a multi-airbag intelligent brace for scoliosis based on LS-DYNA simulation, belonging to the field of intelligent control technology for rehabilitation medical devices. The method includes: establishing a coupled finite element model of the human spine-trunk-multi-airbag brace using LS-DYNA; simulating and solving the target pressure values ​​and optimal force direction vectors of each airbag under different scoliosis stages and angles; then, combining a pressure acquisition unit and a posture detection unit to acquire the brace's working status in real time; using a single-neuron PID adaptive algorithm to dynamically fine-tune the airbag pressure; and using a direction vector calculation formula to dynamically correct the force direction. The system is compatible with the entire LS-DYNA finite element analysis design process. The single-neuron controller relies on supervised Hebb learning rules to update weight coefficients in real time, exhibiting strong robustness and adaptability, and can adapt to patients with different scoliosis levels and body types, significantly improving correction accuracy and dynamic response performance.

Description

Technical Field

[0001] This invention relates to the field of intelligent control technology for rehabilitation medical devices, and more specifically to a control method and system for a multi-bag intelligent scoliosis brace based on LS-DYNA simulation-driven design. Background Technology

[0002] Scoliosis is a common spinal deformity among adolescents, and the mainstream clinical treatment is conservative correction using orthotic braces. However, existing pneumatic scoliosis braces have several technical shortcomings: Traditional airbag braces can only roughly adjust the airbag pressure, controlling only the magnitude of the corrective force but not the precise direction of force application. The spine is a complex spatial biomechanical structure; deviations in the direction of force application can significantly reduce the corrective effect and even cause secondary injury. Existing braces generally use traditional fixed-parameter PID control, which suffers from difficulties in parameter tuning, weak anti-interference capabilities, and slow dynamic response. Pressure control accuracy drops drastically when faced with changes in patient position or airbag deformation. Control parameters are mostly set manually based on experience, without prior calibration using finite element biomechanical simulation, making it impossible to match the personalized correction needs of patients with different scoliosis angles and body types.

[0003] The LS-DYNA explicit dynamics simulation software can accurately simulate the contact, deformation, and mechanical transmission characteristics of the human spine, trunk soft tissue, and flexible airbags, and can output benchmark parameters such as optimal pressure and force direction. The single-neuron PID adaptive control algorithm, combined with supervised Hebb learning rules, can adaptively adjust the control weights online without the need for repeated manual parameter tuning, and has the advantages of fast response speed, strong robustness, and outstanding anti-interference ability.

[0004] Therefore, how to propose a control method and system for a multi-airbag intelligent brace for scoliosis based on LS-DYNA simulation-driven technology, and overcome the shortcomings of existing technologies, is a problem that urgently needs to be solved by those skilled in the art. Summary of the Invention

[0005] In view of this, the present invention provides a control method and system for a multi-airbag intelligent scoliosis brace driven by LS-DYNA simulation. This aims to overcome the shortcomings of traditional airbag scoliosis braces, such as a single force application dimension, poor performance of conventional PID control, and lack of biomechanical support for control parameters. The method utilizes the optimal control reference parameters output from LS-DYNA finite element simulation, employs a single-neuron PID adaptive controller combined with supervised Hebb learning rules to achieve dynamic fine-tuning of airbag pressure, and uses a space vector algorithm to correct the force application direction. It is fully compatible with the entire LS-DYNA simulation design process, improving the brace's correction accuracy, dynamic response capability, and environmental adaptability. To achieve the above objectives, the present invention adopts the following technical solution: A control method for a multi-airbag intelligent brace for scoliosis based on LS-DYNA simulation-driven design includes: S1: Establish a coupled finite element model, complete the preliminary simulation analysis based on LS-DYNA, and determine the basic control parameters of the multi-airbag array; S2: Collect real-time working condition data. Collect real-time pressure inside each airbag, overall posture of the brace and airbag deformation posture data through pressure sensor and attitude tilt sensor respectively. S3: Closed-loop control of the corrective force magnitude. Combining the target pressure value obtained from LS-DYNA simulation, a single-neuron PID adaptive controller is used and the weight coefficients are dynamically updated based on supervised Hebb learning rules to independently and adaptively adjust the pressure of each airbag for numerical control of the corrective force. S4: Corrective force direction vector calculation and dynamic correction. Based on the attitude acquisition data, a mathematical model of the force direction vector is established to calculate the deviation between the actual force direction and the simulation optimal direction, complete the adaptive adjustment of airbag position / deformation, and control the force direction. S5: Full-state cyclic monitoring and iterative control, continuously collects working condition data and repeats steps S2~S4 to form closed-loop control, maintaining stable output in both the magnitude and direction of the corrective force.

[0006] Optionally, S1 specifically includes: S101: Based on human CT / MRI image data, construct a three-dimensional solid model of human spine, ribs and trunk soft tissue in LS-DYNA. At the same time, establish an array-type airbag brace model composed of N independent airbags, complete model assembly, material property assignment, contact surface friction parameter setting, and form a coupled finite element model of spine-trunk-multi-airbag brace. S102: Mesh the model and set explicit dynamic solution parameters for LS-DYNA, targeting different scoliosis angles. Simulation experiments were conducted in groups based on different scoliosis types and patient body types. S103: Extract simulation results using the LS-DYNA post-processing module to obtain the simulated target pressure of the i-th airbag under each working condition. Optimal force direction unit vector airbag deformation threshold The parameters are stored in the controller parameter library; among them, N represents the total number of airbags in the brace.

[0007] Optionally, the formula for acquiring operating condition data in S2 is as follows: No. Real-time working pressure of each airbag: ; In the formula: For the first Real-time pressure of each airbag; This refers to the sensitivity coefficient of the pressure sensor. For the first The output voltage value of each pressure sensor; Euler angles of the support space attitude: ; In the formula: , , These are the Euler angles of the support's deflection about the X, Y, and Z axes, respectively. , , This refers to the sensitivity coefficient of the triaxial tilt sensor; , , This is the output voltage corresponding to the three-axis tilt sensor.

[0008] Optionally, in S3, the closed-loop control of the corrective force magnitude adopts single-neuron PID adaptive control, and the weight coefficients are updated in real time based on supervised Hebb learning rules, including pressure deviation calculation, neuron input calculation, control quantity solution, and weight coefficient iterative update. The specific formula is as follows: (1) No. Individual airbag pressure deviation value: ; In the formula: For the k-th control period Pressure deviation of each airbag; The first result obtained from LS-DYNA simulation Target pressure for each airbag; For the k-th period Real-time pressure of each airbag; (2) Three-way input features of a single neuron: ; In the formula: , The first , Cyclic pressure deviation; (3) Normalized weight coefficients of neurons: ; In the formula: , , The first Each airbag corresponds to the original weighting coefficients of the integral, proportional, and differential components; , , These are the normalized weighting coefficients; (4) Output of single-neuron PID control: ; In the formula: , The first , Cycle number Control output of each airbag; The proportionality coefficient of the neuron corresponding to the i-th air sac. ; (5) Iterative formula for weights in supervised Hebb learning rules: ; in: , , The first The learning rate of the integral, proportional, and differential components of each airbag; (6) No. Cycle number Each airbag ultimately controls the pressure: ; Controller according to Drive the inflation and deflation valve assembly to achieve adaptive adjustment of airbag pressure.

[0009] Optionally, the calculation and correction of the corrective force direction vector in S4 includes the calculation of the actual applied force vector, the solution of the vector deviation, and the calculation of the direction correction amount. The specific formulas are as follows: (1) Solve for the first Euler angle based on the support. The actual force vector per airbag: : ; (2) Magnitude of the force direction vector deviation: ; In the formula: For the first The deviation angle of the force applied by each airbag; , It is the dot product of the optimal direction vector and the actual direction vector; (3) Airbag deformation / position correction amount: ; In the formula: This represents the deformation / position adjustment amount of the i-th airbag; For direction correction scaling factor; when Exceeding the deformation threshold obtained from LS-DYNA simulation When this occurs, the limit protection is triggered, and adjustment stops.

[0010] Optionally, in the LS-DYNA, the trunk soft tissue adopts the Mooney-Rivlin hyperelastic constitutive model, and the constitutive equation is: ; In the formula: Strain energy density; , These are material constants; , The first and second strain invariants are defined; the airbag uses a thin-film unit, and the inflation pressure load boundary condition is set.

[0011] Optionally, the multi-airbag array is a partitioned independent structure, divided into multiple control zones according to the convex side, concave side, thoracic vertebrae, and lumbar vertebrae of the spine. The airbags in each zone are independently controlled in pressure and direction.

[0012] Optionally, when pressure deviation When the corrective force is deemed adequate, pressure adjustment is stopped; among other things, The pressure tolerance threshold is calibrated using LS-DYNA simulation results.

[0013] Optionally, when the direction deviation angle When the applied force direction is deemed satisfactory, the airbag position / deformation adjustment is stopped; among which, This is the allowable error threshold for the direction angle.

[0014] Optionally, a multi-airbag scoliosis intelligent brace control system based on LS-DYNA simulation-driven includes: a coupled finite element model establishment module: used to complete the pre-simulation analysis based on LS-DYNA and determine the basic control parameters of the multi-airbag array; Data Acquisition Module: Used to collect real-time working condition data, and collects real-time pressure inside each airbag, overall posture of the brace and airbag deformation posture data through pressure sensors and attitude tilt sensors respectively; Closed-loop control module: Used for closed-loop control of the corrective force. Combining the target pressure value obtained by LS-DYNA simulation, a single neuron PID adaptive controller is used and the weight coefficients are dynamically updated based on supervised Hebb learning rules to independently and adaptively adjust the pressure of each airbag and perform numerical control of the corrective force. The calculation and dynamic correction module is used to calculate and dynamically correct the force direction vector. It establishes a mathematical model of the force direction vector based on the attitude acquisition data, calculates the deviation between the actual force direction and the simulation optimal direction, completes the adaptive adjustment of the airbag position / deformation, and controls the force direction. Iterative control module: used for full-state cyclic monitoring and iterative control, continuously collects working condition data and repeats steps S2 to S4 to form closed-loop control, maintaining stable output in both the magnitude and direction of the corrective force.

[0015] As can be seen from the above technical solution, compared with the prior art, the present invention discloses a multi-airbag intelligent brace control method and system for scoliosis based on LS-DYNA simulation driving, which has the following beneficial effects: This invention uses finite element simulation results as the control benchmark, and all control parameters are based on biomechanical principles. It streamlines the entire process from simulation design to physical control, differing from traditional empirical parameter settings. Employing a single-neuron PID controller combined with supervised Hebb learning rules, it can autonomously update control weights online, solving the problems of difficult PID parameter tuning and weak anti-interference in traditional PID controllers. This results in fast pressure regulation response and strong robustness. It overcomes the limitation of traditional airbag braces that can only adjust force values, achieving simultaneous and precise control of the magnitude and direction of the corrective force, conforming to the biomechanical characteristics of the spine and reducing the risk of secondary injury. Through independent pressure control and orientation adjustment of multiple airbag zones, it can adapt to patients with different scoliosis angles, types, and body types, broadening its clinical applicability. The single-neuron adaptive algorithm and Hebb learning rules are mature, reliable, and highly practical for engineering implementation. Attached Figure Description

[0016] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0017] Figure 1 This invention provides a schematic flowchart of a multi-airbag intelligent brace control method for scoliosis based on LS-DYNA simulation. Detailed Implementation

[0018] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0019] This invention discloses a control method for a multi-airbag intelligent brace for scoliosis based on LS-DYNA simulation-driven technology, such as... Figure 1 As shown, it includes: S1: Establish a coupled finite element model, complete the preliminary simulation analysis based on LS-DYNA, and determine the basic control parameters of the multi-airbag array; S2: Collect real-time working condition data. Collect real-time pressure inside each airbag, overall posture of the brace and airbag deformation posture data through pressure sensor and attitude tilt sensor respectively. S3: Closed-loop control of the corrective force magnitude. Combining the target pressure value obtained from LS-DYNA simulation, a single-neuron PID adaptive controller is used and the weight coefficients are dynamically updated based on supervised Hebb learning rules to independently and adaptively adjust the pressure of each airbag for numerical control of the corrective force. S4: Corrective force direction vector calculation and dynamic correction. Based on the attitude acquisition data, a mathematical model of the force direction vector is established to calculate the deviation between the actual force direction and the simulation optimal direction, complete the adaptive adjustment of airbag position / deformation, and control the force direction. S5: Full-state cyclic monitoring and iterative control, continuously collects working condition data and repeats steps S2~S4 to form closed-loop control, maintaining stable output in both the magnitude and direction of the corrective force.

[0020] Furthermore, S1 specifically includes: S101: Based on human CT / MRI image data, construct a three-dimensional solid model of human spine, ribs and trunk soft tissue in LS-DYNA. At the same time, establish an array-type airbag brace model composed of N independent airbags, complete model assembly, material property assignment, contact surface friction parameter setting, and form a coupled finite element model of spine-trunk-multi-airbag brace. S102: Mesh the model and set explicit dynamic solution parameters for LS-DYNA, targeting different scoliosis angles. Simulation experiments were conducted in groups based on different scoliosis types and patient body types. S103: Extract simulation results using the LS-DYNA post-processing module to obtain the simulated target pressure of the i-th airbag under each working condition. Optimal force direction unit vector airbag deformation threshold The parameters are stored in the controller parameter library; among them, N represents the total number of airbags in the brace.

[0021] Furthermore, the formula for acquiring operating condition data in S2 is as follows: No. Real-time working pressure of each airbag: ; In the formula: For the first Real-time pressure of each airbag; This refers to the sensitivity coefficient of the pressure sensor. For the first The output voltage value of each pressure sensor; Euler angles of the support space attitude: ; In the formula: , , These are the Euler angles of the support's deflection about the X, Y, and Z axes, respectively. , , This refers to the sensitivity coefficient of the triaxial tilt sensor; , , This is the output voltage corresponding to the three-axis tilt sensor.

[0022] Furthermore, in S3, the closed-loop control of the corrective force magnitude adopts single-neuron PID adaptive control, and the weight coefficients are updated in real time based on supervised Hebb learning rules, including pressure deviation calculation, neuron input calculation, control quantity solution, and weight coefficient iterative update. The specific formula is as follows: (1) No. Individual airbag pressure deviation value: ; In the formula: For the k-th control period Pressure deviation of each airbag; The first result obtained from LS-DYNA simulation Target pressure for each airbag; For the k-th period Real-time pressure of each airbag; (2) Three-way input features of a single neuron: ; In the formula: , The first , Cyclic pressure deviation; (3) Normalized weight coefficients of neurons: ; In the formula: , , The first Each airbag corresponds to the original weighting coefficients of the integral, proportional, and differential components; , , These are the normalized weighting coefficients; (4) Output of single-neuron PID control: ; In the formula: , The first , Cycle number Control output of each airbag; The proportionality coefficient of the neuron corresponding to the i-th air sac. ; (5) Iterative formula for weights in supervised Hebb learning rules: ; in: , , The first The learning rate of the integral, proportional, and differential components of each airbag; In a specific embodiment, the original supervised Hebb learning rule is optimized by abandoning the fixed learning rate model for the integral, proportional, and derivative stages, and designing a bias-related variable learning rate strategy: The learning rate dynamically and adaptively adjusts based on real-time pressure and orientation deviations. A larger deviation results in a higher learning rate, accelerating response speed; when the deviation approaches a threshold, the learning rate automatically decreases to suppress overshoot and oscillations. The variable learning rate calculation formula is as follows: ; In the formula: , , This is the initial base learning rate; denoted as the rate decay coefficient. The optimized algorithm balances dynamic response speed and steady-state control accuracy, further enhancing the adaptability of the single-neuron controller under complex disturbance conditions.

[0023] In a specific embodiment, to address the pressure crosstalk and mechanical coupling issues present in multi-airbag arrays, a grouped, batch-based burst inflation / deflation control logic with branch pressure decoupling is proposed. The entire airbag system is divided into several control groups based on the thoracic vertebral segment, lumbar vertebral segment, convex side, and concave side. The synchronous adjustment mode for all airbags is abandoned, and a grouped, sequential burst adjustment method is adopted. During the adjustment of each airbag group, the inter-group coupling interference coefficient is pre-calibrated using LS-DYNA simulation, and a coupling compensation correction term is superimposed using a single-neuron PID algorithm to dynamically offset the effects of pressure transmission, deformation interference, and mechanical coupling between adjacent airbag groups. A corresponding inter-group coupling compensation formula is also added. ; In the formula: The coupling compensation deviation of the i-th group of airbags; The coupling interference coefficients of the m-th group to the i-th group calibrated for LS-DYNA simulation; Let M be the pressure deviation of the m-th airbag group; M is the total number of airbag groups. This strategy addresses the industry pain points of mutual interference and control oscillation during synchronous adjustment of multiple airbags, while simultaneously achieving coupling and decoupling of the flexible airbag system, significantly improving the control stability of the multi-unit array.

[0024] (6) No. Cycle number Each airbag ultimately controls the pressure: ; Controller according to Drive the inflation and deflation valve assembly to achieve adaptive adjustment of airbag pressure.

[0025] Furthermore, the calculation and correction of the corrective force direction vector in S4 includes the calculation of the actual applied force vector, the solution of the vector deviation, and the calculation of the direction correction amount. The specific formulas are as follows: (1) Solve for the first Euler angle based on the support. The actual force vector per airbag: : ; (2) Magnitude of the force direction vector deviation: ; In the formula: For the first The deviation angle of the force applied by each airbag; , It is the dot product of the optimal direction vector and the actual direction vector; (3) Airbag deformation / position correction amount: ; In the formula: This represents the deformation / position adjustment amount of the i-th airbag; For direction correction scaling factor; when Exceeding the deformation threshold obtained from LS-DYNA simulation When this occurs, the limit protection is triggered, and adjustment stops.

[0026] Furthermore, in the LS-DYNA, the trunk soft tissue adopts the Mooney-Rivlin hyperelastic constitutive model, and the constitutive equation is: ; In the formula: Strain energy density; , These are material constants; , The first and second strain invariants are defined; the airbag uses a thin-film unit, and the inflation pressure load boundary condition is set.

[0027] Furthermore, the multi-airbag array is a partitioned independent structure, divided into multiple control zones according to the convex side, concave side, thoracic vertebrae, and lumbar vertebrae of the spine. The airbags in each zone are independently controlled in pressure and direction.

[0028] Furthermore, when pressure deviation When the corrective force is deemed adequate, pressure adjustment is stopped; among other things, The pressure tolerance threshold is calibrated using LS-DYNA simulation results.

[0029] Furthermore, when the direction deviation angle When the applied force direction is deemed satisfactory, the airbag position / deformation adjustment is stopped; among which, This is the allowable error threshold for the direction angle.

[0030] Furthermore, based on the dual-dimensional control of the magnitude and direction of the corrective force, airbag deformation is introduced. As the third core control variable, based on LS-DYNA simulation to obtain the optimal deformation threshold for each airbag, a three-parameter linkage constraint model of pressure, direction, and deformation is established: when the airbag deformation approaches the limit threshold, the applied force direction and pressure output are adjusted first to limit excessive stretching / compression of the airbag; when the pressure deviation exceeds the threshold, deformation and direction are simultaneously fine-tuned to assist in pressure stabilization; the three parameters mutually constrain and coordinately regulate, rather than being controlled independently by a single variable. The three-parameter linkage constraint criteria are constructed as follows: ; In the formula: For the real-time deformation of the airbag, To simulate the optimal deformation, This is the allowable error threshold for deformation. It achieves integrated and coordinated control of mechanical output, spatial attitude, and device deformation, ensuring both correction effectiveness and protection of the airbag structure and human soft tissue, while balancing control precision, wearing comfort, and equipment lifespan.

[0031] In a specific implementation, a multi-airbag scoliosis intelligent brace control system based on LS-DYNA simulation-driven includes: a coupled finite element model establishment module: used to complete the pre-simulation analysis based on LS-DYNA and determine the basic control parameters of the multi-airbag array; Data Acquisition Module: Used to collect real-time working condition data, and collects real-time pressure inside each airbag, overall posture of the brace and airbag deformation posture data through pressure sensors and attitude tilt sensors respectively; Closed-loop control module: Used for closed-loop control of the corrective force. Combining the target pressure value obtained by LS-DYNA simulation, a single neuron PID adaptive controller is used and the weight coefficients are dynamically updated based on supervised Hebb learning rules to independently and adaptively adjust the pressure of each airbag and perform numerical control of the corrective force. The calculation and dynamic correction module is used to calculate and dynamically correct the force direction vector. It establishes a mathematical model of the force direction vector based on the attitude acquisition data, calculates the deviation between the actual force direction and the simulation optimal direction, completes the adaptive adjustment of the airbag position / deformation, and controls the force direction. Iterative control module: used for full-state cyclic monitoring and iterative control, continuously collects working condition data and repeats steps S2 to S4 to form closed-loop control, maintaining stable output in both the magnitude and direction of the corrective force.

[0032] In a specific implementation, a control method for a multi-airbag intelligent scoliosis brace based on LS-DYNA simulation-driven architecture is divided into five core components: LS-DYNA pre-simulation calibration, real-time working condition data acquisition, single-neuron PID pressure adaptive regulation, force direction vector correction, and cyclic closed-loop control. The specific steps include: Step S11: Establish a coupled finite element model based on LS-DYNA and complete the pre-simulation to calibrate the basic control parameters. This step connects the finite element simulation with the physical control system to obtain the optimal reference parameters under different side bending conditions.

[0033] (1) Model building

[0034] Based on the patient's CT / MRI medical images, a 3D model of the human spine, ribs, and trunk soft tissues was reconstructed. Simultaneously, an array-type brace model consisting of N independent air bladders was constructed, arranged in zones according to the thoracic vertebrae, lumbar vertebrae, and the convex / concave side of scoliosis. Model assembly and element generation were completed in LS-DYNA: the skeleton used solid elements, the trunk soft tissues used solid elements and were assigned a Mooney-Rivlin hyperelastic constitutive model, and the air bladders used thin-film elements.

[0035] Constitutive equation of soft tissue: ; In the formula: Strain energy density; , For human soft tissue materials, constants are used. , These are the first and second strain invariants.

[0036] (2) Simulation working condition settings

[0037] Boundary constraints and load conditions were set for different scoliosis angles, scoliosis types, and patient body types. The LS-DYNA explicit dynamic solution was then initiated to simulate the mechanical response of the spine and trunk under airbag compression.

[0038] (3) Parameter extraction and storage

[0039] The core parameters for each working condition are extracted using the LS-DYNA post-processing module: the simulated target pressure of the i-th airbag. Optimal force direction unit vector Maximum permissible deformation of airbags All parameters are categorized and stored in the intelligent controller parameter library as reference values ​​for physical control.

[0040] Step S12: Real-time working condition data acquisition. The brace integrates distributed pressure sensors and three-axis attitude tilt sensors to collect airbag pressure and brace spatial attitude data in real time, providing raw input for closed-loop control.

[0041] (1) Real-time pressure acquisition of airbag

[0042] The pressure sensor output voltage is linearly related to the airbag pressure. The real-time pressure of the i-th airbag is: ; In the formula: The real-time pressure (Pa) of the i-th airbag; The pressure sensor sensitivity coefficient (Pa / V); The sensor output voltage (V).

[0043] (2) Acquisition of brace attitude data

[0044] A three-axis tilt sensor acquires the Euler angles of the support around the X, Y, and Z axes: ; In the formula: , , Euler angles for three-axis deflection; , , This refers to the sensitivity coefficient of the tilt sensor. , , This is the output voltage of the sensor.

[0045] Step S13: Closed-loop adaptive control of corrective force magnitude

[0046] Target pressure calibrated by LS-DYNA To control the desired value, a single-neuron PID adaptive control algorithm is adopted, which relies on supervised Hebb learning rules to update the weight coefficients online, thereby achieving adaptive fine-tuning of airbag pressure.

[0047] (1) Calculate the pressure deviation

[0048] Pressure deviation of the i-th airbag in the k-th cycle: ; (2) Constructing three-way input features for a single neuron These correspond to the deviation, rate of change of deviation, and second-order rate of change of deviation in PID control, respectively: ; (3) Normalization of weight coefficients To ensure control stability, the original weights corresponding to the integral, proportional, and derivative are normalized: ; (4) Output calculation of single neuron PID control By combining the neuron scaling factor and the normalized weights, the control output for this cycle can be calculated: ; (5) Update weight coefficients based on supervised Hebb learning rules Following supervised learning rules, the weights of the integral, proportional, and derivative steps are iterated in real time to achieve algorithmic adaptation. ; in: , , These are the integral, proportional, and differential learning rates, respectively.

[0049] (6) Airbag pressure output control

[0050] By combining real-time pressure with neuron control output, the final control pressure is obtained: ; The controller drives the inflation / deflation solenoid valve to complete the inflation / deflation regulation. A pressure error threshold is set. ,when At that time, the pressure regulation is complete.

[0051] Step S14: Correction force direction vector calculation and dynamic correction. The actual force direction vector is calculated based on the attitude Euler angles. The deviation is calculated and the airbag position / deformation is adjusted to achieve precise control of the force direction.

[0052] (1) Calculation of the actual force direction vector

[0053] Derive the actual unit force vector of the i-th airbag from triaxial Euler angles: ; (2) Calculation of direction deviation angle The angle deviation between the actual and optimal directions can be solved using the dot product of spatial vectors. ; (3) Calculation of airbag orientation correction Calculate the airbag position / deformation adjustment amount based on the deviation angle: ; Set angle error threshold ,when Stop direction adjustment at the specified time; simultaneously, combine with the deformation threshold calibrated by LS-DYNA. Implement extreme protection to prevent excessive deformation of the airbag.

[0054] Step S15: The full-state cyclic iterative closed-loop control system executes processes S12~S14 cyclically according to a fixed control cycle, continuously collecting operating data, updating neuron weights online, and dynamically correcting the magnitude and direction of the corrective force. When the patient's position or spinal posture changes, the algorithm can autonomously adapt to the disturbance and maintain long-term stability of the corrective state.

[0055] In a specific embodiment, this embodiment is designed for adolescent patients with idiopathic thoracic scoliosis with a lateral curvature angle of 25°. The brace is equipped with 8 independent controllable airbags, which are divided into three major zones: the convex side of the thoracic spine, the concave side of the thoracic spine, and the lumbar spine. Each airbag is equipped with an independent pressure sensor, posture sensor, inflation and deflation valve, and single neuron control unit.

[0056] (1) LS-DYNA pre-simulation calibration

[0057] Based on patient images, a spine-trunk coupling model was reconstructed, and Mooney-Rivlin parameters were assigned to soft tissues. , After group simulation, the target pressure of 8 airbags was extracted. The optimal force vector and deformation threshold are stored in the controller parameter library.

[0058] (2) Parameter initialization

[0059] Initialize the parameters of the single neuron corresponding to each air sac: neuron scaling factor Integral / proportional / differential learning rate , , Initial weights , , Set pressure threshold Angle threshold .

[0060] (3) Real-time data acquisition

[0061] After the device is worn, the sensors collect the pressure of each airbag and the triaxial Euler angles of the brace in real time, and complete the physical quantity conversion through the corresponding formula.

[0062] (4) Single neuron PID pressure adaptive regulation

[0063] The system calculates pressure deviation, neuron input, normalized weights, and control output for each airbag, and updates the weight coefficients in real time using supervised Hebb learning rules to drive the inflation and deflation valves to achieve adaptive pressure adjustment. The system can quickly suppress pressure fluctuations caused by body position disturbances without the need for manual parameter tuning.

[0064] (5) Correction of the direction of force application

[0065] The actual applied force vector and deviation angle are calculated, the airbag adjustment amount is calculated, and the micro-actuator is driven to complete the position / deformation adjustment until the directional deviation meets the threshold requirement.

[0066] (6) Circular closed-loop operation

[0067] The system continuously cycles with a control period of 100ms, achieving stable control of both the magnitude and direction of the corrective force throughout the entire process.

[0068] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. For the apparatus disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the description is relatively simple; relevant parts can be referred to the method section.

[0069] The above description of the disclosed embodiments enables those skilled in the art to make or use the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A control method for a multi-airbag intelligent brace for scoliosis based on LS-DYNA simulation-driven design, characterized in that, include: S1: Establish a coupled finite element model, complete the preliminary simulation analysis based on LS-DYNA, and determine the basic control parameters of the multi-airbag array; S2: Collect real-time working condition data. Collect real-time pressure inside each airbag, overall posture of the brace and airbag deformation posture data through pressure sensor and attitude tilt sensor respectively. S3: Closed-loop control of the corrective force magnitude. Combining the target pressure value obtained from LS-DYNA simulation, a single-neuron PID adaptive controller is used and the weight coefficients are dynamically updated based on supervised Hebb learning rules to independently and adaptively adjust the pressure of each airbag for numerical control of the corrective force. S4: Corrective force direction vector calculation and dynamic correction. Based on the attitude acquisition data, a mathematical model of the force direction vector is established to calculate the deviation between the actual force direction and the simulation optimal direction, complete the adaptive adjustment of airbag position / deformation, and control the force direction. S5: Full-state cyclic monitoring and iterative control, continuously collects working condition data and repeats steps S2~S4 to form closed-loop control, maintaining stable output in both the magnitude and direction of the corrective force.

2. The method for controlling a multi-airbag intelligent brace for scoliosis based on LS-DYNA simulation-driven according to claim 1, characterized in that, S1 specifically includes: S101: Based on human CT / MRI image data, construct a three-dimensional solid model of human spine, ribs and trunk soft tissue in LS-DYNA. At the same time, establish an array-type airbag brace model composed of N independent airbags, complete model assembly, material property assignment, contact surface friction parameter setting, and form a coupled finite element model of spine-trunk-multi-airbag brace. S102: Mesh the model and set explicit dynamic solution parameters for LS-DYNA, targeting different scoliosis angles. Simulation experiments were conducted in groups based on different scoliosis types and patient body types. S103: Extract simulation results using the LS-DYNA post-processing module to obtain the simulated target pressure of the i-th airbag under each working condition. Optimal force direction unit vector airbag deformation threshold The parameters are stored in the controller parameter library; among them, N represents the total number of airbags in the brace.

3. The method for controlling a multi-airbag intelligent brace for scoliosis based on LS-DYNA simulation-driven according to claim 1, characterized in that, The formula for acquiring operating condition data in S2 is as follows: No. Real-time working pressure of each airbag: ; In the formula: For the first Real-time pressure of each airbag; This refers to the sensitivity coefficient of the pressure sensor. For the first The output voltage value of each pressure sensor; Euler angles of the support space attitude: ； In the formula: , , These are the Euler angles of the support's deflection about the X, Y, and Z axes, respectively. , , This refers to the sensitivity coefficient of the triaxial tilt sensor; , , This is the output voltage corresponding to the three-axis tilt sensor.

4. The method for controlling a multi-airbag intelligent brace for scoliosis based on LS-DYNA simulation-driven according to claim 1, characterized in that, The closed-loop control of the corrective force in S3 employs single-neuron PID adaptive control. The weight coefficients are updated in real time based on supervised Hebb learning rules, including pressure deviation calculation, neuron input calculation, control quantity solution, and iterative update of weight coefficients. The specific formula is as follows: (1) No. Individual airbag pressure deviation values: ; In the formula: For the k-th control period Pressure deviation of each airbag; The first result obtained from LS-DYNA simulation Target pressure for each airbag; For the k-th period Real-time pressure of each airbag; (2) Three-way input features of a single neuron: ； In the formula: , The first , Cyclic pressure deviation; (3) Normalized weight coefficients of neurons: ； In the formula: , , The first Each airbag corresponds to the original weighting coefficients of the integral, proportional, and differential components; , , These are the normalized weighting coefficients; (4) Output of single-neuron PID control: ； In the formula: , The first , Cycle number Control output of each airbag; The proportionality coefficient of the neuron corresponding to the i-th air sac. ; (5) Iterative formula for weights in supervised Hebb learning rules: ； in: , , The first The learning rate of the integral, proportional, and differential components of each airbag; (6) No. Cycle number Each airbag ultimately controls the pressure: ； Controller according to Drive the inflation and deflation valve assembly to achieve adaptive adjustment of airbag pressure.

5. The method for controlling a multi-airbag intelligent brace for scoliosis based on LS-DYNA simulation-driven according to claim 1, characterized in that, The calculation and correction of the corrective force direction vector in S4 includes the calculation of the actual applied force vector, the solution of the vector deviation, and the calculation of the direction correction amount. The specific formulas are as follows: (1) Solve for the first Euler angle based on the support. The actual force vector per airbag: ： ； (2) Magnitude of the force direction vector deviation: ； In the formula: For the first The deviation angle of the force applied by each airbag; , It is the dot product of the optimal direction vector and the actual direction vector; (3) Airbag deformation / position correction amount: ； In the formula: This represents the deformation / position adjustment amount of the i-th airbag; For direction correction scaling factor; when Exceeding the deformation threshold obtained from LS-DYNA simulation When this occurs, the limit protection is triggered, and adjustment stops.

6. The method for controlling a multi-airbag intelligent brace for scoliosis based on LS-DYNA simulation-driven according to claim 2, characterized in that, In the LS-DYNA, the soft tissue of the trunk adopts the Mooney-Rivlin hyperelastic constitutive model, and the constitutive equation is: ； In the formula: Strain energy density; , These are material constants; , The first and second strain invariants are defined; the airbag uses a thin-film unit, and the inflation pressure load boundary condition is set.

7. The method for controlling a multi-airbag intelligent brace for scoliosis based on LS-DYNA simulation-driven according to claim 1, characterized in that, The multi-airbag array is a partitioned independent structure, divided into multiple control zones according to the convex side, concave side, thoracic vertebrae, and lumbar vertebrae of the spine. The airbags in each zone are independently controlled in pressure and direction.

8. The control method for a multi-airbag intelligent brace for scoliosis based on LS-DYNA simulation driving according to claim 4, characterized in that, When pressure deviation When the corrective force is deemed adequate, pressure adjustment is stopped; among other things, The pressure tolerance threshold is calibrated using LS-DYNA simulation results.

9. The control method for a multi-airbag intelligent brace for scoliosis based on LS-DYNA simulation driving according to claim 5, characterized in that, When the direction deviation angle When the applied force direction is deemed satisfactory, the airbag position / deformation adjustment is stopped; among which, This is the allowable error threshold for the direction angle.

10. A multi-airbag intelligent brace control system for scoliosis based on LS-DYNA simulation-driven design, characterized in that, include: Coupled Finite Element Model Establishment Module: Used to complete the pre-simulation analysis based on LS-DYNA and determine the basic control parameters of the multi-airbag array; Data Acquisition Module: Used to collect real-time working condition data, and collects real-time pressure inside each airbag, overall posture of the brace and airbag deformation posture data through pressure sensors and attitude tilt sensors respectively; Closed-loop control module: Used for closed-loop control of the corrective force. Combining the target pressure value obtained by LS-DYNA simulation, a single neuron PID adaptive controller is used and the weight coefficients are dynamically updated based on supervised Hebb learning rules to independently and adaptively adjust the pressure of each airbag and perform numerical control of the corrective force. The calculation and dynamic correction module is used to calculate and dynamically correct the force direction vector. It establishes a mathematical model of the force direction vector based on the attitude acquisition data, calculates the deviation between the actual force direction and the simulation optimal direction, completes the adaptive adjustment of the airbag position / deformation, and controls the force direction. Iterative control module: used for full-state cyclic monitoring and iterative control, continuously collects working condition data and repeats steps S2 to S4 to form closed-loop control, maintaining stable output in both the magnitude and direction of the corrective force.

Images