A plantar stress measurement method and system based on finite element analysis

Abstract

This invention relates to the field of plantar stress measurement technology, and provides a plantar stress measurement method based on finite element analysis, comprising: S1: a flexible bionic interface is made of gradient hardness speckled silicone material with a silicone rubber matrix; S2: the subject stands or walks on the flexible bionic interface; S3: three-dimensional morphological data of the sole and instep are simultaneously acquired along with real-time deformation data of the flexible bionic interface; S4: the acquired three-dimensional morphological data and deformation data are preprocessed through a modular bus architecture system; S5: an algorithm combining finite element analysis and CNN deep learning network is used to perform pressure distribution inversion and near-real-time prediction on the deformation data; S6: the three-dimensional morphological data and the inverted pressure distribution data are spatially matched and coupled through a fusion module to form shape-pressure matching data. Under real weight-bearing conditions, the three-dimensional morphology and pressure distribution of the sole are simultaneously obtained to form shape-pressure matching data, enabling precise and rapid adaptation of braces and assistive devices.

Description

Technical Field

[0001] This invention relates to the technical field of plantar stress measurement, and more particularly to a plantar stress measurement method and system based on finite element analysis. Background Technology

[0002] As the core part of the human body in contact with the ground, the biomechanical characteristics of the sole are directly related to gait stability, foot injury risk assessment, foot disease rehabilitation effects, and the accuracy of personalized assistive device fitting. In fields such as human movement biomechanics research, clinical diagnosis and treatment of foot diseases (e.g., flat feet, high arches, plantar fasciitis), rehabilitation assistive device design, and sports protection product development, accurately acquiring three-dimensional morphology and pressure distribution data of the sole under real physiological weight-bearing conditions is the core foundation for achieving scientific assessment and precise intervention. With the increasing demand for precision in health management, sports protection, and rehabilitation treatment, traditional plantar stress measurement technology is no longer sufficient to meet the application needs in multiple scenarios. Developing a plantar morphology-pressure synchronous detection technology that combines realism, completeness, and practicality has become a key direction for industry development.

[0003] Currently, the industry has developed various foot detection technologies and representative products, including thin-film sensors from German brand Novel and American brand Tekscan, as well as rigid foot pressure plates from Belgian brand RScan Footscan and German brand Novel E-med. However, these existing technologies and products still have many insurmountable shortcomings, failing to achieve synchronous and accurate acquisition of the three-dimensional morphology and pressure distribution of the foot under real physiological weight-bearing conditions. Specific problems are as follows: First, the structural characteristics and acquisition modes of existing detection tools result in insufficient data authenticity and completeness. Among them, thin-film pressure sensors, due to their rigid material and lack of tensile properties, have poor conformability to the curved surface of the foot. In dynamic walking and running scenarios, gaps or local compression are easily generated, resulting in deviations between the collected pressure distribution data and the actual stress state of the foot, causing data distortion. Rigid foot pressure plates (such as Footscan and E-med) are limited by their structure and can only capture pressure data under a single fixed gait of the subject. They cannot cover the multi-gait changes during dynamic movement. Moreover, when their rigid interface comes into contact with the soft tissue of the foot, it will forcibly change the natural stress deformation state of the foot, causing abnormal compression of the soft tissue. Therefore, the detected deformation and pressure data cannot reflect the true biomechanical characteristics of the human body under physiological weight-bearing conditions. Plaster, foam molding, and rigid projection technologies all collect morphological data in a non-physiological state where the subject's foot is not bearing real weight or exercise load. They cannot capture the dynamic changes in the shape of the foot under weight-bearing conditions, nor can they establish the intrinsic relationship between morphological changes and pressure distribution. As a result, related products designed based on this data are difficult to adapt to real-world usage scenarios.

[0004] Secondly, existing technologies and products have significant shortcomings in terms of application orientation and commercialization adaptability. Currently, most advanced technologies and methods at home and abroad focus on achieving corrective functions through arch support structure design or local pressure relief through pressure-reducing pads made of a single material. They fail to fundamentally address the core requirement of "shape-pressure matching," which involves simultaneous acquisition and correlation analysis of form and pressure. Furthermore, these technologies struggle to meet the requirements for personalized and precise adaptation in terms of detection accuracy, and also have limitations in load adaptability to different weights and exercise intensities. This restricts their widespread application in diverse scenarios such as clinical rehabilitation, public sports protection, and personalized insole customization, thus limiting their commercialization.

[0005] Furthermore, the aforementioned representative products, such as the German Novel and American Tekscan thin-film sensors, the Belgian RScan Footscan, and the German Novel E-med pressure plate, generally suffer from the following common defects: First, they lack adhesion and can only achieve static data acquisition in a single step, lacking dynamic adaptability. Additionally, the rigid interface directly interferes with the natural deformation of the sole, further reducing the authenticity and reliability of the data. Second, in the design and adaptation of foot braces, there is a lack of coupled analysis of pressure and morphological data, resulting in the brace's structural design failing to accurately match the force requirements of different areas of the sole, leading to poor adaptation. Third, there is an overemphasis on the corrective effect of the arch shape, neglecting the multidimensional interactive forces generated by the sole during movement, as well as the complete relationship between morphology and pressure under real physiological weight-bearing conditions. This significantly reduces the effectiveness of braces and assistive devices, making it difficult to achieve the expected rehabilitation or protection goals.

[0006] Therefore, the industry urgently needs a technical solution that can simultaneously and accurately acquire three-dimensional morphology and pressure distribution data of the sole under real physiological weight-bearing conditions. By establishing a "shape-pressure matching" analysis mechanism, accurate acquisition of the real morphology and pressure distribution of the sole can be achieved, providing reliable data support for the precise design and rapid adaptation of foot braces, rehabilitation aids and sports protection products, thereby solving many defects of existing technologies. Summary of the Invention

[0007] To address the aforementioned problems, the present invention aims to provide a method and system for measuring plantar stress based on finite element analysis. Under actual load-bearing conditions, it simultaneously obtains the three-dimensional morphology and pressure distribution of the plantar surface, forming "shape-pressure matching" data to achieve precise and rapid adaptation of braces and assistive devices.

[0008] The above-mentioned objective of this invention is achieved through the following technical solutions: A method for measuring plantar stress based on finite element analysis includes the following steps: S1: Prepare and provide a flexible biomimetic interface, which is made of a gradient hardness speckled silicone material with a silicone rubber matrix. It has superelasticity and nonlinear stress-strain characteristics and is used to simulate the physiological interaction environment of the sole of the foot to capture the deformation of the sole of the foot under real weight-bearing conditions. S2: The subject stands or walks on the flexible bionic interface, so that the sole of the foot is in full contact with the flexible bionic interface, and produces a real deformation that conforms to the physiological state based on the gradient hardness characteristics of the interface. S3: Activate the multi-camera laser / structured light scanning system to simultaneously acquire three-dimensional morphological data of the sole and instep, as well as real-time deformation data of the flexible bionic interface. The scanning system undergoes dual calibration and adaptive algorithm correction to ensure data accuracy. S4: The system uses a modular bus architecture to preprocess the acquired 3D morphological and deformation data to obtain standardized analysis data. S5: Based on a material model calibrated in advance through mechanical experiments, an algorithm combining finite element analysis and CNN deep learning network is used to perform pressure distribution inversion and near real-time prediction on the preprocessed deformation data. S6: The fusion module spatially matches and couples the 3D morphological data with the inverted pressure distribution data to form morphological-pressure matching data, and outputs relevant application results for precise adaptation.

[0009] Further, step S1 specifically includes: Silicone rubber was selected as the matrix material, and a continuous gradient hardness distribution was designed to match the stress adaptation requirements of different areas of the sole. Scattered particles were uniformly introduced into the silicone rubber matrix material, and the deformation was visualized through the distribution of the scattered particles, providing a basis for subsequent digital image correlation analysis. The flexible biomimetic interface is divided into a surface deformation acquisition layer, an intermediate gradient transition layer, and a bottom load-bearing support layer along the thickness direction; the continuous gradient hardness distribution is achieved by at least one of the following methods: controlling the crosslinking agent ratio, filler concentration, curing temperature field, or layered casting sequence at different thickness positions of the silicone rubber matrix, so that the flexible biomimetic interface forms a monotonically varying elastic modulus distribution in the thickness direction. The flexible biomimetic interface material exhibits significant hyperelasticity and nonlinear stress-strain characteristics. Its stress-strain relationship is comprehensively calibrated through uniaxial tensile tests, biaxial tensile tests, and plane shear tests to ensure the accuracy of the material's mechanical property data. The flexible biomimetic interface was prepared using an integrated process of vacuum mixing, layered curing, and surface treatment. After preparation, the hardness gradient distribution and speckle particle distribution of the material were experimentally verified to ensure that they met the design requirements. The speckle particles were high-contrast inert particles or pigment microparticles that were compatible with the silicone rubber matrix. The speckle particles were arranged along the thickness direction of the flexible biomimetic interface in a manner with high density on the surface, transitional distribution in the middle layer, and low density or no speckle distribution at the bottom layer. By using mechanical performance testing and high-resolution imaging technology, the nonlinear strain response of the flexible bionic interface is systematically evaluated to ensure that the interface maintains good repeatability and stability within different stress load ranges on the sole of the foot, while always maintaining the bionic contact performance of the sole of the foot under different loads.

[0010] Further, step S3 specifically includes: The multi-camera laser / structured light scanning system has a detection range that fully covers the sole and instep areas. The system is equipped with a laser line source with a wavelength range of 450–850nm. The multi-camera laser / structured light scanning system includes 4 to 12 camera acquisition units, and each camera acquisition unit is distributed circumferentially relative to the foot center reference coordinate system. Before starting data acquisition, the scanning system is subjected to both geometric and optical calibration using standard target points. This dual calibration ensures that the coordinate systems of multiple cameras maintain strict consistency, thus avoiding data acquisition errors caused by coordinate system deviations. To address the differences in foot shape among different subjects and the possible posture changes that subjects may experience during the acquisition process, the scanning system incorporates an adaptive stitching algorithm. This algorithm can automatically identify image distortion problems caused by posture changes and occlusion of the sole and instep areas, and perform corresponding error corrections in real time, effectively avoiding stitching errors that are easily generated by traditional single-camera systems. Through the synergistic effect of system configuration, dual calibration, and adaptive algorithm correction, high-precision, continuous, and complete three-dimensional morphological data of the sole and instep are finally acquired. At the same time, real-time deformation data of the flexible bionic interface is captured simultaneously, avoiding the stitching errors of traditional single-camera systems and providing accurate and reliable geometric input data for subsequent mechanical inversion calculations.

[0011] Further, step S4 specifically includes: The system corresponding to the measurement method adopts a modular bus architecture, and each sub-module, including the data acquisition module, preprocessing module, and analysis module, communicates in real time through TCP / IP or USB 3.0 high-speed data channels. The data acquisition module synchronously records the three-dimensional point cloud data of the sole and instep and the deformation field data of the flexible bionic interface, and then transmits the acquired raw data to the preprocessing module. The preprocessing module sequentially performs denoising, gridding, coordinate correction, and time synchronization on the raw data to remove interference information, standardize the data format and time reference, and form standardized analysis data, providing high-quality data support for subsequent stress inversion calculations.

[0012] Further, step S5 specifically includes: With the establishment of the mapping relationship between deformation, stress, and pressure as the core, the mechanical parameters of the flexible biomimetic interface are comprehensively collected through uniaxial tensile tests, biaxial tensile tests, and plane shear tests. These mechanical parameters are used as core basic data to fit the relevant parameters of the Ogden model or Mooney-Rivlin model, thereby determining the characterization model that best matches the real mechanical properties of silicone, and ensuring the hyperelasticity accuracy in the subsequent finite element analysis process. A fusion computing architecture that deeply integrates finite element analysis and artificial intelligence deep learning algorithms is constructed. The finite element analysis module first performs professional calculations on the deformation data and outputs accurate deformation-stress feature parameters. Then, the deformation-stress feature parameters are used as dedicated training samples and input into the CNN deep learning network. The CNN deep learning network adopts a multi-layer residual structure design to specifically improve the model's generalization ability to nonlinear deformation data. Then, the AI ​​model is fully trained and optimized through large sample data to enable it to have the core computing ability to efficiently infer the pressure distribution from deformation data. In the actual test, the pre-processed flexible biomimetic interface deformation data is input into the trained AI model. The AI ​​model closely combines the calibrated hyperelastic material model and the physical constraints of finite element analysis to achieve rapid inversion and near real-time prediction from silicone deformation data to plantar pressure distribution. This algorithm, which integrates finite element analysis with CNN deep learning network, can significantly reduce the overall computational load compared to the traditional pure finite element calculation method, successfully achieve near real-time prediction of pressure distribution, and effectively retain the physical constraint characteristics of finite element analysis, ensuring that the final output pressure distribution prediction results are both reliable and interpretable.

[0013] Furthermore, the fusion of the AI ​​model and finite element analysis employs a physical constraint fusion logic. During model training and prediction, the mechanical conservation relations, material constitutive relations, boundary condition constraints, and contact relationship constraints obtained from finite element analysis are introduced as joint constraints into the CNN deep learning network. This ensures that the plantar pressure distribution results output by the AI ​​model simultaneously meet the requirements of data-driven prediction capability and mechanical-physical consistency. Specifically: The nodal displacement, strain tensor, stress tensor, contact pressure, contact area ratio, and contact state label output by the finite element analysis module are constructed as physical feature priors and input together with the preprocessed deformation data into the CNN deep learning network to form a dual-channel fusion input structure of data features and physical features. During network training, a joint loss function is constructed, which includes a data error loss term and a physical constraint loss term. The data error loss term is used to characterize the deviation between the predicted pressure distribution and the labeled pressure distribution, while the physical constraint loss term is used to constrain the prediction results to meet the hyperelastic response law of the material, the continuity of the contact boundary, the overall force balance relationship, and the local deformation coordination relationship. The weights of each loss term in the joint loss function can be adaptively adjusted according to the training stage, sample working conditions, or prediction error level. During the model prediction phase, after the CNN deep learning network outputs the initial pressure distribution results, the finite element analysis module performs a rapid physical consistency check on the initial results to determine whether they meet the preset mechanical rationality threshold. When there is abnormal pressure concentration in local areas, contact boundary crossing, false pressure in non-contact areas, or overall reaction force deviation exceeding the limit, the physical constraint correction unit is called to iteratively correct the prediction results and output the final plantar pressure distribution results that meet the physical consistency requirements. The physical constraint correction unit adopts a closed-loop correction mechanism based on finite element residual feedback, which feeds back the residual field, boundary deviation field or contact state deviation obtained from finite element verification to the intermediate layer or output layer of the CNN deep learning network to perform one or more local updates on the network prediction results.

[0014] Furthermore, steps S5 and S6 also include system operation assurance and scenario adaptation processes, specifically: The system has a built-in anomaly handling mechanism that monitors in real time for anomalies, including occlusion, deformation oversaturation, or data synchronization delay, during scanning and data processing. When an anomaly is detected, the system automatically starts an error correction program to correct it. If the anomaly cannot be eliminated through correction, the operator is prompted to recollect the data. For different application scenarios, including rehabilitation assistance, sports biomechanical analysis, and personalized insole customization, the system presets corresponding parameter templates and implements adaptive data processing strategies by switching templates to ensure the accuracy and efficiency of data processing in different scenarios. The inversion calculation stage is completed collaboratively by the finite element and AI modules. The pressure distribution data and coupled comprehensive data obtained from the inversion calculation are stored in a standardized JSON format. This format facilitates direct access by external software such as brace design software and rehabilitation assessment systems, thereby improving the convenience of data application.

[0015] Further, step S6 specifically includes: After receiving the three-dimensional morphological standardized data and pressure distribution prediction data, the fusion module performs precise spatial matching and deep coupling of the two types of data based on the spatial coordinate matching algorithm to form complete shape-pressure matching data. This data can comprehensively reflect the intrinsic relationship between foot morphology and pressure distribution under real physiological weight-bearing conditions. The system outputs application results including three-dimensional morphology-pressure coupling data, personalized brace design schemes, and plantar stress analysis reports; The shape-compression matching data and application results can be directly used for the precise design and rapid adaptation of foot braces, rehabilitation aids and sports protection products, providing scientific data support for clinical diagnosis and treatment, sports protection and health management.

[0016] A finite element analysis-based plantar stress measurement system for performing the above-described finite element analysis-based plantar stress measurement method includes: A flexible biomimetic interface fabrication module is used to prepare and provide a flexible biomimetic interface. The flexible biomimetic interface is made of gradient hardness speckled silicone material with silicone rubber matrix. It has superelasticity and nonlinear stress-strain characteristics and is used to simulate the physiological interaction environment of the sole of the foot to capture the deformation of the sole of the foot under real weight-bearing conditions. The subject interface deformation generation module is used when the subject stands or walks on the flexible bionic interface, so that the sole of the foot is in full contact with the flexible bionic interface, and generates a realistic deformation that conforms to the physiological state based on the gradient hardness characteristics of the interface. The three-dimensional morphology and deformation acquisition module is used to start the multi-camera laser / structured light scanning system to simultaneously acquire the three-dimensional morphological data of the sole and instep and the real-time deformation data of the flexible bionic interface. The scanning system is double-calibrated and corrected by an adaptive algorithm to ensure data accuracy. The data standardization preprocessing module is used to preprocess the acquired three-dimensional morphological data and deformation data through a modular bus architecture system to obtain standardized analysis data. The pressure distribution inversion and prediction module is used to perform pressure distribution inversion and near real-time prediction on preprocessed deformation data based on a material model calibrated through mechanical experiments and an algorithm that combines finite element analysis and CNN deep learning network. The form-pressure coupling result output module is used to spatially match and couple the three-dimensional morphological data with the inverted pressure distribution data through the fusion module to form form-pressure matching data and output relevant application results for accurate adaptation.

[0017] A computer device includes a memory and one or more processors, the memory storing computer code that, when executed by the one or more processors, causes the one or more processors to perform the method described above.

[0018] A computer-readable storage medium storing computer code that, when executed, performs the method described above.

[0019] Compared with the prior art, the present invention has at least one of the following beneficial effects: (1) Achieve accurate detection of "shape-pressure matching" of the sole. For the first time, a coupling acquisition method of three-dimensional shape and pressure distribution of the sole is proposed. This breaks through the limitation of existing technology that cannot simultaneously acquire two types of core data under real physiological weight-bearing conditions. It can fully and realistically restore the force characteristics and shape changes of the sole in natural standing, walking and other states, and fully present the intrinsic relationship between shape and pressure. This solves the defect of traditional technology that can only collect shape or pressure and cannot reflect the coupling relationship.

[0020] (2) Effectively avoid detection interference caused by hard interfaces. By adopting a gradient hardness speckled silicone flexible biomimetic interface with silicone rubber matrix, its continuous gradient hardness distribution and superelastic and nonlinear stress-strain characteristics can truly simulate the physiological interaction environment of the sole of the foot. This avoids the problems of abnormal compression of soft tissue and poor fit of the sole caused by the rigidity of materials in existing thin film sensors and hard pressure plates, ensuring that the collected morphological and deformation data fit the real physiological state of the human body and significantly improving the reliability of the data.

[0021] (3) Overcoming the challenge of nonlinear modeling of silicone hyperelasticity, the algorithm of finite element analysis and CNN deep learning network is combined with the material model (Ogden model or Mooney-Rivlin model) calibrated by uniaxial tension, biaxial tension and plane shear experiments. It retains the physical constraint characteristics of finite element analysis and improves the generalization ability of nonlinear deformation by using the AI ​​model of multilayer residual structure. Compared with pure finite element calculation, the amount of calculation is significantly reduced, and the pressure distribution is predicted in near real time. At the same time, the reliability and interpretability of the results are guaranteed.

[0022] (4) Significantly improve the accuracy of foot detection and related product design. Relying on the three-dimensional morphological data without splicing error obtained by the high-precision multi-camera laser / structured light scanning system, and the accurate pressure distribution data inverted by the finite element-AI fusion algorithm, the three-dimensional morphological-pressure coupling data formed by spatial matching and coupling provides comprehensive and reliable data support for the accurate design and rapid adaptation of foot braces, rehabilitation aids and sports protection products. It solves the problem that the existing technology has poor adaptation effect due to lack of coupling data and only focuses on the correction of foot arch shape while ignoring multi-dimensional interactive forces.

[0023] (5) Provides a commercially viable multidimensional interactive detection solution. The system adopts a modular bus architecture, supports TCP / IP or USB3.0 high-speed communication, and has functions such as automatic anomaly correction and multi-scenario parameter template switching. It is suitable for diverse application scenarios such as rehabilitation, sports analysis, and insole customization. Its advantages in detection accuracy, load adaptability and ease of operation break the limitations of commercial promotion of existing advanced methods. It can be widely used in clinical diagnosis and treatment, sports biomechanics research, personalized health protection and other fields, and has significant practical value and market prospects. Attached Figure Description

[0024] Figure 1 This is an overall flowchart of the plantar stress measurement method based on finite element analysis of the present invention; Figure 2 This is a structural diagram of the plantar stress measurement system based on finite element analysis of the present invention. Detailed Implementation

[0025] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0026] Those skilled in the art will understand that, unless specifically stated otherwise, the singular forms “a,” “an,” “the,” and “the” used herein may also include the plural forms. It should be further understood that the term “comprising” as used in this specification means the presence of the stated features, integers, steps, operations, elements, and / or components, but does not exclude the presence or addition of one or more other features, integers, steps, operations, elements, components, and / or groups thereof.

[0027] First Embodiment like Figure 1 As shown, this embodiment provides a method for measuring plantar stress based on finite element analysis, including the following steps: S1: Prepare and provide a flexible biomimetic interface, which is made of a gradient hardness speckled silicone material with a silicone rubber matrix. It has superelasticity and nonlinear stress-strain characteristics and is used to simulate the physiological interaction environment of the foot to capture the deformation of the foot under real weight-bearing conditions.

[0028] In this embodiment, step S1 specifically includes: Silicone rubber was selected as the matrix material, and a continuous gradient hardness distribution was designed. The surface hardness range was set to Shore A 0-10, and the bottom layer hardness reached Shore A 20-30 to match the stress adaptation requirements of different areas of the sole. Scattered speckle particles were uniformly introduced into the silicone rubber matrix material, and the deformation was visualized through the distribution of the speckle particles, providing a basis for subsequent digital image correlation analysis. The total thickness of the flexible bionic interface is set to 4mm-18mm, and it is divided into a surface deformation acquisition layer, an intermediate gradient transition layer and a bottom load-bearing support layer along the thickness direction. The surface deformation acquisition layer has a thickness of 0.5mm-3mm, the intermediate gradient transition layer has a thickness of 2mm-8mm, and the bottom load-bearing support layer has a thickness of 1mm-7mm, so as to ensure the sensitivity of foot contact while taking into account the overall structural stability and load recovery ability. The continuous gradient hardness distribution is achieved through at least one of the following methods: controlling the crosslinking agent ratio, filler concentration, curing temperature field, or layered casting sequence at different thickness locations of the silicone rubber matrix, so that the flexible biomimetic interface forms a monotonically varying elastic modulus distribution in the thickness direction; the hardness difference between adjacent layers is controlled within the range of Shore A 3-8, or the equivalent elastic modulus change rate is controlled at 5%-20%, so as to reduce interlayer stress abrupt changes and suppress local deformation distortion. The intermediate gradient transition layer is formed by continuous diffusion curing or multi-layer micro-differential hardness stack composite method. The multi-layer micro-differential hardness stack composite method includes setting 2 to 6 transition sub-layers, with each transition sub-layer arranged sequentially from low hardness to high hardness, so that the displacement field and strain field of the interface under pressure remain continuously changing, thereby improving the boundary consistency and inversion stability of subsequent finite element modeling. The flexible biomimetic interface material exhibits significant hyperelasticity and nonlinear stress-strain characteristics. Its stress-strain relationship is comprehensively calibrated through uniaxial tensile tests, biaxial tensile tests, and plane shear tests to ensure the accuracy of the material's mechanical property data. The flexible biomimetic interface was prepared using an integrated process of vacuum mixing, layered curing, and surface treatment. After preparation, the hardness gradient distribution and speckle particle distribution of the material were experimentally verified to ensure that they met the design requirements. The speckle particles are high-contrast inert particles or pigment microparticles that are compatible with the silicone rubber matrix. Their equivalent particle size is 20μm-300μm, the speckle coverage per unit area is 15%-45%, and the center-to-center distance between adjacent speckle particles is controlled to be 1.5 times-6 times the particle size, so as to maintain stable image grayscale features and traceability under different loading conditions. The speckle particles are arranged along the thickness direction of the flexible biomimetic interface in a manner with high density on the surface, transitional distribution in the middle layer, and low density or no speckle distribution at the bottom layer. The speckle concentration in the surface layer within a depth range of 0.1mm-1.0mm is higher than that in the lower layer, in order to enhance the ability to recognize small deformations near the foot contact surface and reduce the interference of deep particles on imaging clarity and deformation reconstruction accuracy. The speckle particles are at least one of spherical, irregular sheet-like, or short rod-shaped, and form a non-repeating speckle pattern by combining random distribution with local non-periodic perturbation, so as to avoid periodic texture matching and aliasing in the process of digital image correlation analysis and improve the sub-pixel level displacement recognition accuracy. By using mechanical performance testing and high-resolution imaging technology, the nonlinear strain response of the flexible bionic interface is systematically evaluated to ensure that the interface maintains good repeatability and stability within different load ranges on the sole of the foot, while maintaining the bionic contact performance of the sole under different loads. This effectively avoids the problem of sole deformation distortion caused by rigid interfaces, and ultimately achieves accurate capture of the sole morphology under real physiological weight-bearing conditions.

[0029] The core of the design of step S1 above is to create a flexible biomimetic interface that adapts to the physiological characteristics of the sole of the foot by controlling the entire process of material selection, structural optimization, process assurance and performance verification, so as to solve the problem of data distortion caused by poor interface rigidity and fit in existing detection technologies from the source.

[0030] From a materials design perspective, the selection of silicone rubber as the matrix and the construction of a continuous gradient hardness distribution of Shore A 0-10 on the surface and Shore A 20-30 on the bottom are targeted designs based on the differences in stress intensity in different areas of the sole (such as the arch, heel, and forefoot). The low hardness of the surface layer adapts to the fit requirements of the soft areas of the sole, while the slightly higher hardness of the bottom layer ensures the stability of the interface support, avoiding the shortcomings of a single hardness material that cannot balance "fit" and "support". The introduction of speckle particles into the material is to provide clear deformation tracking marks for subsequent digital image correlation analysis, so that the small deformations of the flexible interface can be accurately captured, providing intuitive deformation evidence for pressure inversion.

[0031] In terms of mechanical property calibration and process fabrication, the stress-strain relationship is comprehensively calibrated through three tests: uniaxial tension, biaxial tension, and plane shear. This is because the hyperelasticity and nonlinearity of flexible silicone directly affect the mapping accuracy of deformation and pressure. Only by obtaining complete parameters through multi-dimensional mechanical tests can an accurate material mechanics basis be provided for subsequent finite element model construction and AI inversion. The integrated process of "vacuum mixing-layer curing-surface treatment" is adopted to ensure the continuous uniformity of gradient hardness and the uniform distribution of speckle particles, avoiding local hardness abrupt changes or speckle aggregation due to process defects, which would affect the realism of foot contact. The experimental verification of hardness gradient and speckle distribution after fabrication, as well as mechanical performance testing and high-resolution imaging evaluation, are to eliminate unqualified products through actual testing, ensuring that each flexible biomimetic interface can maintain stable repeatability and nonlinear strain response under different loads on the foot (from static standing to dynamic walking), and always maintain a biomimetic contact state with the foot, fundamentally avoiding problems such as abnormal compression and deformation distortion of soft tissue on the foot caused by hard interfaces.

[0032] In summary, step S1, through multi-dimensional technical design and strict quality control, enables the flexible bionic interface to possess both the soft and conforming characteristics that simulate physiological interaction of the sole of the foot, and the stable mechanical properties that support deformation detection. This successfully achieves the accurate capture of the sole morphology under real physiological load conditions, laying a core foundation for subsequent three-dimensional morphology acquisition, pressure distribution inversion, and the formation of "shape-pressure matching" data.

[0033] S2: The subject stands or walks on the flexible bionic interface, so that the sole of the foot is in full contact with the flexible bionic interface, and produces a realistic deformation that conforms to the physiological state based on the gradient hardness characteristics of the interface.

[0034] The core design logic of step S2 above is to simulate the real physiological activity scenario of the human body, so that the flexible bionic interface can form an interaction with the sole of the foot in a way that conforms to the natural state, providing a real and effective deformation basis for subsequent data collection, and avoiding the data distortion problem caused by the detection environment deviating from the physiological reality in existing technologies from the scenario level.

[0035] From the perspective of scenario simulation, requiring subjects to contact the flexible bionic interface while standing or walking is because these two states cover the core physiological weight-bearing scenarios of the human foot. Standing corresponds to static weight-bearing, while walking corresponds to dynamic weight-bearing, which can fully reproduce the force changes of the foot under different movement states, overcoming the limitation of traditional rigid pressure plates that can only collect single-step data. At the same time, the design requirement of "full contact" is achieved based on the hyperelasticity and gradient hardness characteristics of the flexible bionic interface: the low hardness design of the surface Shore A 0-10 allows soft areas of the foot (such as the inner arch) to fully conform to the interface, while the slightly higher hardness design of the bottom Shore A 20-30 provides stable support in areas with concentrated force, such as the heel and forefoot, avoiding insufficient contact or local suspension due to insufficient interface support, and ensuring that every area of ​​the foot can form an effective interaction with the interface.

[0036] From the perspective of ensuring the authenticity of deformation, the "physiological conformity" of foot deformation relies entirely on the gradient hardness characteristics of the flexible biomimetic interface. Different areas of the human foot exhibit significant differences in stress intensity and soft tissue thickness. The heel and forefoot, as the main stress points, require a certain amount of support to maintain natural deformation; while areas such as the arch experience less stress and require a soft, adaptable interface to preserve their physiological curvature. The gradient hardness interface can generate differentiated elastic deformation based on the stress feedback from different areas of the foot. Areas with high stress cause moderate deformation of the slightly harder material at the bottom of the interface, while areas with low stress achieve conformal deformation through the softer material on the surface. This avoids the abnormal deformation caused by forcibly compressing soft tissue like a rigid interface, and also avoids the loss of support due to excessive deformation like a single soft material. Ultimately, it forms a realistic deformation consistent with the natural physiological weight-bearing state of the human body, providing the most accurate original deformation data for subsequent 3D morphological acquisition and pressure distribution inversion.

[0037] In summary, step S2 successfully constructed a physiological interaction environment between the sole of the foot and the detection interface through a combination of "real scene simulation + interface feature adaptation". The resulting real deformation is a key prerequisite for the subsequent "shape-pressure matching" data acquisition, directly ensuring the authenticity and reliability of the subsequent detection data.

[0038] S3: Activate the multi-camera laser / structured light scanning system to simultaneously acquire three-dimensional morphological data of the sole and instep, as well as real-time deformation data of the flexible bionic interface. The scanning system undergoes dual calibration and adaptive algorithm correction to ensure data accuracy.

[0039] In this embodiment, step S3 specifically includes: The multi-camera laser / structured light scanning system has a detection range that fully covers the sole and instep areas. The system is equipped with a laser line source with a wavelength range of 450–850nm, a scanning resolution of 0.05~0.1mm, and a frame rate of 30~60fps. It can simultaneously meet the acquisition requirements of static and dynamic scanning and is suitable for data acquisition in different motion states of the subject, including standing and walking. The multi-camera laser / structured light scanning system includes 4 to 12 camera acquisition units, of which at least 2 camera acquisition units are arranged on the sole side to acquire deformation information of the sole contact area and flexible bionic interface, at least 2 camera acquisition units are arranged on both sides of the foot to acquire lateral contour information of the foot edge and arch, and at least 1 to 2 camera acquisition units are arranged obliquely above the dorsum of the foot to acquire morphological information of the dorsum of the foot and the upper surface of the toes, thereby constructing a multi-view closed-loop acquisition space covering the sole, side and top of the foot; Each camera acquisition unit is circumferentially distributed relative to the foot center reference coordinate system and is set towards the target area with a pitch angle of 10°-60° and an included angle of 20°-90°. The foot-side camera acquisition unit adopts a low-elevation angle through-beam or cross-view arrangement, the foot-side camera acquisition unit adopts an inclined surround arrangement, and the dorsum-side camera acquisition unit adopts a top oblique view arrangement to reduce the impact of toe occlusion, arch shadow area and edge reflection area on the integrity of point cloud reconstruction. Before starting data acquisition, the scanning system is subjected to both geometric and optical calibration using standard target points. This dual calibration ensures that the coordinate systems of multiple cameras maintain strict consistency, thus avoiding data acquisition errors caused by coordinate system deviations. Synchronous acquisition employs a mechanism combining hardware-triggered synchronization and timestamp alignment correction. Each camera acquisition unit, laser line light source or structured light projection unit, and deformation data acquisition module is connected to a unified clock reference. The central control module issues a synchronization trigger signal to ensure that each acquisition channel completes exposure, projection, and data buffering within the same sampling period. In dynamic acquisition scenarios, sub-frame level alignment correction is performed based on timestamps. To address the differences in foot shape among different subjects and the possible posture changes that subjects may experience during the acquisition process, the scanning system incorporates an adaptive stitching algorithm. This algorithm can automatically identify image distortion problems caused by posture changes and occlusion of the sole and instep areas, and perform corresponding error corrections in real time, effectively avoiding stitching errors that are easily generated by traditional single-camera systems. The geometric calibration includes the joint calibration of the intrinsic and extrinsic parameters and relative pose relationships of each camera acquisition unit. The optical calibration includes the calibration of the projection angle, projection plane position, light intensity uniformity, and imaging distortion of the laser line source or structured light projection unit. The dual calibration adopts a process that combines static calibration and dynamic verification. First, static parameters are solved using a planar calibration plate, a three-dimensional calibration body, or a multi-layer target array with known characteristic dimensions. Then, temporal consistency verification and field-of-view edge error compensation are performed using a dynamic calibration component that simulates foot trajectory movement. After calibration, the system reconstruction accuracy is evaluated for errors. The evaluation indicators include at least reprojection error, root mean square error of 3D reconstruction, cross-view stitching error, and time synchronization error. The reprojection error is no greater than 0.3 pixels, the root mean square error of 3D reconstruction is no greater than 0.20 mm, and the cross-view stitching error is no greater than 0.30 mm. When any error indicator exceeds the preset threshold, the system automatically performs local recalibration, view weight reallocation, or acquisition parameter reconfiguration to ensure the reliability and consistency of the geometric boundary data required for subsequent finite element inversion. Through the synergistic effect of system configuration, dual calibration, and adaptive algorithm correction, high-precision, continuous, and complete three-dimensional morphological data of the sole and instep are finally acquired. At the same time, real-time deformation data of the flexible bionic interface is captured simultaneously, avoiding the stitching errors of traditional single-camera systems and providing accurate and reliable geometric input data for subsequent mechanical inversion calculations.

[0040] The core design logic of step S3 above is to achieve synchronous and high-precision acquisition of foot morphology and interface deformation data through the synergy of a high-configuration scanning system, a strict calibration process and intelligent algorithm optimization. This provides an accurate geometric and deformation basis for subsequent "shape-compression matching" analysis and specifically addresses the pain points of traditional acquisition technology, such as "incomplete data, insufficient accuracy and poor adaptability".

[0041] From a system hardware configuration perspective, the scanning system's design, covering the entire area of ​​the sole and instep, breaks through the limitations of traditional acquisition technologies that only focus on local areas of the sole, ensuring the acquisition of complete three-dimensional morphological information of the foot. The configuration of a 450–850nm laser line light source is based on the imaging stability of this wavelength on biological tissues and flexible material surfaces, effectively reducing interference such as reflection and insufficient penetration. The high-precision design with a scanning resolution of 0.05–0.1mm can capture subtle textures and morphological undulations of the sole, meeting the need for precise orthotic fitting and detailed data. The 30–60fps frame rate caters to both static standing (low frame rate for efficient acquisition) and dynamic walking (high frame rate for capturing instantaneous changes), overcoming the limitation of traditional rigid pressure plates that can only acquire single-step data, achieving full data coverage under different movement states.

[0042] In terms of data accuracy assurance, dual geometric and optical calibration before data acquisition is a key design feature: geometric calibration ensures that the physical installation positions of multiple cameras are consistent with the measurement coordinate system, avoiding spatial positioning errors caused by hardware placement deviations; optical calibration corrects optical problems such as camera lens distortion and light refraction, ensuring the authenticity of the image. This combination of dual calibrations eliminates system errors from both physical and optical perspectives, laying the foundation for the fusion of multi-source data. Furthermore, the built-in adaptive stitching algorithm achieves "dynamic error correction" to address differences in foot types (such as flat feet, high arches, and normal feet) and fluctuations in the subject's posture during acquisition (such as instability or gait deviation). By automatically identifying image distortion in deformed areas and occlusions on the sole and instep (such as overlapping toes or arch shadows), the stitching parameters are adjusted in real time, completely avoiding the problems of traditional single-camera systems that rely on fixed-angle acquisition and are prone to stitching breaks or overlap errors, ultimately obtaining continuous and complete three-dimensional morphological data.

[0043] Most importantly, step S3 emphasizes the "synchronous acquisition" of three-dimensional morphological data of the sole and instep, as well as real-time deformation data of the flexible bionic interface. This design directly serves the core requirement of "shape-pressure matching": morphological data reflects the spatial structure of the sole, and deformation data is associated with the force feedback of the sole. The synchronicity of the two ensures that each morphological feature can correspond to the corresponding force deformation state, providing original data from the same source and synchronously for the subsequent finite element analysis and AI inversion to establish a precise mapping relationship of "shape-deformation-pressure". This avoids the defects of asynchronous acquisition of morphological and pressure data and the inability to establish a coupling relationship in traditional technologies.

[0044] In summary, step S3, through a multi-layered design of "high-configuration hardware to ensure the range and accuracy of data acquisition + dual calibration to eliminate system errors + adaptive algorithm to deal with dynamic interference + synchronous acquisition to lock the coupling relationship," successfully acquired high-precision, complete, and synchronized core data. This became a key bridge connecting the front-end bionic interface and the back-end pressure inversion, directly determining the reliability and accuracy of subsequent "shape-pressure matching" data.

[0045] S4: The system uses a modular bus architecture to preprocess the acquired 3D morphological and deformation data to obtain standardized analysis data.

[0046] In this embodiment, step S4 specifically includes: The system corresponding to the measurement method adopts a modular bus architecture, and each sub-module, including the data acquisition module, preprocessing module, and analysis module, communicates in real time through TCP / IP or USB 3.0 high-speed data channels. The data acquisition module synchronously records the three-dimensional point cloud data of the sole and instep and the deformation field data of the flexible bionic interface, and then transmits the acquired raw data to the preprocessing module. The preprocessing module sequentially performs denoising, gridding, coordinate correction, and time synchronization on the raw data to remove interference information, standardize the data format and time reference, and form standardized analysis data, providing high-quality data support for subsequent stress inversion calculations. The denoising process includes removing outliers, isolated points, suppressing edge spikes, and filtering reflection anomalies in the 3D point cloud data. It adopts a combination of statistical outlier detection and radius neighborhood constraints to identify anomalies based on the mean distance of the point neighborhood, local point density, and curvature change, so as to reduce the interference of pseudopoints introduced by uneven laser reflection, structured light shadow occlusion, and high curvature areas at the foot edges. The denoising process also includes performing image grayscale equalization, speckle contrast enhancement, and local distortion correction on the deformation field data of the flexible bionic interface to improve the speckle traceability on which digital image correlation analysis depends; for dynamically acquired sequences, a combination of temporal smoothing and inter-frame consistency constraints is used to suppress instantaneous jump noise so that the displacement changes between consecutive frames meet the requirements of physiological motion continuity. The meshing process includes surface reconstruction and topological regularization of the denoised 3D point cloud data. A triangular mesh reconstruction method based on local curvature adaptive sampling is adopted. The mesh density is increased in the high deformation sensitive areas of the heel, metatarsal heads and toes, and the mesh density is reduced in the arch and relatively flat areas, so as to retain the geometric details of the key stress areas of the sole while controlling the amount of computation. The meshing process also includes a mesh quality optimization process, which repairs the elongated cells, flipped cells, self-intersecting cells and hole regions in the initial mesh, and improves the quality of the mesh cells through edge length uniformization, normal unification and local smoothing, so that the generated surface mesh meets the continuity and stability requirements required for subsequent finite element model construction. The coordinate correction includes mapping the three-dimensional point cloud data of the sole and instep, the deformation field data of the flexible bionic interface, and the coordinate system of the scanning system to the same foot reference coordinate system. The foot reference coordinate system is established based on the heel center, the direction of the second metatarsal bone, and the sole contact reference plane. The anterior-posterior direction is defined as the long axis direction of the foot, the lateral direction is defined as the width axis direction of the foot, and the vertical direction is defined as the normal direction, so as to ensure the consistency of different modal data in the spatial registration process. The coordinate correction adopts a multi-stage registration method that combines coarse registration based on feature marker points with iterative nearest point fine registration. In the coarse registration stage, the initial correspondence is established using the geometric features of the heel point, the first metatarsal head point, the fifth metatarsal head point, and the toe tip region. In the fine registration stage, iterative optimization is performed by combining the local surface normal, the regional curvature distribution, and the contact boundary contour, so that the average spatial deviation after registration is no more than 0.50 mm. The time synchronization process includes aligning the 3D point cloud data and deformation field data at the frame level according to a unified timestamp, and completing the time sequence unification between heterogeneous data streams based on trigger pulses, buffer delay compensation and interpolation resampling mechanisms; for data channels with frame rate differences, time reconstruction is performed through at least one of linear interpolation, spline interpolation or Kalman prediction, so that data at different sampling frequencies correspond at the same analysis time. The preprocessing module also performs integrity assessment and confidence marking on the standardized analysis data. The integrity assessment includes at least point cloud coverage, hole area ratio, speckle recognition rate, and temporal continuity index. When a local region is missing, occlusion exceeds the limit, or the displacement of adjacent frames changes beyond a preset threshold, a local completion, abnormal frame removal, or reconstruction weight adjustment mechanism is triggered to improve the stable support capability of the standardized analysis data for subsequent stress inversion results. Among them, the local completion process adopts at least one of the following methods: neighborhood surface fitting, symmetric region mapping or temporal frame joint compensation, to restore the data of the foot edge occlusion area, the toe gap defect area and the local distortion area of ​​the flexible bionic interface, so as to reduce the problem of finite element boundary discontinuity caused by the acquisition blind zone. The standardized analysis data generated after preprocessing includes at least the point cloud of the sole and instep surfaces in a unified coordinate system, the topologically continuous mesh of the foot surface, the flexible biomimetic interface deformation field corresponding to the time, the quality assessment labels of each data frame, and the boundary index information for subsequent inversion calculations, thus providing a data foundation that is structurally complete, spatiotemporally consistent, and directly callable for solving the plantar stress field.

[0047] The core design logic of step S4 above is to achieve "de-cluttering, standardization, and synchronization" of the raw collected data through an efficient system architecture and standardized preprocessing procedures. This transforms the discrete and interference-laden raw data into high-quality analytical data suitable for subsequent calculations, playing a crucial role in bridging the gap between the preceding and following steps. It not only inherits the synchronous acquisition results from step S3 but also provides an accurate and unified data foundation for the stress inversion in step S5, solving the pain points of "communication lag, data disorder, and format incompatibility" in traditional data processing.

[0048] From a system architecture perspective, the adoption of a modular bus architecture and the use of TCP / IP or USB 3.0 high-speed data channels for real-time communication between sub-modules is based on the requirements for high efficiency and collaboration in data processing. The modular design separates data acquisition, preprocessing, and analysis functions into independent modules, ensuring clear division of labor and flexible adaptation to different scenarios (such as rehabilitation and sports analysis). Meanwhile, TCP / IP and USB 3.0, as mature high-speed communication protocols, can meet the requirements for high-capacity, low-latency transmission of 3D point cloud data and deformation field data. Step S3 acquires rich detail in the plantar morphology data (resolution of 0.05~0.1mm) and strong real-time performance of the deformation data (frame rate of 30~60fps). The high-speed data channel avoids data packet loss or delay during transmission, ensuring the integrity and synchronization of the original data and laying the hardware foundation for subsequent time synchronization processing.

[0049] In terms of data transmission and preprocessing, the design logic revolves around the transformation from "raw data to standardized data": First, the data acquisition module synchronously records 3D point cloud data (discrete plantar / foot surface spatial coordinate information) and deformation field data (continuous flexible interface deformation information). These two types of data come from different sources and have different formats; directly using them for subsequent calculations would lead to chaotic model input. Therefore, they need to be transmitted to the preprocessing module for unified standardization. Second, the preprocessing module executes a four-step process: "denoising → meshing → coordinate correction → time synchronization." Each step specifically addresses the core issues of the raw data: denoising removes ambient light interference and electrical interference from the scanning device itself. Irrelevant information such as sub-noise is eliminated to ensure that the data reflects the true shape and deformation characteristics; meshing processing transforms discrete 3D point cloud data into structured mesh data, which not only meets the mesh division requirements of geometric models for finite element analysis, but also facilitates feature extraction by AI models; coordinate correction further optimizes the coordinate consistency of multi-camera acquisition, forming a dual guarantee of "acquisition-processing" with the dual calibration in step S3, completely eliminating coordinate deviation; time synchronization processing accurately aligns the timestamps of 3D shape data and deformation data, ensuring that each set of shape information can correspond to the deformation state at the same moment, perfectly matching the "shape-deformation" spatiotemporal coupling relationship required for "shape-compression matching".

[0050] In summary, step S4, through a two-layer design of "efficient communication architecture to ensure data transmission quality + standardized preprocessing process to optimize data format," successfully transforms the raw collected data into standardized analysis data with unified format, clean data, and spatiotemporal synchronization. This not only solves the problem of subsequent calculation errors caused by data disorder in traditional data processing, but also strengthens the coupling relationship between morphology and deformation data through time synchronization and coordinate correction. This provides solid data support for the accurate calculation of finite element analysis and AI deep learning in step S5, and is an important data preprocessing hub for achieving "form-compression matching".

[0051] S5: Based on a material model calibrated in advance through mechanical experiments, an algorithm combining finite element analysis and CNN deep learning network is used to perform pressure distribution inversion and near real-time prediction on the preprocessed deformation data.

[0052] In this embodiment, step S5 specifically includes: With the establishment of the mapping relationship between deformation, stress, and pressure as the core, the mechanical parameters of the flexible biomimetic interface are comprehensively collected through uniaxial tensile tests, biaxial tensile tests, and plane shear tests. These mechanical parameters are used as core basic data to fit the relevant parameters of the Ogden model or Mooney-Rivlin model, thereby determining the characterization model that best matches the real mechanical properties of silicone, and ensuring the hyperelasticity accuracy in the subsequent finite element analysis process. In the finite element analysis modeling process, the preprocessed three-dimensional geometric data of the sole and instep and the real-time deformation data of the flexible bionic interface are used as boundary inputs to construct a multi-region coupled finite element model including the foot tissue region, the flexible bionic interface region and the support contact region. The foot tissue region includes at least the soft tissue surface of the sole and the outer contour boundary of the foot. The flexible bionic interface region serves as the medium for transmitting compressive deformation, and the support contact region is used to characterize the mechanical action of the external contact surface during standing or gait. The boundary conditions of the finite element model include at least geometric boundary conditions, displacement boundary conditions, load boundary conditions, and constraint boundary conditions. The geometric boundary conditions are determined by the standardized surface point cloud and mesh output in step S4. The displacement boundary conditions are applied by the surface displacement field or key node displacement measured by the flexible bionic interface. The load boundary conditions are calculated from the subject's body mass, posture state, or gait period parameters. The constraint boundary conditions are used to limit the rigid body drift and overall degrees of freedom in the non-contact area. A fusion computing architecture that deeply integrates finite element analysis and artificial intelligence deep learning algorithms is constructed. The finite element analysis module first performs professional calculations on the deformation data and outputs accurate deformation-stress feature parameters. Then, the deformation-stress feature parameters are used as dedicated training samples and input into the CNN deep learning network. The constraint boundary conditions are as follows: the heel reference area is set as a low-degree-of-freedom constraint area to restrict its abnormal drift outside the vertical direction; the toe front edge area is set as a variable contact response area to adapt to the local rotation and unfolding during the forefoot bearing phase; weak constraints or natural boundaries are applied to the non-force-bearing edges of the foot to avoid stress concentration distortion due to excessive fixation; for dynamic measurement scenarios, the boundary conditions are further updated in time according to the gait phase so that the heel strike phase, the full foot support phase, and the forefoot take-off phase correspond to different displacement load combinations and constraint states. The CNN deep learning network adopts a multi-layer residual structure design to specifically improve the model's generalization ability to nonlinear deformation data. Then, the AI ​​model is fully trained and optimized through large sample data to enable it to have the core computing ability to efficiently infer the pressure distribution from deformation data. The contact relationship is defined using a dual-interface contact method, specifically between the plantar tissue and the flexible bionic interface, and between the flexible bionic interface and the external support surface. The contact between the plantar tissue and the flexible bionic interface is set as surface-to-surface or node-to-surface contact to simulate the continuous adhesion behavior between the plantar surface and the bionic interface. The contact between the flexible bionic interface and the external support surface is set as a separable contact relationship to reflect changes in the contact area during plantar loading and unloading. The contact relationship combines normal hard contact and tangential frictional contact. The normal direction prohibits mutual penetration, while the tangential direction is assigned a friction coefficient range of 0.1-1.0 based on different test scenarios to simulate actual friction characteristics under static standing, slow walking, or functional evaluation states. The contact relationship introduces an adaptive update mechanism for the contact state. Based on the local compression of the flexible bionic interface, the nodal contact pressure, and the rate of change of the contact area, the three contact states of adhesion, slippage, and separation are dynamically identified and switched. When the contact pressure in a local area is lower than a preset threshold, it is determined to be in a separation state. When the tangential relative displacement exceeds the slippage threshold, it is determined to be in a slippage state, thereby improving the accuracy of the characterization of contact behavior in the arch suspension area, toe edge area, and gait rolling transition area. In the actual test, the pre-processed flexible biomimetic interface deformation data is input into the trained AI model. The AI ​​model closely combines the calibrated hyperelastic material model and the physical constraints of finite element analysis to achieve rapid inversion and near real-time prediction from silicone deformation data to plantar pressure distribution. The finite element analysis module employs a partitioned and refined mesh and a local iterative solution strategy for the contact area during the solution process. A higher density mesh is used in high load concentration areas, including the heel, first metatarsal head, fifth metatarsal head, and toe, while a relatively sparse mesh is used in the arch and low contact sensitivity areas. The local contact boundary is incrementally updated in conjunction with the contact nonlinear convergence criterion to balance solution accuracy and computational efficiency. To improve the consistency between finite element results and neural network training samples, the deformation-stress characteristic parameters output by the finite element analysis include at least nodal displacement, principal strain, equivalent stress, contact pressure, contact area ratio, and contact state label. Based on the changes in boundary conditions and contact relationships, a multi-condition sample set is generated so that the CNN deep learning network can learn the pressure response laws under different foot types, postures, and load conditions. This algorithm, which integrates finite element analysis with CNN deep learning network, can significantly reduce the overall computational load compared to the traditional pure finite element calculation method, successfully achieve near real-time prediction of pressure distribution, and effectively retain the physical constraint characteristics of finite element analysis, ensuring that the final output pressure distribution prediction results are both reliable and interpretable.

[0053] Furthermore, the fusion of the AI ​​model and finite element analysis employs a physical constraint fusion logic. During model training and prediction, the mechanical conservation relations, material constitutive relations, boundary condition constraints, and contact relationship constraints obtained from finite element analysis are introduced as joint constraints into the CNN deep learning network. This ensures that the plantar pressure distribution results output by the AI ​​model simultaneously meet the requirements of data-driven prediction capability and mechanical-physical consistency. Specifically: The nodal displacement, strain tensor, stress tensor, contact pressure, contact area ratio, and contact state label output by the finite element analysis module are constructed as physical feature priors and input together with the preprocessed deformation data into the CNN deep learning network to form a dual-channel fusion input structure of data features and physical features, so as to enhance the network's ability to identify local highly nonlinear deformation regions and contact transition regions. During network training, a joint loss function is constructed, which includes a data error loss term and a physical constraint loss term. The data error loss term is used to characterize the deviation between the predicted pressure distribution and the labeled pressure distribution, while the physical constraint loss term is used to constrain the prediction results to meet the hyperelastic response law of the material, the continuity of the contact boundary, the overall force balance relationship, and the local deformation coordination relationship. The weights of each loss term in the joint loss function can be adaptively adjusted according to the training stage, sample working conditions, or prediction error level. The physical constraint loss term includes at least one of the following: the mapping residual loss between the pressure field and the deformation field, the pressure mutation penalty loss in the boundary region, the spurious pressure suppression loss in the non-contact region, the consistency loss between the total reaction force and the body load, and the consistency loss between the pressure integral in the contact region and the interface deformation response, thereby suppressing local distortion, boundary oscillation, or non-physical anomaly predictions caused by relying solely on data fitting. During the model prediction phase, after the CNN deep learning network outputs the initial pressure distribution results, the finite element analysis module performs a rapid physical consistency check on the initial results to determine whether they meet the preset mechanical rationality threshold. When there is abnormal pressure concentration in local areas, contact boundary crossing, false pressure in non-contact areas, or overall reaction force deviation exceeding the limit, the physical constraint correction unit is called to iteratively correct the prediction results and output the final plantar pressure distribution results that meet the physical consistency requirements. The physical constraint correction unit adopts a closed-loop correction mechanism based on finite element residual feedback, which feeds back the residual field, boundary deviation field or contact state deviation obtained from finite element verification to the intermediate layer or output layer of the CNN deep learning network, and performs one or more local updates on the network prediction results, so as to improve the prediction stability and interpretability under complex working conditions while ensuring near real-time performance. For different subjects' foot types, different chronophases, and different load levels, the physical constraint fusion logic adopts a case-specific constraint strategy. For the standing case, it focuses on strengthening the overall force balance constraint; for the gait rolling case, it focuses on strengthening the contact state transfer constraint; and for the forefoot pushing case, it focuses on strengthening the continuity constraint of the local high-pressure area. This improves the generalization ability and prediction reliability of the AI ​​model in multi-scenario applications.

[0054] The core design logic of step S5 above is to overcome the industry pain point of difficult modeling of the hyperelastic nonlinearity of silicone through a two-layer architecture of "precise calibration of material model + deep integration of finite element and AI", and achieve "precise inversion + near real-time output" of pressure distribution. This ensures both the physical reliability of the results and the efficiency of engineering applications. It is the core transformation link from "deformation data" to "pressure data" in the entire measurement method, and directly determines the accuracy and practicality of "shape-pressure matching".

[0055] From a fundamental perspective, the calibration of the material model is a prerequisite for accurate inversion. The gradient-hardness speckled silicone used in the flexible biomimetic interface is a typical hyperelastic nonlinear material. The relationship between its deformation and stress cannot be described by traditional linear models. If the material model deviates from the actual characteristics, subsequent pressure inversion will produce systematic errors. Therefore, step S5 designed a multi-dimensional mechanical experiment of "uniaxial tension + biaxial tension + plane shear" to comprehensively capture the mechanical response of silicone under different stress states. Uniaxial tension reflects the unidirectional force characteristics, biaxial tension simulates the multi-directional force scenario of the sole of the foot, and plane shear corresponds to the deformation characteristics during sliding friction of the sole of the foot. The combination of the three can obtain complete mechanical parameters. The Ogden model or Mooney-Rivlin model was chosen for parameter fitting because these two types of models are classic models for characterizing hyperelastic materials and can accurately adapt to the nonlinear stress-strain relationship of silicone. The model calibrated by experimental data can ensure that the calculation of silicone deformation during finite element analysis conforms to the real physical laws, providing a solid material mechanics foundation for subsequent inversion.

[0056] In terms of computational architecture design, the core innovation lies in the fusion strategy of "finite element analysis + CNN deep learning," which perfectly compensates for the shortcomings of a single algorithm. The core value of finite element analysis lies in providing "physical constraints." Based on the principles of continuum mechanics, it strictly follows physical laws such as energy conservation and stress balance, avoiding "physical contradictions" (such as local pressure distribution not conforming to mechanical logic) that may occur when AI models perform pure data fitting. This is a key guarantee for the "reliability and interpretability" of pressure prediction results. The advantage of CNN deep learning lies in its "efficient computation and nonlinear fitting ability." For the complex mapping relationship of hyperelastic nonlinear deformation of silicone, traditional pure finite element calculation requires dividing a large number of meshes and iterative solving, resulting in a huge amount of computation that cannot meet the needs of real-time detection. However, CNN networks can quickly learn the correlation between deformation and pressure through data-driven learning. Crucially, the CNN network adopts a "multi-layer residual structure" design. This structure is optimized for feature extraction of nonlinear data, which can effectively alleviate the gradient vanishing problem in deep network training and improve the model's generalization ability for complex foot deformations (such as arch collapse and multi-point force on the forefoot), ensuring the accuracy of pressure inversion under different foot types and loads.

[0057] Furthermore, the "customized design" of the training samples further enhances the accuracy of the fusion algorithm: the "deformation-stress features" output by the finite element module are used as dedicated training samples for the AI ​​model, rather than directly using the original deformation data. This is equivalent to allowing the AI ​​model to incorporate physical and mechanical knowledge in advance during the learning process, establishing a "physically meaningful mapping relationship" rather than an indiscriminate data association. This not only improves the model training efficiency but also makes the final pressure prediction results interpretable. Each pressure distribution feature corresponds to a specific deformation-stress mechanism, rather than a simple numerical output.

[0058] In the real-world testing phase, preprocessed deformation data is input into the trained AI model. The model simultaneously utilizes a calibrated hyperelastic material model and finite element physical constraints, achieving dual control through "data-driven + physical constraints": the AI ​​model rapidly completes nonlinear mapping to achieve near real-time inversion of pressure distribution, while the physical rules of the finite element method correct for potential deviations in AI predictions, ultimately achieving a balance between "low computational load" and "high accuracy." Compared to traditional pure finite element calculations, this significantly reduces the time spent on mesh generation and iterative solutions, meeting the real-time requirements of clinical testing and rapid brace design; simultaneously, it retains the constraints of mechanical principles, ensuring that the pressure distribution results conform to the biomechanical laws of the human foot, avoiding the unreliability of "black box predictions" from pure AI models.

[0059] In summary, step S5, through a three-layer design of "multi-dimensional experimental calibration of the material model (ensuring accuracy) + finite element-AI fusion architecture (ensuring reliability and improving efficiency) + dedicated samples and residual networks (strong adaptability)," successfully overcomes the challenge of pressure inversion in the nonlinear deformation of hyperelastic materials. It efficiently and accurately transforms the deformation data of the flexible biomimetic interface into plantar pressure distribution data, providing high-quality pressure data support for the "shape-pressure coupling" in step S6. This is the core technology for achieving the core objective of "shape-pressure matching."

[0060] S6: The fusion module spatially matches and couples the 3D morphological data with the inverted pressure distribution data to form morphological-pressure matching data, and outputs relevant application results for precise adaptation.

[0061] In this embodiment, step S6 specifically includes: After receiving the three-dimensional morphological standardized data and pressure distribution prediction data, the fusion module performs precise spatial matching and deep coupling of the two types of data based on the spatial coordinate matching algorithm to form complete shape-pressure matching data. This data can comprehensively reflect the intrinsic relationship between foot morphology and pressure distribution under real physiological weight-bearing conditions. The system outputs application results including three-dimensional morphology-pressure coupling data, personalized brace design schemes, and plantar stress analysis reports; The shape-compression matching data and application results can be directly used for the precise design and rapid adaptation of foot braces, rehabilitation aids and sports protection products, providing scientific data support for clinical diagnosis and treatment, sports protection and health management; The personalized brace design scheme and insole design scheme are automatically output by the parameter generation module. The parameter generation module extracts at least one of the following parameters based on the shape-pressure matching data: foot length, foot width, arch height, inner and outer longitudinal arch curvature, transverse arch width, heel containment boundary, forefoot spread width, pressure peak value of the first metatarsal head region, pressure peak value of the fifth metatarsal head region, pressure peak value of the heel region, pressure center trajectory, and high pressure area distribution range, so as to automatically generate geometric parameters and functional parameters suitable for brace or insole structure design. The geometric parameters include at least the insole length, insole width, forefoot thickness, arch support height, heel cup depth, medial support extension length, lateral limiting width, metatarsal pad placement position, metatarsal pad length, metatarsal pad width, metatarsal pad height, and the contour boundary parameters of the pressure relief groove; the functional parameters include at least the target hardness, elastic gradient, local cushioning coefficient, support stiffness distribution, and zoned material configuration parameters of the corresponding area. The parameter generation module determines the foot shape constraint boundary based on three-dimensional morphological data, and determines the functional zoning boundaries of the pressure relief zone, support zone, buffer zone and stability zone based on pressure distribution data. The pressure relief zone corresponds to the plantar area where pressure is concentrated and tissues are sensitive, the support zone corresponds to the arch of the foot and the area related to posture stability, the buffer zone corresponds to the heel and forefoot impact absorption area, and the stability zone corresponds to the outer edge of the foot or the area related to posture control, so as to form a zoning parameter model that takes into account fit, pressure relief and stability. The parameter generation module automatically determines the parameters of the brace or insole by combining a rule base with an optimization algorithm. The rule base includes at least foot type classification rules, arch support rules, forefoot decompression rules, heel cushioning rules, and pronation / inversion correction rules. The optimization algorithm iteratively optimizes the generated parameters based on the target of reducing peak plantar pressure, the target of pressure uniformity, material volume constraints, and wearing comfort constraints, so as to output the design result that meets the preset target. The foot type classification rules include at least the classification rules for normal feet, low arch feet, high arch feet, and forefoot eccentric feet, and automatically call different support templates and initial parameter values ​​according to different foot types; for low arch feet, the arch support height and medial support range are increased; for high arch feet, the arch cushioning transition area is increased and local stress concentration is reduced; for high forefoot high-pressure feet, the metatarsal pad parameter weight is automatically increased and the forefoot cushioning thickness is adjusted. The parameter generation module can also automatically generate partition gradient structure parameters based on the shape and pressure matching data, so that the brace or insole forms a continuously changing thickness gradient, hardness gradient or elastic gradient along the heel area, arch area, metatarsal area and toe area, so as to reduce the mechanical abrupt change between adjacent areas and improve the continuity of force and comfort during wearing. To improve the accuracy of individualized adaptation, the parameter generation module further establishes a reverse mapping relationship between the pressure target distribution and structural parameters. By setting target pressure reduction values ​​for the high-pressure area of ​​the sole, target support values ​​for the arch area, and target limit values ​​for the unstable posture area, the corresponding combination of support or insole structural parameters is automatically deduced, thereby realizing the automatic conversion from measurement results to design parameters. The automatically generated brace or insole parameters can be converted into manufacturable data, which includes at least three-dimensional surface contour parameters, partition thickness parameters, partition hardness parameters, material stacking sequence parameters, and processing boundary parameters for use in CNC machining, additive manufacturing, or mold forming. The system automatically adjusts the minimum thickness, rounded corner transitions, and partition splicing boundaries according to different manufacturing methods, so that the output results simultaneously meet the mechanical design requirements and manufacturing process requirements. After the personalized brace design scheme and insole design scheme are generated, the system also performs virtual verification on the generated scheme based on the shape-pressure matching data. The virtual verification includes at least the peak pressure improvement rate assessment, pressure distribution uniformity assessment, pressure center trajectory deviation assessment, and local over-support risk assessment. When the verification result does not reach the preset threshold, the parameter generation module automatically adjusts the arch support height, metatarsal pad parameters, heel cup depth, or zone hardness parameters, and re-outputs the optimized brace or insole design result.

[0062] The core design logic of step S6 above is to achieve deep data coupling through precise spatial matching, and to realize the core value of 'form-pressure matching' through multi-dimensional application results. It is the final implementation link of the entire measurement method. It not only inherits the standardized morphological data of step S4 and the precise pressure data of step S5, but also forms core data assets through coupling, and finally outputs results that meet the needs of actual applications. It completely solves the pain points of existing technologies such as "lack of form-pressure coupling data and insufficient fitting accuracy", and realizes a closed loop from "data acquisition" to "precise application".

[0063] From the core design perspective of data coupling, the "spatial coordinate matching algorithm" employed in the fusion module is key to achieving deep coupling between form and pressure. Three-dimensional morphological data records the spatial structural information of the sole and instep (such as arch height, forefoot width, and heel shape), while pressure distribution data reflects the stress intensity and distribution patterns in different areas of the sole (such as peak heel pressure and weak pressure areas in the arch). Although these two types of data originate from synchronous acquisition, they essentially belong to two different dimensions: "spatial structure" and "mechanical response." A lack of precise spatial correlation can lead to "morphological and pressure misalignment" (such as a morphological feature corresponding to incorrect pressure data). The spatial coordinate matching algorithm, by unifying the spatial coordinate systems of the two types of data, precisely maps each pressure distribution feature to its corresponding spatial location on the sole, achieving deep "point-to-point" coupling. For example, the arc shape on the inner side of the arch of the foot is precisely associated with its corresponding pressure value and distribution range. The resulting "shape-pressure matching data" is no longer a simple superposition of two types of data, but can fully reveal the intrinsic relationship of "why a certain shape feature of the sole corresponds to a specific pressure distribution". This perfectly implements the core innovative mechanism of "shape-pressure matching" in this application and fills the gap in existing technologies that cannot establish the coupling relationship between shape and pressure.

[0064] At the application output level, the design logic revolves around "adapting to diverse needs": the output "three-dimensional morphology-pressure coupling data" is the core data asset, providing fundamental materials for in-depth analysis in professional fields (such as sports biomechanics research and clinical diagnosis and treatment of foot diseases); the "personalized brace design solution" directly addresses industry application needs, solving the shortcomings of traditional brace design that relies solely on morphological data and ignores pressure adaptation. Based on morphological-pressure matching data, designers can specifically optimize the support points of the brace (such as strengthening support in the weak pressure area of ​​the arch) and decompression areas (such as using flexible materials in the peak pressure area of ​​the forefoot), achieving dual precision of "morphological fit + pressure adaptation"; the "foot stress analysis report" takes into account the needs of clinical assessment and personal health management, providing a scientific basis for disease diagnosis and rehabilitation training plan development by intuitively presenting abnormal stress areas of the foot (such as overload on the outer side of the heel in patients with flat feet).

[0065] Most importantly, step S6 clarifies that the application scenarios of the output results cover "foot braces, rehabilitation aids, and sports protection products," and serve three major areas: "clinical diagnosis and treatment, sports protection, and health management." This design directly echoes the technical effect of this application's "providing a commercially viable multi-dimensional interactive testing solution": the standardized characteristics of form-compression matching data (such as the JSON format storage mentioned above) allow it to be directly called by various brace design software, significantly shortening the product development and customization cycle and improving adaptation efficiency; while the accuracy of the data ensures the effectiveness of the relevant products, avoiding problems such as poor rehabilitation effects and sports protection failure caused by inaccurate adaptation in existing technologies, laying a practical foundation for the commercial promotion of the technology.

[0066] In summary, step S6, through the design of "spatial coordinate matching algorithm to ensure coupling accuracy + multi-dimensional output to adapt to diverse needs," successfully transforms the high-quality data accumulated in each of the front-end steps into results with practical application value. This not only realizes the implementation of the core mechanism of "form-pressure matching" but also completes the closed loop from technology research and development to industrial application and clinical services, ultimately achieving the invention goal of "precise adaptation and scientific support." It is the key concluding link connecting technological innovation and practical value in the entire plantar stress measurement method.

[0067] In addition, steps S5 and S6 also include system operation assurance and scenario adaptation processes, specifically: The system has a built-in anomaly handling mechanism that monitors in real time for anomalies, including occlusion, deformation oversaturation, or data synchronization delay, during scanning and data processing. When an anomaly is detected, the system automatically starts an error correction program to correct it. If the anomaly cannot be eliminated through correction, the operator is prompted to recollect the data. For different application scenarios, including rehabilitation assistance, sports biomechanical analysis, and personalized insole customization, the system presets corresponding parameter templates and implements adaptive data processing strategies by switching templates to ensure the accuracy and efficiency of data processing in different scenarios. The inversion calculation stage is completed collaboratively by the finite element and AI modules. The pressure distribution data and coupled comprehensive data obtained from the inversion calculation are stored in a standardized JSON format. This format facilitates direct access by external software such as brace design software and rehabilitation assessment systems, thereby improving the convenience of data application.

[0068] The system operation assurance and scenario adaptation process supplemented in steps S5 to S6 above is designed with the core logic of "abnormal fallback + scenario adaptation + format compatibility" to ensure that the entire plantar stress measurement system has both stable and reliable operation capabilities and can flexibly adapt to diverse application needs. At the same time, it lowers the threshold for data application, provides key support for the commercial promotion and cross-scenario implementation of the technology, and solves the pain points of traditional detection systems such as "weak anti-interference ability, poor scenario compatibility, and difficulty in data reuse".

[0069] From the perspective of system operational stability, the built-in anomaly handling mechanism is a key design feature for dealing with the complex variables in real-world testing scenarios. In actual testing, anomalies such as subject posture fluctuations (e.g., overlapping toes, feet obstructing the scanning lens), individual load differences (e.g., overweight subjects causing excessive saturation of the silicone interface deformation, exceeding the preset measurement range), and device transmission delays (e.g., lag in synchronization between 3D point cloud and deformation data during high-speed acquisition) are difficult to avoid. Without an effective handling mechanism, these anomalies will directly lead to data distortion, analysis interruption, or unreliable results. This invention achieves full-process control of anomalies through a closed-loop logic of "real-time monitoring - automatic correction - prompt for re-acquisition": First, it captures specific anomaly types such as occlusion, deformation oversaturation, and data synchronization delay in real time, and then initiates targeted error correction procedures (such as supplementing data in occluded areas through algorithms, adjusting the deformation threshold range, and resynchronizing timestamps) to salvage valid data to the greatest extent possible; for serious anomalies that cannot be eliminated by correction (such as large-area occlusion leading to missing core data or deformation exceeding the material's elastic limit), it promptly prompts operators to re-acquire data to prevent invalid data from flowing into subsequent stages, thus ensuring the accuracy of pressure inversion and deformation-pressure coupling from the source, and enabling the system to output reliable results stably even in non-ideal testing environments.

[0070] In terms of scenario adaptability, the adaptive strategy of the preset parameter templates accurately matches the core needs of different users. Rehabilitation assistance scenarios require a focus on the pressure peaks and distribution patterns of lesion areas on the sole of the foot (such as plantar fasciitis pain points), demanding extremely high data accuracy and medical relevance. Sports biomechanical analysis scenarios need to capture instantaneous deformation and pressure changes during dynamic walking and running, requiring high data acquisition frame rates and dynamic response speeds. Personalized insole customization scenarios need to balance morphological details and pressure adaptation, emphasizing the direct guidance of data on insole structural design. Using a unified data processing strategy would inevitably fail to meet the core needs of all scenarios. This invention, by pre-setting exclusive parameter templates for different scenarios (such as enhancing the pressure peak recognition algorithm in rehabilitation scenarios, improving the data sampling frame rate in sports scenarios, and optimizing the morphology-pressure mapping accuracy in insole customization scenarios), allows users to achieve adaptive data processing simply by switching templates, without manually adjusting complex parameters. This lowers the operational threshold while ensuring data processing accuracy and efficiency in different scenarios, breaking through the shortcomings of traditional detection systems that are "single-function and scenario-limited," and laying the foundation for the technology's promotion in diverse fields such as clinical, sports, and consumer-grade customization.

[0071] The standardized data format further enhances the practical value of the technology and its adaptability to the industry chain. The inverted pressure distribution data and the coupled shape-compression matching data are stored in a standardized JSON format. The core advantage lies in its "universality and convenience": JSON is a universally accepted data exchange format in the industry, compatible with most external software such as orthosis design software, rehabilitation assessment systems, and biomechanical analysis tools. No additional format conversion or data reconstruction is required; it can be directly used by downstream applications. This design connects the "data acquisition-analysis-application" links in the industry chain. For example, insole customization manufacturers can directly import the JSON-formatted shape-compression coupling data into design software to quickly generate personalized products that fit the user's foot shape and stress requirements. Hospital rehabilitation departments can connect the data to rehabilitation assessment systems, providing quantitative evidence for the diagnosis and treatment of foot diseases. This significantly reduces the adaptation cost of technology implementation, enhances the reusability of data, and directly echoes the technical effect of "providing a commercially viable multi-dimensional interactive detection solution."

[0072] In summary, the system's operation and scenario adaptation process ensures system stability through anomaly handling mechanisms, enhances scenario compatibility through parameter template switching, and strengthens data usability through JSON standardization. The synergistic effect of these three aspects makes the entire plantar stress measurement method not only technologically innovative but also reliable, convenient, and scalable in practical applications. It serves as an important bridge connecting technology research and development with commercialization, ensuring that the core value of "shape-pressure matching" can be efficiently implemented in different scenarios.

[0073] Second Embodiment like Figure 2 As shown, this embodiment provides a finite element analysis-based plantar stress measurement system for performing the finite element analysis-based plantar stress measurement method as described in the first embodiment, comprising: A flexible biomimetic interface fabrication module is used to prepare and provide a flexible biomimetic interface. The flexible biomimetic interface is made of gradient hardness speckled silicone material with silicone rubber matrix. It has superelasticity and nonlinear stress-strain characteristics and is used to simulate the physiological interaction environment of the sole of the foot to capture the deformation of the sole of the foot under real weight-bearing conditions. The subject interface deformation generation module is used when the subject stands or walks on the flexible bionic interface, so that the sole of the foot is in full contact with the flexible bionic interface, and generates a realistic deformation that conforms to the physiological state based on the gradient hardness characteristics of the interface. The three-dimensional morphology and deformation acquisition module is used to start the multi-camera laser / structured light scanning system to simultaneously acquire the three-dimensional morphological data of the sole and instep and the real-time deformation data of the flexible bionic interface. The scanning system is double-calibrated and corrected by an adaptive algorithm to ensure data accuracy. The data standardization preprocessing module is used to preprocess the acquired three-dimensional morphological data and deformation data through a modular bus architecture system to obtain standardized analysis data. The pressure distribution inversion and prediction module is used to perform pressure distribution inversion and near real-time prediction on preprocessed deformation data based on a material model calibrated through mechanical experiments and an algorithm that combines finite element analysis and CNN deep learning network. The form-pressure coupling result output module is used to spatially match and couple the three-dimensional morphological data with the inverted pressure distribution data through the fusion module to form form-pressure matching data and output relevant application results for accurate adaptation.

[0074] A computer-readable storage medium stores computer code that, when executed, performs the methods described above. Those skilled in the art will understand that all or part of the steps in the various methods of the above embodiments can be implemented by a program instructing related hardware. This program can be stored in a computer-readable storage medium, which may include: read-only memory (ROM), random access memory (RAM), a magnetic disk, or an optical disk, etc.

[0075] The above description is merely a preferred embodiment of the present invention. The scope of protection of the present invention is not limited to the above embodiments. All technical solutions falling within the scope of the present invention's concept are within the scope of protection of the present invention. It should be noted that for those skilled in the art, any improvements and modifications made without departing from the principles of the present invention should also be considered within the scope of protection of the present invention.

[0076] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0077] It should be noted that the above embodiments can be freely combined as needed. The above description is only a preferred embodiment of the present invention. It should be pointed out that for those skilled in the art, several improvements and modifications can be made without departing from the principle of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A plantar stress measurement method based on finite element analysis, characterized by, Includes the following steps: S1: Prepare and provide a flexible biomimetic interface, which is made of a gradient hardness speckled silicone material with a silicone rubber matrix. It has superelasticity and nonlinear stress-strain characteristics and is used to simulate the physiological interaction environment of the sole of the foot to capture the deformation of the sole of the foot under real weight-bearing conditions. S2: The subject stands or walks on the flexible bionic interface, so that the sole of the foot is in full contact with the flexible bionic interface, and produces a real deformation that conforms to the physiological state based on the gradient hardness characteristics of the interface. S3: Activate the multi-camera laser / structured light scanning system to simultaneously acquire three-dimensional morphological data of the sole and instep, as well as real-time deformation data of the flexible bionic interface. The scanning system undergoes dual calibration and adaptive algorithm correction to ensure data accuracy. S4: The system uses a modular bus architecture to preprocess the acquired 3D morphological and deformation data to obtain standardized analysis data. S5: Based on a material model calibrated in advance through mechanical experiments, an algorithm combining finite element analysis and CNN deep learning network is used to perform pressure distribution inversion and near real-time prediction on the preprocessed deformation data. S6: The fusion module spatially matches and couples the 3D morphological data with the inverted pressure distribution data to form morphological-pressure matching data, and outputs relevant application results for precise adaptation.

2. The plantar stress measurement method based on finite element analysis according to claim 1, characterized by, Step S1 is as follows: Silicone rubber was selected as the matrix material, and a continuous gradient hardness distribution was designed to match the stress adaptation requirements of different areas of the sole. Scattered particles were uniformly introduced into the silicone rubber matrix material, and the deformation was visualized through the distribution of the scattered particles, providing a basis for subsequent digital image correlation analysis. The flexible biomimetic interface is divided into a surface deformation acquisition layer, an intermediate gradient transition layer, and a bottom load-bearing support layer along the thickness direction; the continuous gradient hardness distribution is achieved by at least one of the following methods: controlling the crosslinking agent ratio, filler concentration, curing temperature field, or layered casting sequence at different thickness positions of the silicone rubber matrix, so that the flexible biomimetic interface forms a monotonically varying elastic modulus distribution in the thickness direction. The flexible biomimetic interface material exhibits significant hyperelasticity and nonlinear stress-strain characteristics. Its stress-strain relationship is comprehensively calibrated through uniaxial tensile tests, biaxial tensile tests, and plane shear tests to ensure the accuracy of the material's mechanical property data. The flexible biomimetic interface was prepared using an integrated process of vacuum mixing, layered curing, and surface treatment. After preparation, the hardness gradient distribution and speckle particle distribution of the material were experimentally verified to ensure that they met the design requirements. The speckle particles were high-contrast inert particles or pigment microparticles that were compatible with the silicone rubber matrix. The speckle particles were arranged along the thickness direction of the flexible biomimetic interface in a manner with high density on the surface, transitional distribution in the middle layer, and low density or no speckle distribution at the bottom layer. By using mechanical performance testing and high-resolution imaging technology, the nonlinear strain response of the flexible bionic interface is systematically evaluated to ensure that the interface maintains good repeatability and stability within different stress load ranges on the sole of the foot, while always maintaining the bionic contact performance of the sole of the foot under different loads.

3. The plantar stress measurement method based on finite element analysis according to claim 1, characterized by, Step S3 is as follows: The multi-camera laser / structured light scanning system has a detection range that fully covers the sole and instep areas. The system is equipped with a laser line source with a wavelength range of 450–850nm. The multi-camera laser / structured light scanning system includes 4 to 12 camera acquisition units, and each camera acquisition unit is distributed circumferentially relative to the foot center reference coordinate system. Before starting data acquisition, the scanning system is subjected to both geometric and optical calibration using standard target points. This dual calibration ensures that the coordinate systems of multiple cameras maintain strict consistency, thus avoiding data acquisition errors caused by coordinate system deviations. To address the differences in foot shape among different subjects and the possible posture changes that subjects may experience during the acquisition process, the scanning system incorporates an adaptive stitching algorithm. This algorithm can automatically identify image distortion problems caused by posture changes and occlusion of the sole and instep areas, and perform corresponding error corrections in real time, effectively avoiding stitching errors that are easily generated by traditional single-camera systems. Through the synergistic effect of system configuration, dual calibration, and adaptive algorithm correction, high-precision, continuous, and complete three-dimensional morphological data of the sole and instep are finally acquired. At the same time, real-time deformation data of the flexible bionic interface is captured simultaneously, avoiding the stitching errors of traditional single-camera systems and providing accurate and reliable geometric input data for subsequent mechanical inversion calculations.

4. The plantar stress measurement method based on finite element analysis according to claim 1, characterized by, Step S4 is as follows: The system corresponding to the measurement method adopts a modular bus architecture, and each sub-module, including the data acquisition module, preprocessing module, and analysis module, communicates in real time through TCP / IP or USB 3.0 high-speed data channels. The data acquisition module synchronously records the three-dimensional point cloud data of the sole and instep and the deformation field data of the flexible bionic interface, and then transmits the acquired raw data to the preprocessing module. The preprocessing module sequentially performs denoising, gridding, coordinate correction, and time synchronization on the raw data to remove interference information, standardize the data format and time reference, and form standardized analysis data, providing high-quality data support for subsequent stress inversion calculations.

5. The plantar stress measurement method based on finite element analysis according to claim 1, characterized by, Step S5 is as follows: With the establishment of the mapping relationship between deformation, stress, and pressure as the core, the mechanical parameters of the flexible biomimetic interface are comprehensively collected through uniaxial tensile tests, biaxial tensile tests, and plane shear tests. These mechanical parameters are used as core basic data to fit the relevant parameters of the Ogden model or Mooney-Rivlin model, thereby determining the characterization model that best matches the real mechanical properties of silicone, and ensuring the hyperelasticity accuracy in the subsequent finite element analysis process. A fusion computing architecture that deeply integrates finite element analysis and artificial intelligence deep learning algorithms is constructed. The finite element analysis module first performs professional calculations on the deformation data and outputs accurate deformation-stress feature parameters. Then, the deformation-stress feature parameters are used as dedicated training samples and input into the CNN deep learning network. The CNN deep learning network adopts a multi-layer residual structure design to specifically improve the model's generalization ability to nonlinear deformation data. Then, the AI ​​model is fully trained and optimized through large sample data to enable it to have the core computing ability to efficiently infer the pressure distribution from deformation data. In the actual test, the pre-processed flexible biomimetic interface deformation data is input into the trained AI model. The AI ​​model closely combines the calibrated hyperelastic material model and the physical constraints of finite element analysis to achieve rapid inversion and near real-time prediction from silicone deformation data to plantar pressure distribution. This algorithm, which integrates finite element analysis with CNN deep learning network, can significantly reduce the overall computational load compared to the traditional pure finite element calculation method, successfully achieve near real-time prediction of pressure distribution, and effectively retain the physical constraint characteristics of finite element analysis, ensuring that the final output pressure distribution prediction results are both reliable and interpretable.

6. The method for measuring plantar stress based on finite element analysis according to claim 5, characterized in that, The fusion of the AI ​​model and finite element analysis employs a physical constraint fusion logic. During model training and prediction, the mechanical conservation relations, material constitutive relations, boundary condition constraints, and contact relationship constraints obtained from finite element analysis are introduced as joint constraints into the CNN deep learning network. This ensures that the plantar pressure distribution results output by the AI ​​model simultaneously meet the requirements of data-driven prediction capability and mechanical-physical consistency. Specifically: The nodal displacement, strain tensor, stress tensor, contact pressure, contact area ratio, and contact state label output by the finite element analysis module are constructed as physical feature priors and input together with the preprocessed deformation data into the CNN deep learning network to form a dual-channel fusion input structure of data features and physical features. During network training, a joint loss function is constructed, which includes a data error loss term and a physical constraint loss term. The data error loss term is used to characterize the deviation between the predicted pressure distribution and the labeled pressure distribution, while the physical constraint loss term is used to constrain the prediction results to meet the hyperelastic response law of the material, the continuity of the contact boundary, the overall force balance relationship, and the local deformation coordination relationship. The weights of each loss term in the joint loss function can be adaptively adjusted according to the training stage, sample working conditions, or prediction error level. During the model prediction phase, after the CNN deep learning network outputs the initial pressure distribution results, the finite element analysis module performs a rapid physical consistency check on the initial results to determine whether they meet the preset mechanical rationality threshold. When there is abnormal pressure concentration in local areas, contact boundary crossing, false pressure in non-contact areas, or overall reaction force deviation exceeding the limit, the physical constraint correction unit is called to iteratively correct the prediction results and output the final plantar pressure distribution results that meet the physical consistency requirements. The physical constraint correction unit adopts a closed-loop correction mechanism based on finite element residual feedback, which feeds back the residual field, boundary deviation field or contact state deviation obtained from finite element verification to the intermediate layer or output layer of the CNN deep learning network to perform one or more local updates on the network prediction results.

7. The method for measuring plantar stress based on finite element analysis according to claim 1, characterized in that, Steps S5 and S6 also include system operation assurance and scenario adaptation processes, specifically: The system has a built-in anomaly handling mechanism that monitors in real time for anomalies, including occlusion, deformation oversaturation, or data synchronization delay, during scanning and data processing. When an anomaly is detected, the system automatically starts an error correction program to correct it. If the anomaly cannot be eliminated through correction, the operator is prompted to recollect the data. For different application scenarios, including rehabilitation assistance, sports biomechanical analysis, and personalized insole customization, the system presets corresponding parameter templates and implements adaptive data processing strategies by switching templates to ensure the accuracy and efficiency of data processing in different scenarios. The inversion calculation stage is completed collaboratively by the finite element and AI modules. The pressure distribution data and coupled comprehensive data obtained from the inversion calculation are stored in a standardized JSON format. This format facilitates direct access by external software such as brace design software and rehabilitation assessment systems, thereby improving the convenience of data application.

8. The method for measuring plantar stress based on finite element analysis according to claim 1, characterized in that, Step S6 is as follows: After receiving the three-dimensional morphological standardized data and pressure distribution prediction data, the fusion module performs precise spatial matching and deep coupling of the two types of data based on the spatial coordinate matching algorithm to form complete shape-pressure matching data. This data can comprehensively reflect the intrinsic relationship between foot morphology and pressure distribution under real physiological weight-bearing conditions. The system outputs application results including three-dimensional morphology-pressure coupling data, personalized brace design schemes, and plantar stress analysis reports; The shape-compression matching data and application results can be directly used for the precise design and rapid adaptation of foot braces, rehabilitation aids and sports protection products, providing scientific data support for clinical diagnosis and treatment, sports protection and health management.

9. A finite element analysis-based plantar stress measurement system for performing the plantar stress measurement method based on finite element analysis as described in any one of claims 1-8, characterized in that, include: A flexible biomimetic interface fabrication module is used to prepare and provide a flexible biomimetic interface. The flexible biomimetic interface is made of gradient hardness speckled silicone material with silicone rubber matrix. It has superelasticity and nonlinear stress-strain characteristics and is used to simulate the physiological interaction environment of the sole of the foot to capture the deformation of the sole of the foot under real weight-bearing conditions. The subject interface deformation generation module is used when the subject stands or walks on the flexible bionic interface, so that the sole of the foot is in full contact with the flexible bionic interface, and generates a realistic deformation that conforms to the physiological state based on the gradient hardness characteristics of the interface. The three-dimensional morphology and deformation acquisition module is used to start the multi-camera laser / structured light scanning system to simultaneously acquire the three-dimensional morphological data of the sole and instep and the real-time deformation data of the flexible bionic interface. The scanning system is double-calibrated and corrected by an adaptive algorithm to ensure data accuracy. The data standardization preprocessing module is used to preprocess the acquired three-dimensional morphological data and deformation data through a modular bus architecture system to obtain standardized analysis data. The pressure distribution inversion and prediction module is used to perform pressure distribution inversion and near real-time prediction on preprocessed deformation data based on a material model calibrated through mechanical experiments and an algorithm that combines finite element analysis and CNN deep learning network. The form-pressure coupling result output module is used to spatially match and couple the three-dimensional morphological data with the inverted pressure distribution data through the fusion module to form form-pressure matching data and output relevant application results for accurate adaptation.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer code, and when the computer code is executed, the method as described in any one of claims 1 to 8 is performed.

Images