A quantitative reconstruction method and system for force chains based on shared contact variables and multi-constraint inversion
By using the method of shared contact variables and multi-constraint inversion, the problem of inconsistent contact topology in the quantitative analysis of contact force of photoelastic particles was solved, and stable reconstruction and reliable evaluation of the contact force network were achieved, thus improving the accuracy and consistency of the analysis results.
Patent Information
- Application Number
- CN202611622873.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-10-20
- Publication Date
- 2026-11-17
AI Technical Summary
In existing quantitative analysis of contact forces in photoelastic particles, candidate contact topologies are affected by image and geometric errors, and the independent inversion results of the same physical contact on adjacent particles are inconsistent. There is a lack of unified iterative verification of the overall mechanical state, which leads to unstable calculation results of contact force networks.
A shared contact variable and multi-constraint inversion method is adopted. By acquiring the geometric image and photoelastic image of the particle system, spatial registration is performed to construct candidate contact topologies. The mapping relationship between shared contact variables and particles on both sides of the contact is established. Multi-constraint objective functions are constructed by combining image residuals and mechanical equilibrium residuals, and nonlinear optimization is performed to update the candidate contact topologies until the convergence conditions are met.
It achieves a unified force relationship between adjacent particles under the same physical contact, reduces the omission of weak contact and the misjudgment of non-force transmission proximity relationship, improves the stability and mechanical consistency of the solution results under complex contact conditions, and provides reliable contact force reconstruction results and error evaluation.
Smart Images

Figure CN123120158A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of micromechanical testing of particulate materials and digital optical mechanical testing, specifically to a method and system for quantitative reconstruction of force chains based on shared contact variables and multi-constraint inversion, and more particularly to a method and system for quantitatively inverting the contact topology and contact forces of a particulate system using photoelastic images, and reconstructing the force chain network based on shared contact variables and multiple mechanical constraints. Background Technology
[0002] Particulate materials are widely found in soil and rock masses, mineral particles, powder materials, granular infill structures, and multiphase media. Their macroscopic load-bearing, deformation, and instability behaviors are closely related to the non-uniform transmission of contact forces between particles. Under external loads, the contact forces between particles are not uniformly distributed within the system, but rather form a force chain network with obvious directionality and non-uniformity through a finite number of strong contacts. Accurately obtaining the particle-scale contact forces and their spatial organization is a crucial foundation for analyzing stress transmission, local rearrangement, structural instability, and the correlation between macroscopic and microscopic responses in granular materials.
[0003] Photoelasticity utilizes the stress-optical effect generated in birefringent materials under stress to transform the internal stress state of particles into observable bright / dark or colored fringes, thus it is widely used for the microscopic stress study of two-dimensional granular systems. Existing photoelastic particle analysis typically includes steps such as particle location identification, neighbor particle judgment, stress contact screening, and contact force calculation. For quantitative analysis, the particle stress level can be estimated based on indicators such as the gradient of the photoelastic image, or the theoretical photoelastic response can be calculated based on a given contact force. Optimization can then be used to match the theoretical photoelastic response with the experimental photoelastic image to invert the magnitude and direction of the contact force.
[0004] With the development of digital image processing and nonlinear optimization methods, the inversion of particle contact force based on photoelastic images has been able to transform from stripe visualization to quantitative contact force calculation. However, when there are many particles, complex force-affected contact, dense photoelastic stripes, or changes in image quality, candidate contact identification errors, initial contact force errors, and local optimization errors are easily propagated to each other. For experiments with transparent boundaries, liquid media, or complex background light fields, the local photoelastic response near weak contact may also be affected by changes in background brightness, reflection, and refraction, further increasing the uncertainty of contact topology and contact force inversion.
[0005] Current quantitative analysis of contact force in photoelastic particles still suffers from the following shortcomings: (1) Candidate contact topology is sensitive to geometric detection accuracy and local photoelastic response. Weak contact is prone to missed or misjudged detection: Particle contact is usually determined by particle center distance, radius and preset tolerance to determine candidate neighbors, and then the brightness or gradient response near the contact is used to determine whether force is transmitted. When there is a deviation in particle edge recognition, there are small gaps between particles, the background brightness is uneven or the weak contact stripes are not obvious, a single geometric threshold or a single local optical threshold is difficult to reliably distinguish between real force transmission contact and non-force transmission proximity relationship; after the incorrect contact topology enters the subsequent inversion, it will cause the number of contact force variables and the loading direction to deviate.
[0006] (2) Independent inversion based on a single particle as the basic object can easily lead to inconsistent calculation results for the same physical contact on adjacent particles: the same particle contact belongs to both sides of the contact. When the two sides of the particles are image fitted separately, the same contact may yield two different sets of contact force results. Post-processing can reduce some differences by comparison, averaging or replacement, but the correction results may not be simultaneously fed back to the photoelastic image fitting and mechanical equilibrium process of the particles on both sides of the contact, which can easily lead to the problem that the image matching results and the action-reaction relationship between particles cannot be uniformly satisfied.
[0007] (3) The contact force inversion is highly dependent on the initial values and the fitting results of local images, and lacks a unified iterative verification mechanism between the contact topology, local images and the overall mechanical state: The contact force solution is a nonlinear inverse problem. When the number of contacts is large, the stripes are dense or the initial force distribution deviation is large, the optimization process may converge to different local solutions. If the solution quality is evaluated based solely on the residual of a single particle image, it is difficult to determine whether the contact force network is reasonable in terms of particle force and torque balance, boundary resultant force and whole-field photoelastic response. For particles with abnormal local residuals, there is also a lack of a unified process for targeted adjustment of the uncertain contact topology based on the source of residuals and re-solving. Summary of the Invention
[0008] The purpose of this invention is to address the problems in existing technologies for quantitative analysis of contact forces in photoelastic particles, such as the influence of image and geometric errors on candidate contact topology, the difficulty in maintaining consistency in independent inversion results of the same physical contact on adjacent particles, and the lack of unified iterative verification between local image fitting and overall mechanical state. This invention provides a method and system for quantitative reconstruction of force chains with shared contact variables and multi-constraint inversion.
[0009] To achieve the above objectives, the technical solution adopted by the present invention is as follows: Firstly, a quantitative reconstruction method for force chains based on shared contact variables and multi-constraint inversion is provided, comprising the following steps: Obtain geometric and photoelastic images of the same particle system under unpolarized and polarized light fields, and spatially register the geometric and photoelastic images to form a set of particle geometric information. The geometric proximity relationships between particles and between particles and boundaries are determined based on the set of particle geometric information, and candidate contact topologies are constructed by combining the local photoelastic response of the candidate contact regions. For the actual physical contact in the candidate contact topology, a shared contact variable is established, and a mapping relationship between the shared contact variable and the particles on both sides of the contact is established. The internal stress field of the particles is calculated based on the shared contact variable, and a theoretical photoelastic response corresponding to each particle is generated. Based on the difference between the photoelastic image and the theoretical photoelastic response, an image residual is constructed. The image residual is then combined with the particle force balance residual, particle torque balance residual, and boundary or overall load residual constructed based on the shared contact variable to form a multi-constraint objective function. Given a candidate contact topology and initial values of shared contact variables, the multi-constraint objective function is subjected to constrained nonlinear optimization to update the candidate contact topology until a preset convergence condition is met. When the preset convergence condition is met, the particle-particle contact force and particle-boundary contact force corresponding to each shared contact variable are output. A full-field synthetic photoelastic image and a quantitative force chain network with contact force magnitude and direction information are generated according to the actual spatial position of the particles. At the same time, an error evaluation index for characterizing the reliability of the reconstruction result is output.
[0010] Furthermore, candidate contacts between particles and between particles and boundaries are constructed based on the geometric proximity relationships between particles and between particles and boundaries. Candidate contacts between particles are determined based on the difference between the particle center distance and the sum of the corresponding particle sizes, while candidate contacts between particles and boundaries are determined based on the shortest distance between the outer edge of the particle and the test boundary. For candidate contacts that meet the geometric proximity conditions, photoelastic brightness, grayscale gradient, stripe directionality, and / or local texture response are extracted in a local region based on the direction of the line connecting the candidate contact point and the particle center. The effectiveness of the candidate contacts is evaluated based on the local photoelastic response.
[0011] Furthermore, the shared contact variables are represented by contact force amplitude and direction parameters, or by normal contact force components and tangential contact force components. For the same physical contact between particle i and particle j, only one set of shared contact variables is set, and they are applied to particle i and particle j respectively by direction transformation, so that the forces of the same physical contact on the two adjacent particles satisfy the relationship of equal magnitude and opposite direction.
[0012] Furthermore, the principal stress difference at each position inside the particle is calculated based on the shared contact variable, particle geometric parameters, and photoelastic material parameters, and the theoretical photoelastic response is generated according to the stress-optical relationship corresponding to the circularly polarized light field or the plane-polarized light field; the theoretical photoelastic response of each particle is mapped to the full-field coordinates according to the particle center position and size to form a full-field synthesized photoelastic image.
[0013] Furthermore, the multi-constraint objective function includes at least an image matching term and a particle mechanics equilibrium term; the image matching term is constructed based on the pixel differences and / or structural similarities between the photoelastic image and the theoretical photoelastic response, and the particle mechanics equilibrium term is constructed based on the vector sum and torque sum of the shared contact variables associated with each particle; when the external load or boundary reaction force is known, the multi-constraint objective function also includes an overall load constraint term consisting of the difference between the calculated boundary resultant force and the measured external load.
[0014] Furthermore, a physical value range is set for the shared contact variables and multiple sets of initial values are generated, and constrained nonlinear optimization is performed on each set. The candidate optimal results are selected based on the objective function value, particle image residual, and overall load error corresponding to the solution results of each set of initial values. When the image residual or mechanical equilibrium residual of the preset particle exceeds the threshold, the topology adjustment of the uncertain contact associated with the particle is prioritized, and the solution is re-solved under the adjusted candidate contact topology.
[0015] Furthermore, the particle system is a dry particle system, a saturated particle system, or a particle system subjected to pore fluid action; when the particle system is subjected to pore fluid action, the flow rate, pressure difference, hydraulic gradient, and / or pore velocity are recorded simultaneously, and the quantitative force chain network obtained under different hydraulic conditions is associated and stored according to the corresponding hydraulic state.
[0016] Secondly, a quantitative force chain reconstruction system based on shared contact variables and multi-constraint inversion is provided to realize the quantitative force chain reconstruction method based on shared contact variables and multi-constraint inversion as described above. The system includes: The image and geometric information acquisition unit is used to acquire geometric and photoelastic images of the same particle system, complete image registration, and identify the particle center position, size, number, and experimental boundary. Candidate contact topology building units are used to construct candidate contact topologies for particle-particle and particle-boundary relationships based on particle geometric proximity and local photoelastic response, and to identify determined and uncertain contacts. The shared contact variable and photoelastic forward modeling unit are used to establish a unique shared contact variable for each physical contact, map the shared contact variable to the particles on both sides of the contact, and calculate the internal stress field and theoretical photoelastic response of the particles based on the shared contact variable. The multi-constraint objective building unit is used to construct image residuals based on photoelastic images and theoretical photoelastic responses, and to construct particle force balance residuals, particle torque balance residuals, and boundary or overall load residuals based on shared contact variables, so as to form a multi-constraint objective function. The adaptive inversion and topology update unit is used to perform constrained nonlinear optimization and multiple initial value repeated solutions on the multi-constraint objective function, and adjust the activation state of uncertain contact according to the residual evaluation results after the solution until the preset convergence condition is met. The results are summarized and output by the force chain output unit, which is used to output particle-particle contact force, particle-boundary contact force, full-field synthesized photoelastic image, quantitative force chain network, and error evaluation index.
[0017] Thirdly, a computer device is provided, including a memory and a processor, wherein the memory stores a computer program, characterized in that the processor executes the computer program to implement the steps of the method for quantitative reconstruction of force chains based on shared contact variables and multi-constraint inversion.
[0018] A non-transitory computer-readable storage medium storing computer instructions thereon, characterized in that, when executed by a processor, the computer instructions cause the computer to execute the aforementioned quantitative reconstruction method of force chains based on shared contact variables and multi-constraint inversion.
[0019] Compared with existing technologies, the beneficial effects of this invention are: 1. This invention uses physical contact as the basic solution object, setting only one set of shared contact variables for each particle-particle or particle-boundary contact in the particle system, and simultaneously mapping these shared contact variables to the particles on both sides of the contact, so that the force relationship of the same physical contact on adjacent particles remains consistent at the variable level, avoiding the problem of repeated estimation and inconsistency caused by the independent inversion of the same contact by the particles on both sides; 2. This invention combines geometric proximity information with the local photoelastic response of the candidate contact region to construct the candidate contact topology, and divides the contact into deterministic contact and uncertain contact; during the contact force solution process, the uncertain contact is adjusted in a targeted manner according to the particle image residual and mechanical equilibrium residual, so that the contact topology can be mutually corrected with the contact force inversion result, reducing the impact of missed weak contact and misjudgment of non-force-transmitting proximity relationship on subsequent steps. 3. This invention simultaneously utilizes experimental photoelastic images, particle force and torque balance, and boundary or overall load information to construct a multi-constraint objective function, and employs a multi-initial-value approach for repeated solutions, so that contact force inversion no longer relies solely on single-particle local image matching; by using observation information at different scales to jointly constrain and share contact variables, the stability and mechanical consistency of the solution results under complex contact conditions can be improved; 4. This invention establishes a unified calculation process of "candidate contact topology construction - shared contact force inversion - residual evaluation - topology update - full-field result output", and after meeting the convergence conditions, simultaneously outputs particle contact force, full-field synthesized photoelastic image, quantitative force chain network, and error evaluation index, so that the contact force reconstruction results have a verifiable, comparable, and traceable quality evaluation basis, and can be further used for quantitative analysis of particle force chain evolution under different loading, hydraulic, or other working conditions. Attached Figure Description
[0020] Figure 1 This is a schematic diagram of the overall process of a quantitative reconstruction method for force chains based on shared contact variables and multi-constraint inversion according to the present invention. Figure 2 A schematic diagram of a photoelastic particle sample and an image after photoelastic particle annealing test provided in an embodiment of the present invention; Figure 3 This is a schematic diagram illustrating the relationship between the photoelastic response of a single particle under different known load conditions and the overall square gradient response of the particle and the calibration of the loading force, provided in an embodiment of the present invention. Figure 4 This is a schematic diagram of the method for determining particle geometric proximity and candidate contact geometric parameters provided in an embodiment of the present invention; Figure 5 This is a schematic diagram illustrating the automatic identification, filtering, and correction process of particle geometry information provided in an embodiment of the present invention. Figure 5 Image (a) is the original unpolarized image. Figure 5 (b) shows the automatically recognized circle. Figure 5 (c) represents the initial circle recognition after filtering. Figure 5 (d) shows the particle number and geometric model after the filter modification; Figure 6 A schematic diagram comparing the measured photoelastic response, independent inversion results, and shared contact force reconstruction results of particles on both sides of the same physical contact, as provided in an embodiment of the present invention. Figure 7 This is a schematic diagram of particle numbering, candidate contact relationships, and contact topology identification results provided in an embodiment of the present invention; Figure 8 This is a schematic diagram of the final particle contact force network and multi-section force transmission verification provided in an embodiment of the present invention; Figure 9 A schematic diagram comparing the experimental photoelastic image provided in this embodiment of the invention with the full-field synthesized photoelastic image obtained by inversion reconstruction based on the final contact force. Detailed Implementation
[0021] The technical solution of the present invention will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are merely some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.
[0022] Example 1: A quantitative reconstruction method for force chains based on shared contact variables and multi-constraint inversion is provided, comprising the following steps: Step 1: Obtain geometric and photoelastic images of the same particle system under unpolarized and polarized light fields, spatially register the geometric and photoelastic images, identify the center position, size, number and experimental boundary of the particles, and form a set of particle geometric information. Step 2: Determine the geometric proximity relationships between particles and between particles and boundaries based on the set of particle geometric information, and construct candidate contact topologies by combining the local photoelastic response of the candidate contact regions. The candidate contact topologies include definite contacts and indefinite contacts. Step 3: Establish shared contact variables for the actual physical contact in the candidate contact topology, and establish the mapping relationship between the shared contact variables and the particles on both sides of the contact. Calculate the internal stress field of each particle based on the shared contact variables associated with each particle, and generate the theoretical photoelastic response corresponding to each particle based on the stress-optical relationship. Step 4: Construct an image residual based on the difference between the photoelastic image and the theoretical photoelastic response. Combine the image residual with the particle force balance residual, particle torque balance residual, and boundary or overall load residual constructed based on the shared contact variables to form a multi-constraint objective function. Step 5: Given the candidate contact topology and the initial values of the shared contact variables, perform constrained nonlinear optimization on the multi-constraint objective function and solve it repeatedly using a multi-initial-value method. Based on the optimized image residual, mechanical equilibrium residual and the effectiveness evaluation results of the candidate contact, retain, deactivate or reactivate uncertain contacts. Repeat steps 3 to 5 when the candidate contact topology is updated until the preset convergence condition is met. Step 6: When the preset convergence condition is met, output the particle-particle contact force and particle-boundary contact force corresponding to each shared contact variable, and generate a full-field synthetic photoelastic image and a quantitative force chain network with contact force magnitude and direction information according to the actual spatial position of the particles. At the same time, output the error evaluation index used to characterize the reliability of the reconstruction result.
[0023] This method for quantitative reconstruction of force chains using shared contact variables and multi-constraint inversion takes physical contact as the basic solution object. For each particle-particle or particle-boundary contact in the particle system, only one set of shared contact variables is set, and these shared contact variables are simultaneously mapped to the particles on both sides of the contact. This ensures that the force relationship of the same physical contact on adjacent particles remains consistent at the variable level, avoiding the problem of repeated estimation and inconsistency caused by the independent inversion of the same contact by the particles on both sides.
[0024] This quantitative force chain reconstruction method establishes a unified calculation process of "candidate contact topology construction - shared contact force inversion - residual evaluation - topology update - full-field result output". After meeting the convergence conditions, it simultaneously outputs particle contact force, full-field synthesized photoelastic image, quantitative force chain network and error evaluation index, so that the contact force reconstruction results have a quality evaluation basis that can be checked, compared and traced, and can be further used for quantitative analysis of particle force chain evolution under different loading, hydraulic or other working conditions.
[0025] Further, in step 2, candidate contacts between particles and between particles and boundaries are constructed based on the geometric proximity relationships between particles and between particles and boundaries. The candidate contacts between particles are determined based on the difference between the particle center distance and the sum of the corresponding particle sizes, and the candidate contacts between particles and boundaries are determined based on the shortest distance between the outer edge of the particle and the test boundary. For candidate contacts that meet the geometric proximity conditions, photoelastic brightness, grayscale gradient, stripe directionality, and / or local texture response are extracted in a local area based on the direction of the line connecting the candidate contact point and the particle center, and the effectiveness of the candidate contacts is evaluated based on the local photoelastic response.
[0026] Furthermore, in step 3, the shared contact variables are represented by contact force amplitude and direction parameters, or by normal contact force components and tangential contact force components; for the same physical contact between particle i and particle j, only one set of shared contact variables is set, and they are applied to particle i and particle j respectively by direction transformation, so that the forces of the same physical contact on the two adjacent particles satisfy the relationship of equal magnitude and opposite direction.
[0027] Furthermore, the principal stress difference at each position inside the particle is calculated based on the shared contact variable, particle geometric parameters, and photoelastic material parameters, and the theoretical photoelastic response is generated according to the stress-optical relationship corresponding to the circularly polarized light field or the plane-polarized light field; the theoretical photoelastic response of each particle is mapped to the full-field coordinates according to the particle center position and size to form a full-field synthesized photoelastic image.
[0028] Further, in step 4, the multi-constraint objective function includes at least an image matching term and a particle mechanics equilibrium term; the image matching term is constructed based on the pixel differences and / or structural similarities between the photoelastic image and the theoretical photoelastic response, and the particle mechanics equilibrium term is constructed based on the vector sum and torque sum of the shared contact variables associated with each particle; when the external load or boundary reaction force is known, the multi-constraint objective function also includes an overall load constraint term consisting of the difference between the calculated boundary resultant force and the measured external load.
[0029] Further, in step 5, a physical value range is set for the shared contact variables, and multiple sets of initial values are generated for constrained nonlinear optimization. Candidate optimal results are selected based on the objective function value, particle image residual, and overall load error corresponding to each set of solution results. When the preset particle image residual or mechanical equilibrium residual exceeds a threshold, topological adjustment is preferentially performed on the uncertain contact associated with that particle, and the solution is re-executed under the adjusted candidate contact topology, that is, the contact force inversion is re-executed under the updated candidate contact topology. The mechanical equilibrium residual includes particle force equilibrium residual, particle moment equilibrium residual, etc.
[0030] Furthermore, the particle system is a dry particle system, a saturated particle system, or a particle system subjected to pore fluid action; when the particle system is subjected to pore fluid action, the flow rate, pressure difference, hydraulic gradient, and / or pore velocity are recorded simultaneously, and the quantitative force chain network obtained under different hydraulic conditions is associated and stored according to the corresponding hydraulic state.
[0031] Example 2: To make the objectives, technical solutions, and advantages of the present invention clearer, the following example uses two-dimensional disc-shaped polycarbonate photoelastic particles, combined with... Figures 1-9 The method of the present invention will be further described below. The particle system described is only used to illustrate one embodiment of the present invention and does not constitute a limitation on the particle material, particle size, particle quantity, or specific loading method.
[0032] Step 1: Preprocessing of photoelastic particles, response calibration, and acquisition of experimental images.
[0033] In step 1, before the contact force inversion of the formal particle system, residual stress is released on the photoelastic particles used, and the photoelastic response calibration relationship is established under the current particle material, particle thickness, polarized light field, and image acquisition conditions. Subsequently, unpolarized geometric images and polarized photoelastic images are acquired under the same conditions of the particle system to be analyzed. Step 1 will be further explained below in conjunction with steps 101 to 105.
[0034] Step 101: Perform residual stress release treatment on the photoelastic particles.
[0035] Residual stress may be generated in polycarbonate photoelastic particles during sheet manufacturing, cutting, and machining. When these particles are placed in a polarized light field, the residual stress will generate initial photoelastic fringes under no external load, and this stress will be superimposed on the photoelastic response generated by subsequent external contact forces, thus affecting the quantitative inversion of the stress on the particles. Therefore, it is necessary to anneal the processed photoelastic particles before formal testing. The photoelastic particles are completely immersed in a high-temperature resistant liquid medium and placed as a whole in a constant temperature device; the temperature is gradually increased at a preset heating rate until the photoelastic particles reach the residual stress release temperature and are held for a preset time; then, the temperature is slowly reduced to room temperature at a preset cooling rate.
[0036] In a specific embodiment, the particles can be immersed in vegetable oil and heated to approximately 150°C at a rate of 10–15°C / h, held at this temperature for approximately 8 hours, and then gradually cooled to room temperature at a rate of approximately 10°C / h. These parameters are only one implementation method for polycarbonate photoelastic particles; the annealing temperature, holding time, and heating / cooling rates can be adjusted accordingly based on the particle material, size, and processing technology.
[0037] After the annealing test, the particles were placed in a polarized light field without external force and unloaded images were collected for inspection before and after annealing (e.g., Figure 2 If the initial striations inside the particles are significantly reduced after annealing, it is considered that the residual stress release meets the requirements for subsequent photoelastic tests.
[0038] Step 102: Establish a known loading condition for a single particle. In step 102, a photoelastic particle with the same material, thickness, and processing method as the one used in the formal experiment is selected as a calibration particle. The calibration particle is placed under the same or equivalent polarization path, light source, camera, exposure parameters, and image spatial resolution conditions as the formal experiment. A known external force passing through the vicinity of the particle's center is applied to both sides of the particle. The known loading force is changed step by step, and photoelastic images corresponding to each loading level are acquired.
[0039] In one embodiment, the loading levels used are: 5 N, 10 N, 15 N, 20 N, 30 N, 40 N, 50 N, 60 N, 70 N, 80 N, 90 N, 100 N. The calibration photoelastic images obtained at each loading level constitute a single-particle calibration image sequence (e.g., Figure 3 (As shown).
[0040] Step 103: Determine the effective image area for the calibration particles. For each calibration image, determine the center coordinates of the photoelastic disk and the particle radius. If the circumference basically coincides with the actual outer edge of the particle, the current particle area is accepted; if the center is offset, the particle center is readjusted; if the radius deviation is large, the particle circumference is redefined.
[0041] In one implementation, a MATLAB plugin is used to automatically identify a point near the particle's center and a point on the particle's circumference, and the initial radius of the particle is obtained based on the distance between the two points. Subsequently, the corresponding circumference is displayed in an image, and the particle's center and radius are checked. This yields the center coordinates of the particle under the q-th calibration condition. x q , y q ), and pixel radius R q .
[0042] Furthermore, to reduce the influence of artificial gradients caused by abrupt changes in grayscale at the edges of particle contours, a shrunken effective region is used when calculating the photoelastic response. Definition: , Where, Δ R e This represents the number of pixels indented at the grain edge. R q,eff To account for the influence of artificial gradients, the effective region radius is set to 1 pixel in one embodiment. Circular particles are then created based on this effective region radius. M q (x , y) : .
[0043] Step 104: Calculate the overall squared gradient response of a single particle and establish calibration relationships.
[0044] Convert the q-th calibration photoelastic image into a red, green channel, or grayscale image: I q (x , y) ; Calculate the image gradients of the grayscale image along the horizontal and vertical directions respectively: , G x , q (x , y) Represents the image gradient in the horizontal direction. G y , q (x , y) This represents the image gradient in the vertical direction.
[0045] In one implementation, the gradient is obtained using central difference or an equivalent discrete gradient calculation method.
[0046] Define the squared gradient at the pixel location as: , The squared gradients are then summed within the effective region of the particle: , It should be noted that this embodiment uses the total amount of the squared gradient within the effective region of the particle, rather than the average value of the squared gradient within the particle region.
[0047] Different known loading forces F q With the corresponding Establish calibration data pairs: .
[0048] In one embodiment, a one-time relationship is used: , Fitting the equation to all calibration points, where F is the independent variable, representing the input standard quantity, and G... Σ 2 Let be the dependent variable, representing the total squared particle gradient, a G b is the slope parameter. G This is the intercept parameter. The corresponding inverse function is F. G for: , When the inverse calculation yields F G When < 0, it is set to 0, that is: .
[0049] Under the current calibration image conditions in this embodiment, we can obtain: , Therefore, we can conclude that: .
[0050] Step 105: Acquire geometric and photoelastic images of the particle system to be analyzed. After particle preprocessing and calibration, the photoelastic particles are loaded into the experimental apparatus and subjected to predetermined external conditions. Geometric images under unpolarized light fields and photoelastic images under polarized light fields are acquired under the same particle spatial state. The unpolarized geometric images are mainly used to identify the particle outline, center, and radius; the polarized photoelastic images are mainly used to obtain the photoelastic response of the particles after being subjected to force.
[0051] In one embodiment, the two processed images are saved as PNG image files. For particulate systems containing fluid, background images can be acquired under conditions of no particles, no external load, or other reference conditions, and background subtraction, brightness normalization, or local contrast correction can be performed as needed.
[0052] Step 2: Identify particle geometry information and establish a particle geometry model.
[0053] In this embodiment, step 2 is used to determine the center position and size of all particles based on the non-polarized geometric image obtained in step 1, and to map the geometric information to a polarized photoelastic image to form a particle geometric model that is used uniformly for subsequent contact search and photoelastic inversion. Step 2 will be further explained below in conjunction with steps 201 to 204.
[0054] Step 201: Use reference particles to determine the size range for automatic circle recognition.
[0055] Several particles with relatively clear edges and not obscured by boundaries are selected from the unpolarized image as reference particles.
[0056] In one embodiment, six reference particles are selected, and six boundary points are selected on the circumference of each reference particle. For the boundary point of the m-th reference particle ( x l , y l The least squares method is used to fit the equation of the circle: , get: , In the formula, A, B, and C represent transition parameters, x c and y c The coordinates of the center of the circle are represented by R, and the radius is represented by R. The reference average radius and standard deviation of the radius are obtained from multiple reference particles. Based on the reference radius, the allowable radius range for subsequent automatic circle recognition is determined.
[0057] Step 202: Preprocess the geometric image.
[0058] The non-polarized geometric image is converted to a grayscale image, and grayscale range adjustment is applied to enhance the contrast between particle edges and the background. For test box boundaries, fixed components, or other areas easily misidentified as particle arcs, invalid regions can be pre-defined; for example, a certain width around the image can be set as a region that does not participate in circle detection. Furthermore, Gaussian smoothing can be used to reduce random noise.
[0059] In one embodiment, the standard deviation of Gaussian smoothing is approximately 0.8 pixels.
[0060] Step 203: Use the circle detection method to identify all particles.
[0061] Based on the radius range obtained in step 201, circular target detection is performed on the preprocessed geometric image.
[0062] In one implementation, a grayscale polarity recognition mode is used simultaneously: a first group identifies the position of the bright-colored disk; a second group identifies the edge of the dark-colored disk in the original grayscale image; and two sets of candidate circle detection results and their circle detection confidence indices are obtained.
[0063] Calculate the combined values of all candidate particle detection indicators under both detection modes, and select the set with the better overall detection performance as the final particle identification result. This yields the identified particle location, particle size, and particle quantity N.
[0064] Step 204: Establish a unified set of particle numbers and geometric information.
[0065] All particles are assigned unique numbers representing particle information according to a predetermined spatial order, and the corresponding particle numbers are stored uniformly. For all subsequent contact searches, particle image cropping, single particle inversion, shared contact solutions, and full-field image reconstruction, the fixed geometric information obtained in this step is used, and the particle center and radius are not repeatedly identified.
[0066] Step 3: Construct candidate contact topologies and determine the initial values of each contact force.
[0067] Based on the particle positions and sizes obtained in step 2, candidate particle-particle and particle-boundary contacts are generated, and the initial force values for each contact are determined by combining the overall photoelastic response of the particles and the local photoelastic response near the candidate contacts. Step 3 will be further explained below with reference to steps 301 to 306.
[0068] Step 301: Extract the measured photoelastic image of a single particle.
[0069] Based on the center of the i-th particle ( x i , y i ) and radius R i Extract the corresponding particle region from the complete polarized photoelastic image.
[0070] A disk mask region is established to exclude background pixels outside the circumference of the particle, thereby obtaining the measured photoelastic image of the i-th particle. I i,exp .
[0071] Step 302: Calculate the overall gradient squared response of the particle to be tested.
[0072] Using the same image gradient calculation method as in the calibration stage, the gradient G in two directions is calculated for the photoelastic image of particle i. x,i and G y,i Furthermore, the gradient squared G is obtained. i 2 : , Using the effective region mask corresponding to step 104, the squared gradients inside the disk are summed: , Substituting this value into the single-particle calibration inverse function obtained in step 104, we obtain the overall force scale estimate of the i-th particle. F G,i Note the calculation at this time. F G,i It is not the final result after the inversion of each contact force of the i-th particle, but rather the overall force scale of the particle used to determine the initial contact force values of subsequent candidate particles.
[0073] Step 303: Construct a set of particle-particle geometric nearest neighbors.
[0074] For any two particles i and j, calculate the distance between their centers. d ij : ; Define the surface gap between particles g ij : , when g ij ≤ d tol When particle i and particle j are considered as geometrical candidate contacts, then... d tol To allow for geometric contact tolerances. In the first round of automatic search in one embodiment, d tol The parameter is set to 10 pixels. This parameter is only used to generate a set of possible force-transmitting contacts and does not directly indicate that all geometric neighbors are actual force-transmitting contacts.
[0075] Step 304: Establish candidate particle-boundary contacts.
[0076] For particles near the test boundary, calculate the shortest distance between the outer edge of the particle and the left, right, top, bottom, or other specified boundaries. If the distance is less than the boundary contact tolerance (empirical value of 10 pixels), establish the corresponding particle-boundary candidate contact. Particle-boundary contact and particle-particle contact are then processed together in the subsequent contact force inversion.
[0077] Step 305: Calculate the local photoelastic response of the candidate contact region.
[0078] For a candidate contact between particle i and particle j, the contact direction from particle i to particle j is determined by the line connecting the centers of the two particles. β ij : , A local contact detection area is established near the corresponding circumferential position.
[0079] In one embodiment, the feature size of the local region can be C. R =6 pixels.
[0080] Candidate contacts are calculated within the local region. For all candidate contacts of particle i, the following is obtained: , where z i Let be the number of candidate contacts for particle i. G c,i 2 Indicates particles i No. k The squared response of the photoelastic gradient in the local detection region near the candidate contact is used to characterize the local response intensity of the photoelastic stripes near the candidate contact.
[0081] Step 306: Distribute the initial contact forces according to the local photoelastic response.
[0082] Calculate the sum of the local contact responses of particle i for each candidate contact: , when At that time, based on the proportion of local photoelastic response of the candidate contact, the overall force scale of the particle obtained in step 302 is distributed to different contacts.
[0083] In one embodiment, the initial force at the k-th contact f ik (0) It can be represented as: , In the formula, F G,i This indicates that step 302 is based on the total squared gradient of particle i. The overall force scale of the particles is obtained by back-calculating the total gradient squared amount and the force calibration relationship established in step 104.
[0084] This allows candidate contacts with stronger local photoelastic responses to obtain relatively larger initial contact forces. When local photoelastic responses cannot be stably distinguished, initial contact force values can also be generated using uniform distribution, random perturbation distribution, or methods based on the force states of adjacent particles.
[0085] Step 4: Perform photoelastic forward modeling and contact force inversion on each particle.
[0086] In step 4, the parameters to be optimized are established based on the candidate contact positions of the particles and the initial contact force. The theoretical stress field and theoretical photoelastic image inside the particles are calculated under a given contact force condition. The contact force parameters are continuously adjusted to match the theoretical photoelastic image with the measured photoelastic image of the particles. Step 4 will be further explained below in conjunction with steps 401 to 406.
[0087] Step 401: Establish single-particle contact force parameters.
[0088] Let the i-th particle have a total of z i If there is a contact to be sought, then its parameter P to be optimized is... i It can be represented as: , in, f ik This represents the amplitude of the contact force at the k-th contact. α ik This represents the force direction correction relative to the candidate contact geometry, k = 1, 2, 3... z i The geometric position of the candidate contact on the circumference of the particle is determined in step 305.
[0089] Step 402: Perform particle mechanical balance correction on the initial contact force.
[0090] Based on the candidate initial contact forces obtained in step 306, their vector sum in the two-dimensional coordinate direction is calculated. Under quasi-static conditions that ignore particle inertia, the particle contact forces are adjusted to satisfy predetermined force balance conditions by adjusting the amplitude and / or direction of the contact forces. For particles with two main contacts, the two contact forces can be balanced in amplitude and direction according to the two contact directions; for particles with three or more contacts, the force amplitude and direction are readjusted according to the vector components of all contact forces. The balance correction is used to improve the initial state of subsequent nonlinear inversion.
[0091] Step 403: Calculate the internal stress field of the particles under the given contact force conditions.
[0092] A computational grid was established within the particle, corresponding one-to-one with the pixels of the measured photoelastic image. For any grid point, the stress components generated by each contact force were calculated and linearly superimposed to obtain... σ xx , σ yy , σ xy, representing the total normal stress along the x-direction, the total normal stress along the y-direction, and the total plane shear stress at a mesh point after all contact forces have acted together. The principal stress difference is calculated based on the two-dimensional stress state. Δσ : . Step 404: Generate a theoretical photoelastic image based on the stress-optics relationship.
[0093] Photoelastic materials exhibit stress birefringence when subjected to contact force. Under circularly polarized light field conditions, the theoretical light intensity at any point inside the particle is... I It can be represented as: , in, I 0 represents the incident light intensity. C The material stress optical coefficient, h For particle thickness, λ For effective light wavelength, σ 1- σ 2. Principal stress difference. This stress-optical relationship is consistent with the basic relationship used in polycarbonate photoelasticity tests.
[0094] In the numerical implementation process, the material, thickness, and light source-related parameters can be combined into equivalent photoelastic parameters, which are expressed as: , In the formula, I syn This represents the normalized theoretical photoelastic intensity calculated under the current trial contact force conditions, with a value ranging from 0 to 1. F σ The equivalent photoelastic calibration parameters used for theoretical photoelastic forward modeling are predetermined based on the single-particle fringe order or other equivalent calibration methods under known load conditions, and remain unchanged during the inversion process of the same set of images. By traversing all valid pixels within the particle, a theoretical photoelastic image is generated under the current calculated contact force conditions. I i,syn .
[0095] Step 405: Construct single-particle image residuals.
[0096] The measured photoelastic image obtained in step 301 is denoted as... I i,exp The theoretical photoelastic image obtained in step 404 is denoted as... I i,syn To reduce the impact of particle edge errors on the inversion, a fitting region is set inside the particle.
[0097] In one embodiment, the radius of the fitted region can be taken as... Rfit =0.90 R i Define the fitting mask M. i fi Thus, the pixel residual is obtained. D i : , Expand all valid pixel residuals to form a residual vector r i : ,
[0098] Corresponding image fitting error E i for: .
[0099] Step 406: The contact force is inverted using a nonlinear least squares algorithm.
[0100] The parameters obtained in steps 306 and 402 are used as P i (0) Adjustment is performed using a nonlinear least squares algorithm. f ik and α ik make: .
[0101] In one embodiment, the Levenberg-Marquardt method is used for solving the problem. To avoid the solution entering local extrema due to a single initial value, multiple sets of initial force and direction parameters can be set for the same particle and solved separately. After optimization, the following are saved: final contact force, contact force direction, single-particle image residual, and corresponding theoretically synthesized photoelastic image.
[0102] Step 5: Verify the contact topology and reconstruct the shared contact force.
[0103] In step 5, the preliminary contact network obtained in step 4 is checked a second time, and a unique shared contact force is established for the same particle-particle physical contact, so that adjacent particles use the same contact force variable at the same physical contact position. Step 5 will be further explained below in conjunction with steps 501 to 507.
[0104] Step 501: Expand the candidate nearest neighbor range to perform a missed contact check.
[0105] Recalculate for all particles: , In addition to the particles that have entered the initial contact network in step 303, the particle pairs that meet the expanded geometric tolerance are then checked.
[0106] In one embodiment, the extended search range can be set to g ij ≤22 pixels. The extended nearest neighbors are not directly used as the final contact, but rather as a set of contacts that may be missed during verification.
[0107] Step 502: Calculate the bilateral directional photoelastic response of the candidate contact.
[0108] For candidate particle pairs ij, local sector detection regions are constructed from particle i pointing to particle j and from particle j pointing to particle i, respectively.
[0109] In one embodiment, the direction of the fan-shaped center is the direction of the line connecting the centers of the particles, and the half-angle of the fan is approximately 15°. The radial range is approximately 0.68R~0.98R (R is the radius of the circle). The local photoelastic gradient response and background response on the contact sides of the two particles are calculated respectively, and a normalized index Z is constructed. ij and Z ji When both facing regions of the particles exhibit a significant photoelastic response, the confidence level of the candidate contact is increased; when only one side shows a photoelastic response, it is marked as an indeterminate contact; when both sides lack corresponding responses, the priority of the candidate contact can be reduced or the contact can be deactivated.
[0110] Step 503: Form the corrected contact topology.
[0111] Taking into account particle geometry, automatic contact recognition results, and local G-forces... 2 The candidate contacts are retained, deleted, or re-enabled based on the response, bilateral directional photoelastic response, single-particle inversion residual, and particle force balance results. This yields the final candidate contact topology for shared contact force calculation.
[0112] Step 504: Check the bidirectional inversion differences for the same physical contact.
[0113] For the same physical contact between particles i and j, f may be obtained in the single-particle independent inversion in step 4. ij and f ji Define the average of the two: , Simultaneously define the degree of relative inconsistency in contact force. δ ij : , The index is used to identify particle pairs with a high degree of inconsistency in single-particle independent inversion.
[0114] Step 505: Establish a unique shared contact variable for each physical contact.
[0115] For any pair of particles in the final candidate contact topology, e=( i,j Only one shared contact force F is set. ij The shared contact force participates in the photoelastic forward modeling of both particle i and particle j.
[0116] For particle i, the corresponding force is written as F. ij For particle j, it is written as F ji This satisfies the interaction and reaction forces between particles during the variable definition stage, rather than simply averaging the two results after the two particles have been solved independently.
[0117] Step 506: Recalculate the photoelastic response of particles on both sides of the contact using shared contact force.
[0118] Shared contact force F ij Substitute the photoelastic forward model of the particles on both sides of the contact into the model to regenerate. I i, shared and I j, shared , respectively, represent the introduction of shared contact variables F ij The theoretical composite photoelastic images of particles i and j were then recalculated, and calculations were performed on them respectively. E i, shared , E j, shared And calculate the average residual of the particle pairs after the shared solution: , Average residuals before shared processing E ij,before Compare the results. Define the residual improvement rate. P ij : . Step 507: Determine the stable shared contact force through repeated calculations with multiple initial values.
[0119] For each shared contact or a group of interconnected shared contacts, multiple initial force values are set and repeated inversion is performed. The calculated contact force is... Calculate its mean: , Standard deviation S F,ij and coefficient of variation CV ij : .
[0120] Taking into account factors such as the image residual after sharing, the coefficient of variation of contact force, particle mechanical equilibrium, and the stability of results with multiple initial values, the shared contact result that meets the preset quality conditions is selected as the final contact force. For contacts or particles that do not meet the quality conditions, steps 501-503 can be returned to re-examine the contact topology, and the shared contact force inversion can be repeated after adjustment.
[0121] Step 6: Generate a full-field synthetic photoelastic image, a quantitative force chain network, and perform overall verification. In step 6, the final shared contact force is re-substituted into the photoelastic forward model of all particles to generate the theoretical photoelastic response of the complete particle system and establish a quantitative contact force network. Furthermore, the quality of the final contact force network is evaluated using particle image residuals and system force transmission results. Step 6 will be further explained below in conjunction with steps 601–606.
[0122] Step 601: Regenerate the final theoretical photoelastic image of all particles.
[0123] For the i-th particle, input all the finally determined shared contact forces into the photoelastic forward model described in steps 403-404 to obtain... The theoretical photoelastic image corresponds one-to-one with the final shared contact force of the particles.
[0124] Step 602: Restore the single-particle theoretical image to full-field coordinates. Create a blank image with the same size as the complete experimental polarization image. According to step S2 ,Will The image is mapped to the actual position of the particles within the complete sample. All particles are processed sequentially to obtain a full-field theoretically synthesized photoelastic image of the complete particle system. This full-field synthesized photoelastic image is then displayed side-by-side or superimposed with the experimental polarization image to check the consistency of the overall fringe distribution.
[0125] Step 603: Establish a contact force weighted network.
[0126] For the final particle contact set E, establish a contact force weighted adjacency matrix W. When particles i With particles j When there is an eventual shared contact W ij = F ij ,otherwise: W ij =0. Since only one shared contact force is set for the same physical contact, therefore... W ij = W ji Draw the contact line based on the particle's center position. The contact line width or other graphic attributes are determined according to... Fij Mapping is performed to form a quantitative force chain network.
[0127] Step 604: Calculate the particle-scale image residual.
[0128] For all effective particles, calculate E i The average particle residual is then obtained: ,
[0129] The solution also includes the median of particle residuals and the number of particles with high residuals. Particles with residuals exceeding preset values are marked as abnormal particles and returned for inspection: particle center, particle radius, missed contact, false contact, particle-boundary contact, and shared contact solution stability.
[0130] Step 605: Perform system force transmission verification using known external loads.
[0131] When the test has a known macroscopic external load, one or more statistical sections are set inside the specimen. For the m-th statistical section, the complete set of contacts C(m-th) passing through that section is identified. S m ).
[0132] Calculate the sum of the components of the contact force along the macroscopic loading direction: , in, n e The contact force is a unit vector. n L The macroscopic loading direction is defined. For multiple statistical sections, the average cross-sectional force transmission is calculated: , Further calculation of the coefficient of variation of force transmission across the cross section: , Will F S External measured load F load Compare the results. Define the macroscopic load error: .
[0133] When the particle image residual, shared contact stability, and macroscopic load error all meet the preset conditions, the current result is used as the final photoelastic contact force reconstruction result; otherwise, return to step 5 to re-check and calculate abnormal contacts or abnormal particles.
[0134] In one embodiment, such as Figure 8As shown, multiple horizontal statistical sections are set inside the particle sample along the direction perpendicular to the vertical loading direction. To reduce the influence of local contact force concentration and boundary effects near the upper and lower loading boundaries on the statistical results, six statistical sections (S2, S3, S4, S5, S6, and S7) located in the middle of the sample are selected for overall force transmission verification (e.g., ...). Figure 8 The transmitted forces at each statistical section from S2 to S7, calculated based on the final shared contact force network, are 86.77 N, 79.51 N, 84.30 N, 77.65 N, 87.67 N, and 86.31 N, respectively.
[0135] The average transmitted force of the six statistical sections is F S =83.55N. The actual external load applied in the test was: F load =80.00 N (The force applied by the testing machine is 100N, the background force of the measuring system is 20N, and the corrected test force is 80N).
[0136] Therefore, the relative error between the average cross-sectional transmitted force and the measured external load is: , The coefficients of variation of the force transmitted at the six statistical cross sections are: CV S =4.78%.
[0137] Step 606: Results summary and archiving output.
[0138] Based on the output of particle contact force, full-field synthesized photoelastic image, quantitative force chain network, and quality evaluation results, the particle contact force reconstruction results for the current working condition are uniformly summarized and archived. Archived content may include: particle number and its center position, radius, and other geometric information; particle-particle and particle-boundary final contact topology; shared contact force and its direction for each physical contact; particle image fitting residuals; calculation results of multiple initial values for shared contacts and their mean, standard deviation, and coefficient of variation; full-field synthesized photoelastic image; contact force weighted chain network; and force transmission results at a specified section and the error between these results and the measured external load.
[0139] Furthermore, particle geometry information, candidate contact topology, final contact topology, shared contact force, and error evaluation results can be indexed according to a unified particle number and contact number. For continuous loading processes or multiple test conditions, a test condition index can be established according to the test time, loading level, sample number, and / or hydraulic state. The particle geometry, contact topology, shared contact force, synthesized photoelastic image, and quality evaluation indicators corresponding to different test conditions can be associated and stored, thereby ensuring that the particle contact force reconstruction results are traceable and verifiable.
[0140] Example 3: Based on the same inventive concept as the aforementioned method, a quantitative force chain reconstruction system with shared contact variables and multi-constraint inversion is provided to realize the quantitative force chain reconstruction method with shared contact variables and multi-constraint inversion as described above.
[0141] The system includes: The particle preprocessing and calibration data management unit is configured to acquire and manage photoelastic particle material parameters, particle thickness, imaging parameters, and single-particle calibration images under known loading conditions; calculate the overall squared gradient response of the particles under different known loads; and establish G... 2 - Force calibration relationship.
[0142] The image and geometric information acquisition unit is configured to acquire non-polarized geometric images and polarized photoelastic images of the particle system to be analyzed, perform spatial correspondence between the two types of images, and identify the particle center position, size, number, and experimental boundary to form the particle geometric information required for subsequent contact search and contact force inversion.
[0143] Candidate contact topology building units are configured to establish candidate particle-particle and particle-boundary contacts based on particle center-to-center distance, particle size, and the distance between the particle outer edge and the test boundary, and to calculate the overall particle G. 2 Local G near the response and candidate contact 2 The response is used to obtain the initial contact force corresponding to each candidate contact.
[0144] The single-particle photoelastic inversion unit is configured to calculate the internal stress field of the particle based on the candidate contact position, contact force amplitude, and contact force direction, and generate a theoretical photoelastic image based on the stress-optic relationship; based on the residual between the theoretical photoelastic image and the measured photoelastic image, the particle contact force and its direction parameters are updated through a nonlinear optimization method to obtain the preliminary contact force inversion results for each particle.
[0145] The contact topology verification and shared contact reconstruction unit is configured to verify the candidate contact topology by combining the extended geometric nearest neighbor, the directional photoelastic response on both sides of the candidate contact, and the particle image residual, and to establish a unique shared contact variable for the same physical contact, so that the shared contact variable participates in the photoelastic forward modeling and contact force inversion of the particles on both sides of the contact.
[0146] The quality evaluation and overall force calibration core unit is configured to regenerate the theoretical photoelastic image of each particle based on the final contact force and form a full-field synthesized photoelastic image. It calculates the particle image residual, the mean, standard deviation and coefficient of variation of the solution results of shared contact multiple initial values, and calculates the internal force transmission results of the particle system based on the contact force components of the specified cross section. It then compares these results with the measured external load to evaluate the reliability of the contact force reconstruction results.
[0147] The results summary and force chain output unit is configured to output particle geometry information, final contact topology, particle-particle contact force, particle-boundary contact force, full-field synthesized photoelastic image, contact force weighted chain network, particle residual, shared contact stability index, and cross-sectional force transmission results when the preset quality conditions are met. It also establishes a results index and completes archiving according to the sample number, working condition, and / or time. When the preset quality conditions are not met, it drives the candidate contact topology and initial value construction unit, single particle photoelastic inversion unit, and contact topology verification and shared contact reconstruction unit to continue to perform contact topology adjustment and contact force inversion until the preset quality conditions are met.
[0148] This invention focuses on the quantitative inversion of contact force in photoelastic particle systems, combining the release of residual stress in photoelastic particles with the single-particle G... 2 This invention unifies force calibration, particle geometry identification, candidate contact topology construction, initial contact force estimation, single-particle photoelastic contact force inversion, contact topology verification, shared contact force reconstruction, and full-field photoelastic reconstruction and overall force verification within a single computational framework. This improves the stability, mechanical consistency, and traceability of particle contact force reconstruction results, even under conditions of complex particle contact relationships, uncertain weak contact identification, and discrepancies in independent inversion results for single particles. It addresses problems in existing photoelastic particle contact force calculations, such as the easy propagation of candidate contact identification errors to subsequent inversions, the difficulty in maintaining consistency between independent solutions on both sides of the same physical contact, and the lack of overall force verification in local image fitting results. Compared to existing technologies, this invention can uniformly invert and verify particle-scale photoelastic response, contact topology, and overall force transmission state, improving the reliability and interpretability of quantitative reconstruction of contact forces and force chain networks in complex particle systems.
[0149] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.
Claims
1. A quantitative reconstruction method for force chains based on shared contact variables and multi-constraint inversion, characterized in that, Includes the following steps: Obtain geometric and photoelastic images of the same particle system under unpolarized and polarized light fields, and spatially register the geometric and photoelastic images to form a set of particle geometric information. The geometric proximity relationships between particles and between particles and boundaries are determined based on the set of particle geometric information, and candidate contact topologies are constructed by combining the local photoelastic response of the candidate contact regions. For the actual physical contact in the candidate contact topology, a shared contact variable is established, and a mapping relationship between the shared contact variable and the particles on both sides of the contact is established. The internal stress field of the particles is calculated based on the shared contact variable, and a theoretical photoelastic response corresponding to each particle is generated. Based on the difference between the photoelastic image and the theoretical photoelastic response, an image residual is constructed. The image residual is then combined with the particle force balance residual, particle torque balance residual, and boundary or overall load residual constructed based on the shared contact variable to form a multi-constraint objective function. Given a candidate contact topology and initial values of shared contact variables, the multi-constraint objective function is subjected to constrained nonlinear optimization to update the candidate contact topology until a preset convergence condition is met. When the preset convergence condition is met, the particle-particle contact force and particle-boundary contact force corresponding to each shared contact variable are output. A full-field synthetic photoelastic image and a quantitative force chain network with contact force magnitude and direction information are generated according to the actual spatial position of the particles. At the same time, an error evaluation index for characterizing the reliability of the reconstruction result is output.
2. The method for quantitative reconstruction of force chains based on shared contact variables and multi-constraint inversion according to claim 1, characterized in that, Candidate contacts between particles and between particles and boundaries are constructed based on the geometric proximity relationships between particles and between particles and boundaries. Candidate contacts between particles are determined based on the difference between the particle center distance and the sum of the corresponding particle sizes. Candidate contacts between particles and boundaries are determined based on the shortest distance between the outer edge of the particle and the test boundary. For candidate contacts that meet the geometric proximity conditions, photoelastic brightness, grayscale gradient, stripe directionality, and / or local texture response are extracted in a local region based on the direction of the line connecting the candidate contact point and the particle center. The effectiveness of the candidate contacts is evaluated based on the local photoelastic response.
3. The method for quantitative reconstruction of force chains based on shared contact variables and multi-constraint inversion according to claim 1, characterized in that, The shared contact variables are represented by contact force amplitude and direction parameters, or by normal contact force components and tangential contact force components. For the same physical contact between particle i and particle j, only one set of shared contact variables is set, and they are applied to particle i and particle j respectively by direction transformation, so that the forces of the same physical contact on the two adjacent particles satisfy the relationship of equal magnitude and opposite direction.
4. The method for quantitative reconstruction of force chains based on shared contact variables and multi-constraint inversion according to claim 1, characterized in that, The principal stress difference at each position inside the particle is calculated based on the shared contact variable, particle geometric parameters, and photoelastic material parameters. The theoretical photoelastic response is generated based on the stress-optical relationship corresponding to the circularly polarized light field or the plane-polarized light field. The theoretical photoelastic response of each particle is mapped to the full-field coordinates according to the particle center position and size to form a full-field synthesized photoelastic image.
5. The method for quantitative reconstruction of force chains based on shared contact variables and multi-constraint inversion according to claim 1, characterized in that, The multi-constraint objective function includes at least an image matching term and a particle mechanics equilibrium term; the image matching term is constructed based on the pixel differences and / or structural similarities between the photoelastic image and the theoretical photoelastic response, and the particle mechanics equilibrium term is constructed based on the vector sum and torque sum of the shared contact variables associated with each particle; when the external load or boundary reaction force is known, the multi-constraint objective function also includes an overall load constraint term consisting of the difference between the calculated boundary resultant force and the measured external load.
6. The method for quantitative reconstruction of force chains based on shared contact variables and multi-constraint inversion according to claim 1, characterized in that, The shared contact variables are given a physical range and multiple initial values are generated. Constrained nonlinear optimization is performed on each set. The candidate optimal result is selected based on the objective function value, particle image residual, and overall load error corresponding to the solution results of each set of initial values. When the image residual or mechanical equilibrium residual of the preset particle exceeds the threshold, the uncertain contact associated with the particle is adjusted first, and the solution is resolved under the adjusted candidate contact topology.
7. The method for quantitative reconstruction of force chains based on shared contact variables and multi-constraint inversion according to claim 1, characterized in that, The particle system is a dry particle system, a saturated particle system, or a particle system subjected to pore fluid. When the particle system is subjected to pore fluid, the flow rate, pressure difference, hydraulic gradient, and / or pore velocity are recorded simultaneously, and the quantitative force chain network obtained under different hydraulic conditions is associated and stored according to the corresponding hydraulic state.
8. A quantitative force chain reconstruction system based on shared contact variables and multi-constraint inversion, used to implement the quantitative force chain reconstruction method based on shared contact variables and multi-constraint inversion as described in any one of claims 1 to 7, characterized in that, The system includes: The image and geometric information acquisition unit is used to acquire geometric and photoelastic images of the same particle system, complete image registration, and identify the particle center position, size, number, and experimental boundary. Candidate contact topology building units are used to construct candidate contact topologies for particle-particle and particle-boundary relationships based on particle geometric proximity and local photoelastic response, and to identify determined and uncertain contacts. The shared contact variable and photoelastic forward modeling unit are used to establish a unique shared contact variable for each physical contact, map the shared contact variable to the particles on both sides of the contact, and calculate the internal stress field and theoretical photoelastic response of the particles based on the shared contact variable. The multi-constraint objective building unit is used to construct image residuals based on photoelastic images and theoretical photoelastic responses, and to construct particle force balance residuals, particle torque balance residuals, and boundary or overall load residuals based on shared contact variables, so as to form a multi-constraint objective function. The adaptive inversion and topology update unit is used to perform constrained nonlinear optimization and multiple initial value repeated solutions on the multi-constraint objective function, and adjust the activation state of uncertain contact according to the residual evaluation results after the solution until the preset convergence condition is met. The results are summarized and output by the force chain output unit, which is used to output particle-particle contact force, particle-boundary contact force, full-field synthesized photoelastic image, quantitative force chain network, and error evaluation index.
9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the quantitative reconstruction method of force chain based on shared contact variables and multi-constraint inversion as described in any one of claims 1 to 7.
10. A non-transitory computer-readable storage medium storing computer instructions thereon, characterized in that, When the computer instructions are executed by the processor, the computer performs the quantitative reconstruction method of force chain based on shared contact variables and multi-constraint inversion as described in any one of claims 1 to 7.
