Unmanned aerial vehicle blade cut simulation evaluation method based on test calibration

Abstract

The application discloses a kind of based on test calibration unmanned aerial vehicle blade cut simulation evaluation method, it is related to finite element simulation technical field, including, according to target supplementary test condition to carry out blade cut supplementary test, supplementary test data and blade cut test data are mapped to unified double residual space, the initial calibration parameter domain is carried out multi-condition cross constraint, according to double residual response difference contraction initial calibration parameter domain, generate credible parameter domain, and credible parameter domain is loaded to blade cut simulation model, executes parameter envelope finite element calculation, generates blade cut response envelope.The application realizes the collaborative evaluation of load response and crack propagation uncertainty, improves the credibility of blade cut risk assessment.

Description

Technical Field

[0001] This invention relates to the field of finite element simulation technology, and in particular to a simulation evaluation method for UAV propeller blade cuts based on experimental calibration. Background Technology

[0002] In the field of human-machine interaction safety and airworthiness verification for unmanned aerial vehicles (UAVs), blade cut assessment based on finite element simulation has become a core means of quantifying contact damage risk. Conventional methods usually rely on standard tests to obtain material properties and load boundaries, construct a biomimetic skin coupling model of the blade, reproduce the mechanical response and damage morphology during the collision process through explicit dynamic solutions, and combine experimental data to statically calibrate key parameters, thereby predicting the severity of cuts under different working conditions and providing a basis for the design of protective structures.

[0003] However, when dealing with multi-parameter coupling effects, conventional methods are prone to parameter confusion when calibrating based on overall error because dynamic parameters and biomimetic skin damage parameters may have similar effects on load response and crack evolution. Parameter calibration often relies on limited test conditions and lacks screening of supplementary test conditions and cross-condition constraints for confused parameters. The resulting parameter range may be wide and it is difficult to stably characterize the uncertainty boundary of the response. Summary of the Invention

[0004] In view of the aforementioned existing problems, the present invention is proposed.

[0005] Therefore, this invention provides a simulation evaluation method for UAV blade cuts based on experimental calibration to solve the problems of identification confusion under multi-parameter coupling and the lack of cross-condition constraints in the calibration parameter domain.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution: This invention provides a simulation evaluation method for UAV blade laceration based on experimental calibration, comprising: establishing a blade laceration simulation model based on blade laceration test data; setting the dynamic parameters involved in the collision and the biomimetic skin damage parameters as parameters to be calibrated; performing explicit dynamic finite element calculations and extracting finite element load response characteristics and finite element crack evolution characteristics; aligning the time reference and spatial coordinate system with the blade laceration test data to obtain load transfer residuals and crack evolution residuals; applying local perturbations to the parameters to be calibrated; constructing a parameter residual causality graph based on the perturbation response of the load transfer residuals and crack evolution residuals; dividing the load action layer and crack action layer according to the dominant residuals; calibrating the parameters of the two layers sequentially; adjusting the cross-layer parameters with double residual constraints; and generating an initial calibration parameter domain. The calibration parameters are locally re-perturbed to extract parameter identification feature vectors. Confused parameter groups are identified based on the degree of overlap and sensitivity degradation of these feature vectors. Simultaneously, candidate supplementary test conditions are generated within the executable operating range. A finite element method (FEM) simulation of the blade cut is used to pre-simulate the problem, calculating the response separation and evidence observability of the confused parameter groups to generate the target supplementary test condition. A blade cut supplementary test is then conducted according to the target condition. The supplementary test data and the blade cut test data are mapped to a unified dual residual space. Multi-condition cross-constraints are applied to the initial calibration parameter domain. The initial calibration parameter domain is shrunk based on the difference in dual residual responses to generate a reliable parameter domain. This reliable parameter domain is then loaded into the blade cut simulation model, and parameter envelope finite element calculations are performed to generate the blade cut response envelope.

[0007] As a preferred embodiment of the experimentally calibrated UAV blade cut simulation evaluation method described in this invention, the steps of performing explicit dynamic finite element calculations and extracting finite element load response characteristics and finite element crack evolution characteristics are as follows: Based on the blade cut test data, the blade and bionic skin were discretized using the finite element method. The blade was set as a rigid finite element body, and the bionic skin was set as a hyperelastic failureable finite element body. A reference point was established at the blade rotation center. The blade motion boundary, the bionic skin constraint boundary, and the contact relationship between the blade and the bionic skin were loaded into the finite element discretized structure to generate a blade cut simulation model. The dynamic parameters involved in the collision and the biomimetic skin damage parameters were extracted from the blade cut test data. The two types of parameters were set as parameters to be calibrated, and explicit dynamic finite element calculations were performed to generate finite element response sequences. The support reaction time history is extracted from the finite element response sequence, and the formation time and adjacent expansion relationship of the damaged area are identified according to the bionic skin failure criterion. The support reaction time history is organized into finite element load response characteristics, and the spatial location and expansion time sequence of the damaged area are organized into finite element crack evolution characteristics.

[0008] As a preferred embodiment of the experimentally calibrated UAV blade cut simulation evaluation method described in this invention, the steps for obtaining the load transfer residual and the crack evolution residual are as follows: Using the moment of first contact of the blade as the test time reference, the test support reaction time history is extracted from the blade laceration test data, and the bionic skin surface coordinates are established. The leading edge of the crack in the continuous test images is mapped to the bionic skin surface coordinates according to the crack initiation position and the progression order, generating test load response characteristics and test crack evolution characteristics. The finite element load response characteristics and the experimental load response characteristics are time-aligned according to the experimental time base. The finite element crack evolution characteristics and the experimental crack evolution characteristics are projected onto the bionic skin surface coordinates. The differences in load amplitude and transmission timing, as well as the differences in crack leading edge position and expansion process, are calculated respectively to obtain the load transmission residual and crack evolution residual.

[0009] As a preferred embodiment of the experimentally calibrated UAV blade cut simulation evaluation method of the present invention, the steps for constructing the parameter residual cause-effect graph are as follows: Forward and reverse disturbances are constructed around the current values ​​of each parameter to be calibrated. According to the parameter isolation principle, each local disturbance combination is loaded into the blade cut simulation model. Explicit dynamic finite element calculation is performed while keeping the non-disturbance parameters and collision conditions unchanged to generate a parameter disturbance simulation set. The parameter disturbance simulation set and the blade cut test data are integrated into a time base and spatial coordinate system. The load transfer residual and crack evolution residual under each combination of disturbance parameters are calculated. The two types of residuals corresponding to the current value of the parameter to be calibrated are used as the disturbance reference. The time series change and amplitude change of the residuals are extracted to generate a parameter dual residual response set. By using the parameter dual residual response set, a directed interaction connection is established between the parameter to be calibrated and two types of residuals. The causal direction is determined according to the order of the residual responses, and the interaction weights are set according to the stability of the response direction and the contribution of residual changes, thus constructing a parameter residual causal graph.

[0010] As a preferred embodiment of the experimental calibration-based UAV blade cut simulation evaluation method of the present invention, the steps for generating the initial calibration parameter domain are as follows: Based on the influence weight of each parameter to be calibrated on the two types of residuals in the parameter residual causality diagram, the degree of dominance of the parameter to be calibrated on the load transfer residual and the crack evolution residual is calculated, the load action layer and the crack action layer are divided, and the parameter to be calibrated that has a directed action connection on both the load transfer residual and the crack evolution residual is determined as the cross-layer parameter, and a layered parameter set is generated. The load-bearing layer parameters are traversed through the layered parameter set and parameter combinations whose load transfer residuals are within a preset limit are selected. Within the selected parameter combinations, the crack-bearing layer parameters are locally searched, and the cross-layer parameters are corrected based on the same direction of change of the two types of residuals. The intersection constraints are applied to parameter combinations whose residuals are both within the limit range to generate the initial calibration parameter domain.

[0011] As a preferred embodiment of the experimental calibration-based UAV blade cut simulation evaluation method of the present invention, the step of extracting parameters and identifying feature vectors is as follows: Within the initial calibration parameter domain, the parameter combination with the minimum normalized sum of load transfer residual and crack evolution residual is selected as the re-perturbation reference. Around the values ​​of each parameter to be calibrated in the re-perturbation reference, positive local re-perturbation and reverse local re-perturbation are constructed according to multiple perturbation amplitudes and loaded into the blade cut simulation model to perform explicit dynamic finite element calculation, generating a parameter re-perturbation simulation set. By integrating the parameter re-perturbation simulation set with the blade cut test data into a unified time and spatial coordinate system, the temporal amplitude response of the load transfer residual and the crack evolution residual relative to the local re-perturbation is extracted and normalized to generate parameter identification feature vectors.

[0012] As a preferred embodiment of the experimental calibration-based UAV blade cut simulation evaluation method of the present invention, the steps for generating candidate supplementary test conditions are as follows: The distance between the parameter identification feature vectors is calculated based on the directional angle between them and the temporal overlap relationship. The response sensitivity is calculated based on the ratio of the change in the double residual to the local reperturbation amplitude. The parameters to be calibrated that are closest to each other and whose response sensitivity decreases with the increase of the local reperturbation amplitude are merged to generate a confused parameter group. Within the executable operating conditions, the collision motion states are discretely combined, and each discrete combination is associated with a set of confusion parameters to generate candidate supplementary test conditions and corresponding finite element pre-simulation task sets.

[0013] As a preferred embodiment of the experimental calibration-based UAV blade cut simulation evaluation method of the present invention, the steps for generating the target retest condition are as follows: Execute the finite element pre-test task set, calculate the response separation degree based on the parameter identification feature vector distance of the confused parameter group under each candidate supplementary test condition, and calculate the evidence observability based on the coverage of the finite element load response characteristics and finite element crack evolution characteristics within the test acquisition range, and generate the candidate test condition evaluation matrix. The candidate working condition evaluation matrix is ​​sorted by dual objectives and non-dominated order. Candidate supplementary working conditions are extracted from the first non-dominated layer. The response separation degree and evidence observability are positively normalized. The maximum value of the two normalized indicators is used to form the dual objective ideal point. The deviation distance between each candidate supplementary working condition and the dual objective ideal point is calculated. The candidate supplementary working condition with the smallest deviation distance is determined as the target supplementary working condition.

[0014] As a preferred embodiment of the experimentally calibrated UAV blade cut simulation evaluation method described in this invention, the steps for generating the reliable parameter domain are as follows: According to the target test conditions, conduct blade cut test, obtain test data, load the parameter combination in the initial calibration parameter domain into the blade cut simulation model to perform explicit dynamic finite element calculation, and integrate the time reference and spatial coordinate system with the test data to obtain the test load residual and the test crack residual. The residuals of the supplementary load and the residuals of the supplementary crack are normalized and mapped to the residuals of load transfer and crack evolution to generate a multi-condition dual residual matrix. The matrix is ​​then reorganized according to the parameter combination identifier and the condition index. The consistency of the change direction of the two types of residuals under different conditions and the difference of normalized amplitude are calculated for the same parameter combination to generate a cross-condition residual difference spectrum. Using the cross-condition residual difference spectrum, the direction of change of the double residuals of the same parameter combination under different conditions is encoded as a constraint vector. Parameter combinations with the same direction of change of double residuals are selected according to the constraint vector, and non-dominated sorting is performed according to the normalized amplitude difference. Parameter combinations that are not simultaneously dominated by other parameter combinations are extracted to generate a cross-compatible parameter sequence. Connect the cross-compatible parameter sequences according to the adjacency relationship of the parameter values, calculate the double residual gradient direction between adjacent parameter combinations, take the parameter combination with the minimum sum of normalized double residuals as the core node, extract the parameter region that is connected to the core node and has a continuous gradient direction, and perform intersection constraint with the initial calibration parameter domain to generate a reliable parameter domain.

[0015] As a preferred embodiment of the experimentally calibrated UAV blade cut simulation evaluation method described in this invention, the generation of blade cut response envelope refers to extracting boundary parameter combinations and internal representative parameter combinations from the reliable parameter domain and loading them into the blade cut simulation model to perform parameter envelope finite element calculation, generating parameter envelope response sequences for each parameter combination, calculating the upper and lower bounds of the load response according to a unified time scale, and reconstructing the spatial envelope of the crack leading edge trajectory to generate the blade cut response envelope.

[0016] The beneficial effects of this invention are as follows: By constructing a parameter residual causal graph, the load action layer and the crack action layer are divided according to the load transfer residual and the crack evolution residual, and double residual constraints are applied to the cross-layer parameters, which realizes the hierarchical identification of the parameters to be calibrated and the shrinkage of the calibration parameter domain, thereby improving the interpretability of finite element simulation parameter calibration; by extracting parameter identification feature vectors to identify confused parameter groups, and combining response separation degree and evidence observability to generate target supplementary test conditions, the collaborative evaluation of load response and crack propagation uncertainty is realized, thereby improving the credibility of blade cut risk assessment. Attached Figure Description

[0017] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 This is a flowchart of a simulation evaluation method for UAV blade cuts based on experimental calibration.

[0019] Figure 2 A flowchart for generating a hierarchical parameter set.

[0020] Figure 3 A flowchart for generating the target supplementary test conditions.

[0021] Figure 4 A flowchart for generating the blade cut response envelope. Detailed Implementation

[0022] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0023] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0024] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it a single or selective embodiment that is mutually exclusive with other embodiments.

[0025] Reference Figures 1-4As one embodiment of the present invention, this embodiment provides a simulation evaluation method for UAV blade cuts based on experimental calibration, including the following steps: S1. Based on the blade laceration test data, establish a blade laceration simulation model, set the dynamic parameters involved in the collision and the bionic skin damage parameters as parameters to be calibrated, perform explicit dynamic finite element calculation and extract the finite element load response characteristics and finite element crack evolution characteristics, and align the time reference and spatial coordinate system with the blade laceration test data to obtain the load transfer residual and crack evolution residual.

[0026] S1.1: Based on the blade cut test data, the blade and bionic skin are discretized using finite element method. The blade is set as a rigid finite element body, and the bionic skin is set as a hyperelastic failureable finite element body. A reference point is established at the blade rotation center. The blade motion boundary, the bionic skin constraint boundary, and the contact relationship between the blade and the bionic skin are loaded into the finite element discretized structure to generate a blade cut simulation model.

[0027] Specifically, the blade shape, bionic skin thickness, clamping position, and collision attitude were extracted from the blade laceration test data. The geometry of the blade and bionic skin was reconstructed according to the test dimensions and relative positions, and the motion range of the blade contact edge was projected onto the bionic skin surface to determine the collision contact area. The blade and bionic skin were meshed, and the impact position of the bionic skin was refined according to the rule that the mesh size of the collision contact area is smaller than that of the outer area, completing the finite element discretization. The blade finite element nodes were constrained to synchronous rigid body motion, forming a rigid finite element body, thus configuring the bionic skin with hyperelasticity. The material relationship and failure exit rules are used to form a hyperelastic failureable finite element body. A reference point is established at the blade rotation center, and the translational and rotational degrees of freedom of the blade finite element nodes are coupled to the reference point. The translational and rotational motions loaded at the reference point are transmitted to the blade through the coupling relationship. The blade motion boundary is applied to the reference point according to the collision posture, the bionic skin edge nodes are constrained according to the clamping position, and the blade contact surface is paired with the bionic skin impact surface. The normal load and tangential load are transmitted according to the surface closure state and relative slip state to generate a blade cut simulation model.

[0028] It should be noted that the hyperelastic material relationship is based on the stress-strain curve obtained from the bionic skin material test to select the hyperelastic strain energy function, and the bionic skin damage parameters are determined by curve fitting. At each calculation time of the explicit dynamic finite element calculation, the strain energy density is calculated according to the deformation gradient of the bionic skin finite element mesh region, and then the corresponding stress is obtained by differentiating the hyperelastic strain energy function, thereby characterizing the nonlinear mechanical response of the bionic skin in the process of stretching and large deformation. The failure exit load rule sets the failure criteria based on the bionic skin damage parameters, and reads the stress state of the bionic skin finite element mesh region at each calculation moment of the explicit dynamic finite element calculation. The stress state is compared with the failure criteria. The finite element mesh region that reaches the failure condition is recorded with the failure time and spatial location, and stops participating in subsequent contact load transfer and internal stress calculation. The finite element mesh regions that continuously reach the failure condition form a crack propagation path.

[0029] S1.2: Extract the dynamic parameters involved in the collision and the biomimetic skin damage parameters from the blade cut test data, set the two types of parameters as parameters to be calibrated, perform explicit dynamic finite element calculation, and generate finite element response sequences.

[0030] Specifically, the mass of the blade, connecting structure, and drive structure moving synchronously in the collision direction is read from the blade cut test data. This synchronous mass is then converted and summed according to the corresponding degrees of freedom to obtain the equivalent mass. Based on the mass distribution of each participating component and its position relative to the blade's rotation axis, the equivalent moment of inertia is determined according to the moment of inertia composition relationship. The equivalent mass and equivalent moment of inertia are then used as dynamic parameters for the collision. The time interval during which the blade maintains contact with the bionic skin and undergoes continuous sliding is selected from the blade cut test data. The ratio of tangential load to normal load is calculated moment by moment. After eliminating abrupt changes in the ratio during contact establishment and disengagement, the remaining ratio is statistically analyzed. The contact friction parameters are obtained. Based on continuous test images, the initiation time and location of the first crack in the bionic skin are determined. The local elongation ratio is calculated using the displacement changes of the marked points around the crack location before and after crack initiation. The local elongation ratio is then converted into the equivalent strain at the crack initiation time to obtain the failure initiation strain. The test load and loaded displacement during the period from the start of crack propagation to the cessation of crack propagation are integrated to obtain the external work done during crack propagation. The ratio of the external work done to the area of ​​the newly added crack during the same period is calculated to obtain the failure evolution energy. The failure initiation strain and failure evolution energy are jointly determined as the failure parameters of the bionic skin. The contact friction parameters and the failure parameters of the bionic skin are used as the damage parameters of the bionic skin.

[0031] Based on the measurement error and sample dispersion of the blade cut test data, the range of values ​​for the dynamic parameters involved in the collision and the bionic skin damage parameters were set, and these parameters were loaded into the blade cut simulation model as parameters to be calibrated. Explicit dynamic finite element calculations were performed according to the time step. At each time step, the displacement and rotation angle of the reference point were updated according to the blade motion boundary, and the blade motion was driven by rigid coupling. The normal distance between the blade contact surface and the impacted surface of the bionic skin was read. A contact algorithm based on the penalty function was used to calculate the normal contact action between the blade and the bionic skin, and the tangential slip behavior of the contact interface was described using the Coulomb friction model. The minimum value between the trial tangential contact load and the product of the contact friction parameters and the normal contact load was taken as the tangential contact load. The deformation state of the biomimetic skin is determined based on the displacement changes of the finite element nodes. The stress state is obtained through hyperelastic material relations. The contact load, internal stress, and support reaction force generated by the constraint boundary of the biomimetic skin are then synthesized. The node acceleration is determined by the ratio of the resultant force to the node mass. The node velocity and node position are updated using the central difference method. When the equivalent strain reaches the failure initiation strain, damage energy consumption begins to accumulate. The failure degree is characterized by the ratio of accumulated damage energy consumption to failure evolution energy. When the failure degree reaches the complete failure state (e.g., failure degree is 1), the transmission of contact load and internal stress in the corresponding finite element mesh area is stopped. The blade motion, biomimetic skin deformation, load transmission, and crack propagation process are recorded in chronological order to generate a finite element response sequence.

[0032] It should be noted that the finite element response sequence includes the blade motion state continuously recorded according to the calculation time, as well as the deformation, stress, failure and support reaction force changes of the bionic skin during the contact process.

[0033] It should also be noted that in this embodiment, when the blade cuts into the bionic skin at high speed, the contact area will change continuously with deformation and tear expansion. The contact algorithm based on the penalty function establishes contact constraints in real time through the penetration state between the contact surfaces, so that the contact force can be dynamically updated with the contact, sliding and separation process, thereby ensuring that the load transfer process in the explicit dynamic calculation matches the actual cut contact state. The Coulomb friction model is used to describe the tangential friction when two contact surfaces slide relative to each other. The principle is to limit the magnitude of the tangential friction force by the product of the normal contact force and the coefficient of friction. In this embodiment, the Coulomb friction model is used to characterize the tangential slip and friction load transmission at the contact interface between the blade and the bionic skin, so that the tangential contact action during the cutting process can change dynamically with the normal contact state.

[0034] S1.3: Extract the support reaction time history from the finite element response sequence, identify the formation time and adjacent expansion relationship of the damaged area according to the bionic skin failure criterion, organize the support reaction time history into finite element load response characteristics, and organize the spatial location and expansion time sequence of the damaged area into finite element crack evolution characteristics.

[0035] Specifically, the support reaction forces generated along the load transfer direction at each finite element node of the bionic skin constraint position are read from the finite element response sequence according to the calculation time. The support reaction forces at the same calculation time are vectorized according to the direction of action. The first contact time of the blade is used as the time starting point, and the synthesized support reaction forces are arranged in chronological order to form the support reaction force time history. The failure degree of each finite element mesh region of the bionic skin is read at each calculation time according to the bionic skin failure criterion. The calculation time when the failure degree first reaches the complete failure state is determined as the formation time of the damage area, and the spatial position of the damage area in the coordinates of the bionic skin surface is recorded.

[0036] For damage regions that are adjacent in formation time and have connected spatial boundaries, an extended connection is established. The order of expansion of the damage region from the crack initiation position to the adjacent position is determined according to the order of formation time. The support reaction time history is organized according to the correspondence between the calculation time and the synthetic support reaction force to form finite element load response characteristics. The spatial position, formation time and adjacent expansion relationship of the damage region are organized according to the expansion order to form finite element crack evolution characteristics.

[0037] It should be noted that the finite element load response characteristics include the time history of the support reaction forces formed by the combined support reaction forces arranged continuously according to the calculation time; the finite element crack evolution characteristics include the spatial location, formation time, and adjacent expansion relationship of the damage area organized according to the expansion order.

[0038] S1.4: Using the moment of first contact of the blade as the test time reference, the test support reaction time history is extracted from the blade cut test data, and the bionic skin surface coordinates are established. The leading edge of the crack in the continuous test images is mapped to the bionic skin surface coordinates according to the crack initiation position and the progression order, generating test load response characteristics and test crack evolution characteristics.

[0039] Specifically, the sampling time of the force sensor and the frame time of the continuous test images are read from the blade cut test data. The image frame in the continuous test images where the blade contact edge first touches the bionic skin surface is located, and the time corresponding to the image frame is determined as the first contact time of the blade. The first contact time of the blade is used as the zero time to unify the conversion of the force sensor sampling time and the continuous test image time. The support reaction force sampling value after the first contact time of the blade is extracted, and the support reaction force baseline of the stable period before the first contact of the blade is subtracted (the stable period is exemplarily the period when the support reaction force of several consecutive sampling points does not show a continuous increase, continuous decrease or sudden change, and the support reaction force baseline is the average value of the support reaction force in the stable period). The support reaction force sampling value is projected onto the load transmission direction and arranged continuously according to the converted sampling time to form the test support reaction force time history.

[0040] The corner point where the length and width edges of the bionic skin clamping boundary intersect and remain unchanged during the experiment is selected as the fixed edge intersection point. The fixed edge intersection point is used as the origin of the coordinate system, and the length and width directions of the bionic skin are used as the coordinate axes, respectively. The bionic skin surface coordinates are established based on the ratio between the actual spacing of the marker points on the bionic skin surface and the image spacing. The crack boundary is extracted according to the frame time of the continuous test images. The position of the foremost point of the crack boundary along the crack extension direction is taken as the crack leading edge, and the position where the crack first appears is taken as the crack initiation position. Then, the crack leading edge in each image frame is converted to the bionic skin surface coordinates based on the marker points on the bionic skin surface, and the progression order is arranged according to the appearance time of the crack leading edge and the positional changes in adjacent image frames. The test support reaction time history is organized according to the correspondence between relative time and support reaction force to generate test load response characteristics. The surface coordinates, appearance time, and progression order of the crack leading edge are organized to generate test crack evolution characteristics.

[0041] It should be noted that the test load response characteristics include the test support reaction time history formed by continuously arranging the sampled times after subtracting the support reaction baseline and projecting it onto the load transfer direction; the test crack evolution characteristics include the crack leading edge surface coordinates, appearance time, crack initiation position and progression sequence organized according to the image frame time.

[0042] S1.5: Align the finite element load response characteristics with the test load response characteristics according to the test time reference, and project the finite element crack evolution characteristics and the test crack evolution characteristics onto the bionic skin surface coordinates. Calculate the load amplitude difference and transmission timing difference, as well as the crack leading edge position difference and expansion process difference, and obtain the load transmission residual and crack evolution residual.

[0043] Specifically, the first contact time of the blade in both the finite element load response characteristics and the test load response characteristics is converted to zero time. The finite element load response characteristics are interpolated according to the sampling time of the test load response characteristics to form two support reaction time histories at a unified time node. The amplitude difference between the finite element support reaction and the test support reaction is calculated at each unified time node. The characteristic time is extracted based on the order of load transmission, load peak appearance and load fall in the two support reaction time histories. The time offset between the corresponding characteristic time is calculated. The amplitude difference and time offset are organized into load transmission residual.

[0044] The spatial positions of the damaged area in the finite element crack evolution characteristics and the spatial positions of the crack leading edge in the experimental crack evolution characteristics are uniformly converted to biomimetic skin surface coordinates. The finite element damaged area and the experimental crack leading edge are matched according to the occurrence time, and the surface coordinate distance between the finite element crack leading edge and the experimental crack leading edge at the same time is calculated. Then, the expansion order is arranged according to the propagation distance from the crack initiation position to each crack leading edge. The occurrence time and propagation distance corresponding to the same expansion order are compared, and the surface coordinate distance, the difference in occurrence time, and the difference in propagation distance are organized into crack evolution residuals.

[0045] It should be noted that in the subsequent determination of the limited range, determination of the direction of residual change, calculation of the normalized sum and the calculation of the response sensitivity, the load amplitude difference and characteristic time offset in the load transfer residual, and the surface coordinate distance, occurrence time difference and propulsion distance difference in the crack evolution residual are dimensionless. When dimensionless, each residual term is compared with the sum of the maximum fluctuation of the same type of residual term and the corresponding acquisition error in the repeated blade cut test. The dimensionless residual terms in the load transfer residual and the dimensionless residual terms in the crack evolution residual are averaged to generate scalar load transfer residual and scalar crack evolution residual. The sum, change and direction of the subsequent load transfer residual and crack evolution residual are determined based on the corresponding scalar residual.

[0046] S2. Apply local perturbation to the parameters to be calibrated, construct a parameter residual causal graph based on the perturbation response of the load transfer residual and the crack evolution residual, divide the load action layer and the crack action layer according to the dominant residual, calibrate the parameters of the two layers in sequence, adjust the cross-layer parameters with double residual constraints, and generate the initial calibration parameter domain.

[0047] S2.1: Construct positive and negative disturbances around the current values ​​of each parameter to be calibrated, load each local disturbance combination into the blade cut simulation model according to the parameter isolation principle, and perform explicit dynamic finite element calculation while keeping the non-disturbance parameters and collision conditions unchanged to generate a parameter disturbance simulation set.

[0048] Specifically, based on the current value of each parameter to be calibrated, a local disturbance is selected according to the corresponding value range (such as 5% of the span of the value range of the parameter to be calibrated). The sum of the current value and the local disturbance is used as the positive disturbance value, and the difference between the current value and the local disturbance is used as the negative disturbance value. When the disturbance value exceeds the corresponding value range, the corresponding range boundary is used.

[0049] Local perturbation combinations are established according to the parameter isolation principle. Each local perturbation combination replaces only one parameter to be calibrated with a positive or negative perturbation value, while the other parameters to be calibrated retain their current values. The blade motion boundary, the bionic skin constraint boundary, and the contact relationship between the blade and the bionic skin remain unchanged. Each local perturbation combination is sequentially loaded into the blade laceration simulation model. Explicit dynamic finite element calculations are performed under the same collision conditions and calculation time steps. The corresponding finite element response sequence is recorded according to the local perturbation combination identifier to generate a parameter perturbation simulation set.

[0050] It should be noted that the parameter isolation principle means that only one parameter to be calibrated is changed each time, while the other parameters to be calibrated and the collision conditions remain unchanged. Forward and reverse perturbations are performed separately to obtain the influence of the target parameter to be calibrated on the finite element response sequence.

[0051] S2.2: The parameter disturbance simulation set and the blade cut test data are integrated into a time reference and spatial coordinate system. The load transfer residual and crack evolution residual under each combination of disturbance parameters are calculated. The two types of residuals corresponding to the current value of the parameter to be calibrated are used as the disturbance reference. The time series change and amplitude change of the residuals are extracted to generate a parameter dual residual response set.

[0052] Specifically, the finite element response sequences formed by the combinations of local disturbances in the parameter disturbance simulation set are compared with the blade laceration test data in terms of time reference and spatial coordinate system. The first contact time of the blade is taken as the unified time starting point. The support reaction time histories formed by the combinations of local disturbances are interpolated according to the sampling time of the test support reaction time histories. The spatial position of the damage area formed by the combinations of local disturbances is mapped to the coordinates of the biomimetic skin surface, so that the finite element load response characteristics and the test load response characteristics are at the same time node, and the finite element crack evolution characteristics and the test crack evolution characteristics are at the same spatial coordinates.

[0053] At each unified time point, the amplitude difference between the finite element support reaction force and the test support reaction force is extracted, and the time offset between the start of load transmission, the occurrence of the load peak and the load fall-off time is extracted to form the load transmission residual under each local disturbance combination. At the same time, the damaged area and the test crack leading edge are matched according to the crack expansion order, and the surface coordinate distance, the time difference of occurrence and the difference of advancement distance are extracted to form the crack evolution residual under each local disturbance combination.

[0054] The load transfer residual and crack evolution residual formed when the parameter to be calibrated is kept at its current value are used as the disturbance reference. The increase or decrease of the load amplitude difference sequence relative to the disturbance reference under positive and negative disturbances, the forward and backward shift of the load characteristic time relative to the disturbance reference are extracted one by one, and the changes of the crack leading edge position difference, the occurrence time difference and the advance distance difference relative to the disturbance reference are extracted. The positive and negative disturbance changes of the same parameter to be calibrated are paired and organized according to the disturbance direction, local disturbance amount and time order to generate a parameter double residual response set.

[0055] S2.3: Using the parameter dual residual response set, a directed interaction connection is established between the parameter to be calibrated and the two types of residuals. The causal direction is determined according to the order of the residual responses, and the interaction weights are set according to the stability of the response direction and the contribution of residual changes, thus constructing a parameter residual causal graph.

[0056] Specifically, each parameter to be calibrated in the parametric dual residual response set is set as a parameter node, and the load transfer residual and crack evolution residual are set as residual nodes respectively. For each parameter to be calibrated, the change sequence of the two types of residuals relative to the disturbance reference under positive and negative disturbances is read. The first time node in which the residuals continuously maintain the same direction of increase or decrease is taken as the starting point of the residual response. According to the time relationship of local disturbance loading first and residual response later, a directed action connection is established between the parameter to be calibrated and the corresponding residual node.

[0057] When both types of residuals generate responses, the order of action is recorded based on the starting positions of the load transfer residual and the crack evolution residual. It is then determined whether the positive and negative disturbances form opposite trends in residual increase or decrease. The ratio between the number of time nodes consistent with the more frequent increase or decrease direction and the total number of time nodes where residual changes occur is taken as the stability of the response direction. The ratio between the sum of the absolute values ​​of the residual changes caused by a single parameter to be calibrated and the sum of the absolute values ​​of the residual changes caused by all parameters to be calibrated to the same type of residual is taken as the contribution of residual change. The product of the stability of the response direction and the contribution of residual change is taken as the weight of the directed action connection. Directed action connections that have not formed a stable increase or decrease direction are not retained, and a parameter residual causal graph is generated.

[0058] The expression for the effect weight is: ; in, Indicates the first The parameter to be calibrated points to the first... The effect weight of class residuals; This indicates the number of time points that align with the direction of increase or decrease in residuals that occur most frequently. This represents the total number of time points at which residual changes occur; Indicates the first The parameters to be calibrated are at the time node. The cause of the first The change of the residual relative to the disturbance reference; Indicates the first The parameters to be calibrated are at the time node. The cause of the first The change of the residual relative to the disturbance reference; Indicates the first The parameter to be calibrated causes the first A set of time points representing residual changes; Indicates the first The parameter to be calibrated causes the first A set of time points representing residual changes; This indicates the residual type index, where the residual type is either load transfer residual or crack evolution residual. Indicates the index of the parameter to be calibrated currently being calculated; This indicates the index of the parameters to be calibrated that participate in the comparison of contributions to the variation of similar residuals. This represents the set of all indexes of parameters to be calibrated. Indicates a time point.

[0059] It should be noted that the parameter residual causal diagram organizes the response direction, response order, and action weight of the parameters to be calibrated to load transfer residuals and crack evolution residuals into a traceable causal structure, realizing the distinction between the dominant residuals and cross-layer effects of different parameters to be calibrated. This provides a basis for the division of load action layers, crack action layers, and cross-layer parameters, and improves the pertinence of layered calibration and dual residual constraint adjustment.

[0060] S2.4: Based on the influence weight of each parameter to be calibrated on the two types of residuals in the causal graph of parameter residuals, calculate the degree of dominance of the parameter to be calibrated on the load transfer residual and the crack evolution residual, divide the load action layer and the crack action layer, and determine the parameter to be calibrated that has a directed action connection on both the load transfer residual and the crack evolution residual as the cross-layer parameter, and generate the layered parameter set.

[0061] Specifically, in the parameter residual causal graph, each parameter to be calibrated points to the weight of the load transfer residual and the crack evolution residual respectively. When no directed action connection is retained, the corresponding action weight is recorded as zero. The ratio of the action weight pointing to the load transfer residual to the sum of the two types of action weights is taken as the load dominance degree, and the ratio of the action weight pointing to the crack evolution residual to the sum of the two types of action weights is taken as the crack dominance degree.

[0062] When the load dominance is greater than the crack dominance, the parameter to be calibrated is assigned to the load action layer; when the crack dominance is greater than the load dominance, the parameter to be calibrated is assigned to the crack action layer. When the two dominance levels are equal, the response start points of the load transfer residual and the crack evolution residual relative to the local disturbance are read respectively. When the load transfer residual first undergoes a stable change, the parameter to be calibrated is assigned to the load action layer; when the crack evolution residual first undergoes a stable change, the parameter to be calibrated is assigned to the crack action layer. Here, a stable change means that the load transfer residual or the crack evolution residual maintains the same direction of increase or decrease relative to the disturbance reference at multiple consecutive time nodes (such as three consecutive time nodes) without an immediate reverse change, and the first time node where the continuous change begins is taken as the residual response start point. When the two types of residuals begin to respond at the same time node, the parameter to be calibrated is included in both the load action layer and the crack action layer and marked as a cross-layer parameter. The parameter is organized according to the name of the parameter to be calibrated, the load dominance, the crack dominance, the action layer to which it belongs, and the cross-layer parameter identifier to generate a layered parameter set.

[0063] S2.5: Use the layered parameter set to traverse the load application layer parameters and filter the parameter combinations whose load transfer residuals are within the preset limit range. Within the filtered parameter combinations, perform a local search on the crack application layer parameters and correct the cross-layer parameters according to the same direction change interval of the two types of residuals. Perform intersection constraints on the parameter combinations whose two types of residuals are both within the limit range to generate the initial calibration parameter domain.

[0064] Specifically, based on the layered parameter set, the relationships between the load-bearing layer parameters, the crack-bearing layer parameters, and the cross-layer parameters are read. The limits for load transfer residuals and crack evolution residuals are set according to the fluctuation range and acquisition error of the repeated blade damage test. The value ranges of each load-bearing layer parameter are divided into continuous discrete values ​​according to the local disturbance amount, and combined to form a load-bearing layer traversal parameter combination. While keeping the current values ​​of the crack-bearing layer parameters and the cross-layer parameters unchanged, the load-bearing layer traversal parameter combination is loaded one by one into the blade damage simulation model. Explicit dynamic finite element calculation is performed, and the load transfer residuals are obtained after processing with the time reference and spatial coordinate system. Parameter combinations whose load transfer residuals are within the limited range are retained.

[0065] Using the current value of the crack action layer parameter in the retained parameter combination as the search center, the crack action layer parameter is gradually changed in both the positive and negative directions with a search step size smaller than the interval traversal interval (such as half of the local disturbance of the parameter to be calibrated). Under the condition of keeping the load action layer parameter unchanged, the crack evolution residual is obtained, and the parameter combination in which the crack evolution residual enters the limited range is retained. For parameter combinations containing cross-layer parameters, the two types of residuals corresponding to adjacent values ​​are compared sequentially in the positive and negative directions of the cross-layer parameters. The range in which the load transfer residual and the crack evolution residual decrease continuously together is taken as the same direction change interval, and the cross-layer parameter is corrected in the direction in which the two types of residuals decrease together. The parameter combinations in which the load transfer residual is within the limited range and the parameter combinations in which the crack evolution residual is within the limited range are collected respectively. The parameter combinations that simultaneously meet the requirements of the two types of residuals and the adjacent value range are retained through intersection constraints to generate the initial calibration parameter domain.

[0066] It should be noted that, based on the maximum fluctuations of load amplitude difference, characteristic time offset, crack leading edge position difference and propulsion time difference in the repeated blade cut test, and superimposed with the corresponding acquisition error, the range from zero to the obtained upper limit is set as the limit range of load transfer residual and crack evolution residual, respectively. The initial calibration parameter domain is calibrated progressively through the load application layer and the crack application layer. It retains the parameter combinations that meet the requirements of load transfer residual and crack evolution residual respectively, and uses the same direction change interval of the two types of residuals to correct the cross-layer parameters. This achieves causal decoupling and cross-constraint between parameters of different application layers, providing a reliable parameter search range for subsequent identification of confused parameters and generation of supplementary test conditions.

[0067] S3. Implement local re-perturbation on the parameters to be calibrated, extract parameter identification feature vectors, and identify confused parameter groups according to the degree of overlap and sensitivity degradation of the parameter identification feature vectors. At the same time, generate candidate supplementary test conditions within the executable operating condition range, perform finite element pre-simulation using the blade cut simulation model, calculate the response separation degree and evidence observability of the confused parameter groups, and generate the target supplementary test condition.

[0068] S3.1: Within the initial calibration parameter domain, select the parameter combination with the minimum normalized sum of load transfer residual and crack evolution residual as the re-perturbation reference. Based on the values ​​of each parameter to be calibrated in the re-perturbation reference, construct positive local re-perturbation and reverse local re-perturbation according to multiple perturbation amplitudes, and load them into the blade cut simulation model to perform explicit dynamic finite element calculation, generating a parameter re-perturbation simulation set.

[0069] Specifically, the load transfer residuals and crack evolution residuals corresponding to each parameter combination within the initial calibration parameter domain are forward normalized and summed. The parameter combination with the smallest normalized sum is used as the re-perturbation reference. For each parameter to be calibrated in the re-perturbation reference, multiple incremental perturbation amplitudes are divided according to the available distance from the current value to the upper and lower boundaries of the value interval (e.g., taking one-third, two-thirds, and all available distances in sequence). The sum of the current value and each perturbation amplitude is used as the positive local re-perturbation value, and the difference between the current value and each perturbation amplitude is used as the negative local re-perturbation value. Local re-perturbation combinations are constructed by changing only one parameter to be calibrated each time, keeping the other parameters to be calibrated and the collision conditions unchanged. Each local re-perturbation combination is sequentially loaded into the blade damage simulation model to perform explicit dynamic finite element calculations, obtaining the finite element response sequence corresponding to each local re-perturbation combination. A parameter re-perturbation simulation set is generated according to the parameter identification, perturbation direction, and perturbation amplitude.

[0070] S3.2: The parameter re-disturbance simulation set and the blade cut test data are integrated into a time base and spatial coordinate system. The time-series amplitude response of the load transfer residual and the crack evolution residual relative to the local re-disturbance is extracted and normalized and encoded to generate parameter identification feature vectors.

[0071] Specifically, the finite element response sequences corresponding to each local re-perturbation combination in the parameter re-perturbation simulation set are compared with the blade laceration test data in terms of time and spatial coordinate system. Taking the first contact time of the blade as a unified time starting point, the support reaction time histories are aligned to the same time node, and the spatial location of the damaged area is mapped to the coordinates of the biomimetic skin surface. The load transfer residuals and crack evolution residuals corresponding to each perturbation direction and amplitude are obtained. Using the two types of residuals corresponding to the re-perturbation reference as comparison benchmarks, the increase or decrease of the load amplitude difference sequence and the load start-up and load peak appearance are extracted. The forward or backward shift of the load fall-off time forms the temporal amplitude response of the load transfer residual relative to local re-perturbation. At the same time, the changes in the position difference of the crack leading edge, the difference in the advance distance, and the corresponding time difference of occurrence are extracted to form the temporal amplitude response of the crack evolution residual relative to local re-perturbation. The change with the largest absolute value among the similar responses is used as the normalization benchmark. The ratio of each change to the normalization benchmark is used as the encoding value, and the positive and negative directions of the encoding value are retained. The encoding values ​​are then arranged according to the parameter identification, perturbation direction, perturbation amplitude, and response occurrence order to generate the parameter identification feature vector.

[0072] It should be noted that the parameter identification feature vector encodes the load transfer residual and crack evolution residual changes caused by the parameter to be calibrated under different disturbance directions and amplitudes into a unified feature sequence, retaining the residual response direction, occurrence order and amplitude change law, realizing direct comparison of the action characteristics of different parameters to be calibrated, and revealing the coincidence of parameter responses and the decline in identification ability.

[0073] S3.3: Calculate the distance between the parameter identification feature vectors according to the directional angle between them and the temporal overlap relationship, and calculate the response sensitivity according to the ratio of the change in the double residual to the local reperturbation amplitude. Merge the parameters to be calibrated that are closest to each other and whose response sensitivity decreases with the increase of the local reperturbation amplitude to generate a confused parameter group.

[0074] Specifically, the parameter identification feature vectors are paired according to the identifier of the parameter to be calibrated, the direction of the disturbance, and the magnitude of the disturbance. The encoded values ​​at the same response position in the two parameter identification feature vectors are combined to form a feature vector. The directional angle is obtained by the ratio of the inner product of the two feature vectors to the product of the magnitudes of the two feature vectors. The ratio of the length of the common time interval in which the two parameter identification feature vectors have non-zero responses to the length of the merged time interval is taken as the temporal overlap. Finally, the sum of the normalized value of the directional angle and the reverse value of the temporal overlap is taken as the distance between the parameter identification feature vectors.

[0075] For each parameter to be calibrated, the ratio of the absolute value of the load transfer residual change to the sum of the absolute values ​​of the crack evolution residual change to the local re-perturbation amplitude is used as the response sensitivity, and the response sensitivities are arranged in ascending order of perturbation amplitude. The expression for response sensitivity is: ; in, Indicates the first The parameter to be calibrated is in the first... Response sensitivity at the local reperturbation amplitude level; Indicates the first The parameter to be calibrated is in the first... The change in load transfer residual caused by a local re-perturbation amplitude level; Indicates the first The parameter to be calibrated is in the first... The change in crack evolution residual caused by a local re-perturbation amplitude level; Indicates the first The parameter to be calibrated is in the first... The local re-perturbation amplitude at each local re-perturbation amplitude level; This represents the local re-perturbation magnitude hierarchy index, arranged from smallest to largest perturbation magnitude.

[0076] The minimum value is selected from the parameter identification feature vector distances between each parameter to be calibrated and the other parameters to be calibrated. When two parameters to be calibrated are the closest to each other in terms of parameter identification feature vector distance, and the response sensitivity of both parameters to be calibrated decreases continuously with the increase of local reperturbation amplitude (e.g., the response sensitivity corresponding to three consecutive increasing perturbation amplitudes decreases sequentially), a confusion connection is established between the two parameters to be calibrated. The parameters to be calibrated that are interconnected through the confusion connection are merged to generate a confusion parameter group.

[0077] S3.4: Discretely combine the collision motion states within the executable working conditions range, associate each discrete combination with the confusion parameter group, and generate candidate supplementary test working conditions and corresponding finite element pre-simulation task sets.

[0078] It should be noted that the executable operating condition range is determined by providing initial value boundaries for the blade speed, cutting angle, and contact position based on the driving capability, attitude adjustment stroke, bionic skin clamping range, and test acquisition range of the test device. Trial runs are conducted within the initial value boundaries to verify whether the blade speed can be maintained stably, whether the blade can contact the bionic skin at the set cutting angle, and whether the contact position can be maintained within the effective impact area of ​​the bionic skin. At the same time, value combinations that exclude support reaction forces exceeding the acquisition range or crack leading edges leaving the continuous test image imaging area are excluded. The collision motion state refers to the relative motion intensity, contact direction, and action position when the blade contacts the bionic skin.

[0079] Specifically, discrete values ​​are defined according to the minimum adjustment interval of the blade rotation speed, cutting angle, and contact position of the test device. Discrete values ​​of different collision motion states are cross-combined, and discrete combinations that cannot form effective contact or exceed the bearing capacity boundary of the blade structure are eliminated. The remaining discrete combinations are used as candidate supplementary test conditions. For each candidate supplementary test condition, the boundary values ​​and internal representative values ​​of each parameter to be calibrated in the confusion parameter group are selected from the initial calibration parameter domain to form a parameter combination that can reflect the parameter differences within the confusion parameter group. The values ​​of the parameters to be calibrated outside the confusion parameter group are kept unchanged. The candidate supplementary test conditions and corresponding parameter combinations are loaded into the blade cut simulation model. Explicit dynamic finite element calculation tasks are organized according to the candidate supplementary test condition identifier, confusion parameter group identifier, and parameter combination identifier to generate a finite element pre-simulation task set.

[0080] S3.5: Execute the finite element pre-test task set, calculate the response separation degree based on the distance of the parameter identification feature vectors of the confused parameter group under each candidate supplementary test condition, and calculate the evidence observability based on the coverage of the finite element load response characteristics and finite element crack evolution characteristics within the test acquisition range, and generate the candidate test condition evaluation matrix.

[0081] Specifically, the finite element pre-simulation task set is executed sequentially. The finite element response sequence formed by each parameter combination is collected according to the candidate supplementary test condition identifier and the confusion parameter group identifier. Based on the blade first contact time and the bionic skin surface coordinates, the temporal amplitude response of the load transfer residual and the crack evolution residual is normalized and encoded to form the parameter identification feature vector of the confusion parameter group under each candidate supplementary test condition. The parameter identification feature vectors of different parameters to be calibrated within the same confusion parameter group are compared pairwise. The distance between the parameter identification feature vectors is obtained based on the direction angle and the temporal coincidence relationship. The minimum parameter identification feature vector distance is used as the response separation degree of the corresponding candidate supplementary test condition.

[0082] The expression for the distance between parameter-identified feature vectors is: ; in, Indicates the first The parameter to be calibrated and the first The distance between parameter identification feature vectors of each parameter to be calibrated; Indicates the first The parameter identification feature vector of each parameter to be calibrated; Indicates the first The parameter identification feature vector of each parameter to be calibrated; This indicates the transpose of the feature vector for parameter identification; This represents the maximum directional angle in the comparison relationship of all parameter-identified feature vectors; This represents the length of the merged time interval when the two parameter-identified feature vectors produce a non-zero response. This represents the length of the common time interval during which the two parameter-identified feature vectors produce a non-zero response.

[0083] The amplitude and duration of the support reaction force in the finite element load response characteristics are compared with the range and sampling period of the force sensor. The ratio of the amount of load data falling within the force sensor's acquisition range to the total amount of finite element load response characteristics is used to characterize the coverage of the load response characteristics. At the same time, the spatial location and formation time of the damage area in the finite element crack evolution characteristics are compared with the imaging area and acquisition period of the continuous test images. The ratio of the amount of crack data falling within the acquisition range of the continuous test images to the total amount of finite element crack evolution characteristics is used to characterize the coverage of the crack evolution characteristics. The average of the two coverage values ​​is used as the observability of evidence. The candidate supplementary test conditions are used as row indices, and the response separation degree and the observability of evidence are used as evaluation columns to generate a candidate test condition evaluation matrix.

[0084] It should be noted that the response separation degree is used to quantify the degree of difference in the identification of different confused parameters under candidate supplementary test conditions, thereby screening supplementary test conditions that can enhance the parameter differentiation ability; the evidence observability degree is used to quantify the degree to which the finite element load response characteristics and finite element crack evolution characteristics fall within the actual test acquisition range, thereby ensuring that the difference response under the target supplementary test condition can be effectively acquired by the test; the candidate test condition evaluation matrix organizes the response separation degree and evidence observability of the confused parameter group under each candidate supplementary test condition into a unified dual-objective evaluation structure, which simultaneously characterizes whether the response of the parameter to be calibrated can be distinguished and whether the finite element load response characteristics and finite element crack evolution characteristics can be effectively acquired by the test, realizing the synergistic quantification of the candidate supplementary test condition identification ability and the test feasibility.

[0085] S3.6: Perform bi-objective non-dominated sorting on the candidate working condition evaluation matrix, extract candidate supplementary working conditions from the first non-dominated layer, perform positive normalization on the response separation degree and evidence observability, use the maximum value of the two normalized indices to form the bi-objective ideal point, calculate the deviation distance between each candidate supplementary working condition and the bi-objective ideal point, and determine the candidate supplementary working condition with the smallest deviation distance as the target supplementary working condition.

[0086] Specifically, the candidate supplementary test conditions in the candidate test condition evaluation matrix are compared pairwise. When the response separation degree and evidence observability of one candidate supplementary test condition are not lower than those of the other candidate supplementary test condition, and at least one of the indicators is higher than that of the other candidate supplementary test condition, the other candidate supplementary test condition is marked as the dominated test condition. The candidate supplementary test conditions that are not dominated by any candidate supplementary test condition are classified into the first non-dominated layer.

[0087] The maximum and minimum values ​​of response separation and evidence observability in the first non-dominated layer are obtained respectively. The ratio between the difference between the index value and the corresponding minimum value and the difference between the corresponding maximum and minimum value is used as the positive normalization index. The maximum value of the response separation normalization index and the maximum value of the evidence observability normalization index are used to form the dual-objective ideal point. The difference between the two normalization indices of each candidate supplementary test condition and the corresponding index of the dual-objective ideal point is calculated. The square root of the sum of the squares of the two differences is used as the deviation distance. The candidate supplementary test conditions in the first non-dominated layer are arranged in ascending order of deviation distance. The candidate supplementary test condition with the smallest deviation distance is determined as the target supplementary test condition.

[0088] S4. Conduct blade cut test according to the target test conditions, map the test data and blade cut test data to a unified double residual space, apply multi-condition cross-constraints to the initial calibration parameter domain, shrink the initial calibration parameter domain according to the difference in double residual response, generate a reliable parameter domain, load the reliable parameter domain into the blade cut simulation model, perform parameter envelope finite element calculation, and generate the blade cut response envelope.

[0089] S4.1: Conduct blade cut repair tests according to the target repair test conditions, obtain repair test data, load the parameter combination in the initial calibration parameter domain into the blade cut simulation model to perform explicit dynamic finite element calculation, and integrate the time reference and spatial coordinate system with the repair test data to obtain the repair load residual and repair crack residual.

[0090] Specifically, the test conditions were adjusted according to the blade rotation speed, cutting angle, and contact position in the target test condition, while keeping the bionic skin clamping method and test acquisition conditions unchanged. The blade cut retest was conducted, and test data were collected simultaneously. The dynamic parameters of each parameter combination within the initial calibration parameter domain and the bionic skin damage parameters were sequentially loaded into the blade cut simulation model. Explicit dynamic finite element calculations were performed while keeping the target test condition, blade motion boundary, bionic skin constraint boundary, and contact relationship unchanged, to form the finite element response sequence for each parameter combination. The finite element support reaction time history was then processed. The time history of the supplementary support reaction force is unified with the first contact time of the blade, and the finite element support reaction force time history is interpolated according to the sampling time of the supplementary support reaction force time history. The support reaction force amplitude at the same time node, as well as the time when the load starts to be transmitted, the load peak appears, and the load falls back, are compared to form the supplementary load residual. The spatial position of the damage area in the finite element response sequence and the leading edge of the supplementary crack are mapped to the coordinates of the bionic skin surface. The damage area and the leading edge of the supplementary crack are matched according to the crack expansion order. The surface coordinate position, the time of occurrence, and the propulsion distance are compared to obtain the supplementary load residual and the supplementary crack residual.

[0091] S4.2: Normalize and map the measured load residual and the measured crack residual to the load transfer residual and the crack evolution residual to generate a multi-condition dual residual matrix. Then, according to the parameter combination identifier and the condition index, matrix recombination is performed. The consistency of the change direction of the two types of residuals under different conditions for the same parameter combination and the difference of normalized amplitude are calculated to generate a cross-condition residual difference spectrum.

[0092] Specifically, the load transfer residuals and crack evolution residuals under the initial working condition and the supplementary load residuals and supplementary crack residuals under the target supplementary working condition are collected separately. Residual terms of the same type are paired according to the composition order of load amplitude difference, transfer timing difference, crack leading edge position difference, and expansion process difference. The minimum value of the same type of residual term in all parameter combinations and under both working conditions is used as the normalization lower limit, and the difference between the maximum and minimum values ​​is used as the normalization span. The ratio of the difference between the residual term and the normalization lower limit to the normalization span is used as the normalized residual term. The mean of the normalized residual terms of the same type is used as the normalized load residual and normalized crack residual under the corresponding working condition. A matrix is ​​generated by using the parameter combination identifier as the row index and the normalized load residual and normalized crack residual under the working condition index as the column index for matrix recombination, resulting in a multi-working-condition dual residual matrix.

[0093] For the same parameter combination, the normalized load residual change and the normalized crack residual change of the target test condition relative to the initial test condition are calculated respectively. When both changes are positive or both are negative, the direction of change is recorded as consistent. When the signs of the two changes are opposite, the direction of change is recorded as inconsistent. The absolute values ​​of the two changes are respectively used as the normalized amplitude difference. The consistency of the direction of change and the normalized amplitude difference are organized according to the parameter combination identifier to generate the cross-test residual difference spectrum.

[0094] S4.3: Using the cross-condition residual difference spectrum, the direction of change of the double residuals of the same parameter combination under different conditions is encoded as a constraint vector. Parameter combinations with the same direction of change of double residuals are selected according to the constraint vector, and non-dominated sorting is performed according to the normalized amplitude difference. Parameter combinations that are not simultaneously dominated by other parameter combinations are extracted to generate a cross-compatible parameter sequence.

[0095] Specifically, the normalized load residual change and normalized crack residual change of the same parameter combination between the initial working condition and the target retest working condition are extracted. An increase in residual is encoded as a positive value, a decrease in residual is encoded as a negative value, and no change in residual is encoded as a zero value. Constraint vectors are formed according to the order of normalized load residual and normalized crack residual. Parameter combinations in the constraint vector with both codes being positive or both being negative are retained, while parameter combinations with opposite signs of codes are eliminated. Pairwise comparisons are performed on the retained parameter combinations. When the difference between the normalized amplitudes of two parameters in one parameter combination is not greater than that of another parameter combination, and at least one normalized amplitude difference is less than that of another parameter combination, the other parameter combination is marked as the dominated parameter combination. Parameter combinations not simultaneously dominated by any other parameter combination are extracted and arranged in ascending order of the sum of their two normalized amplitude differences to generate a cross-compatible parameter sequence.

[0096] S4.4: Connect the cross-compatible parameter sequences according to the adjacency relationship of the parameter values, calculate the double residual gradient direction between adjacent parameter combinations, take the parameter combination with the minimum sum of normalized double residuals as the core node, extract the parameter region that is connected to the core node and has a continuous gradient direction, and perform intersection constraint with the initial calibration parameter domain to generate a reliable parameter domain.

[0097] Specifically, for each parameter combination in the cross-compatibility parameter sequence, the average of the normalized load residuals under the initial working condition and the target supplementary working condition is used to generate a cross-working condition normalized load residual, and the average of the normalized crack residuals under the initial working condition and the target supplementary working condition is used to generate a cross-working condition normalized crack residual. Each parameter combination in the cross-compatibility parameter sequence is set as a parameter node. When two parameter combinations differ in only one parameter to be calibrated and the difference in value is equal to the corresponding search step size, an adjacency connection is established between the two parameter nodes. Next, for each adjacent connection, calculate the difference between the normalized load residuals and the difference between the normalized crack residuals across the two parameter combinations. Then, form a double residual gradient vector by the ratio of the two differences to the difference in the value of the parameter to be calibrated that has changed. The positive and negative directions of the double residual gradient vector represent the directions of increase or decrease of the two types of residuals caused by the change of the parameter to be calibrated. For example, if both components are positive, it means that the two types of residuals increase synchronously when the parameter to be calibrated increases; if both are negative, it means that the two types of residuals decrease synchronously; and if the signs are opposite, it means that the two types of residuals change in opposite directions.

[0098] The expression for the double residual gradient vector is: ; in, Indicates the first The parameter combination points to the double residual gradient vector of the adjacent v-th parameter combination; and These represent the normalized load residuals across two adjacent parameter combinations, respectively. and These represent the cross-condition normalized crack residuals corresponding to two adjacent parameter combinations; and These represent the values ​​of the parameter to be calibrated that are the only ones to change in two adjacent parameter combinations; and This represents the index of two adjacent parameter combinations connected according to the adjacency relationship of parameter values; the first component of the double residual gradient vector represents the direction of change of the cross-condition normalized load residual relative to the value of the calibrated parameter, and the second component represents the direction of change of the cross-condition normalized crack residual relative to the value of the calibrated parameter.

[0099] The parameter combination that minimizes the sum of the normalized load residual and the normalized crack residual across working conditions is taken as the core node. From the core node, the search proceeds outwards along the adjacent connections. When the inner product of the gradient vectors of the adjacent double residuals is positive, the corresponding adjacent connection is retained; when the inner product is not greater than zero, the corresponding adjacent connection is truncated. The parameter nodes that can reach the core node by retaining the adjacent connections and their adjacent value ranges are collected to form a parameter region that is connected to the core node and has a continuous gradient direction. The intersection constraint between the parameter region and the initial calibration parameter domain is applied to retain the parameter combinations and their continuous value ranges that are simultaneously located in the parameter region and the initial calibration parameter domain, thus generating a reliable parameter domain.

[0100] It should be noted that the credible parameter domain, through multi-condition dual residual consistency screening and parameter adjacency gradient continuity constraint, excludes parameter combinations that can only explain a single condition or abrupt changes in residual change direction, shrinks the scattered feasible parameters into a stable range of values ​​connected around the core node, and achieves simultaneous limitation of the reliability of parameter calibration and cross-condition applicability.

[0101] S4.5: Extract the boundary parameter combination and the internal representative parameter combination from the reliable parameter domain and load them into the blade cut simulation model to perform parameter envelope finite element calculation, generate the parameter envelope response sequence of each parameter combination, calculate the upper and lower bounds of the load response according to a unified time scale, and reconstruct the spatial envelope of the crack leading edge trajectory to generate the blade cut response envelope.

[0102] Specifically, at least one parameter combination with a parameter to be calibrated located at the value boundary is selected from the trusted parameter domain as the boundary parameter combination. The trusted parameter domain is then divided into layers according to the value range of each parameter to be calibrated. Within each layer, the parameter combination with the largest normalized distance from the selected parameter combination is selected as the internal representative parameter combination. The boundary parameter combination and the internal representative parameter combination are loaded into the blade laceration simulation model, and parameter envelope finite element calculation is performed under the same collision conditions. The synthetic support reaction force of the bionic skin constraint position, the spatial location of the damage area, and the formation time of the damage area are recorded according to the calculation time to generate the parameter envelope response sequence of each parameter combination.

[0103] Using the initial contact moment of the blade as a unified time starting point, the composite support reaction force in the envelope response sequence of each parameter is interpolated according to the unified time nodes. At each unified time node, the maximum and minimum values ​​of the composite support reaction force are selected to form the upper and lower bounds of the load response. For each parameter combination, the damage area that has reached the complete failure state is mapped to the coordinates of the bionic skin surface. The order of crack propagation is determined according to the formation time of the damage area and the adjacent expansion relationship. At each unified time node, the boundary position that is at the forefront along the crack propagation direction and connected to the existing damage area is extracted as the crack leading edge position. Then, the crack leading edge positions are connected sequentially according to the unified time nodes to form the crack leading edge trajectory of the corresponding parameter combination. The crack leading edge positions of all parameter combinations at the same unified time node are collected. The outer boundary that can surround all crack leading edge positions is extracted along the coordinates of the bionic skin surface to form the crack leading edge spatial envelope section of each unified time node. Then, the crack leading edge spatial envelope sections of adjacent unified time nodes are connected according to the crack propagation order to generate the blade laceration response envelope.

[0104] It should be noted that the blade laceration response envelope includes the upper and lower bounds of the load response at a unified time point, as well as the spatial envelope of the laceration leading edge formed by the reconstruction of the laceration leading edge trajectory of each parameter combination. The blade laceration response envelope transmits the differences in parameter values ​​within the credible parameter domain as the range of load variation and the range of laceration expansion, realizing the synchronous characterization of the uncertainty of the blade laceration mechanical response and the uncertainty of damage evolution, so that the simulation evaluation is no longer limited to a single response corresponding to a single parameter combination.

[0105] In summary, this invention achieves hierarchical identification of parameters to be calibrated and shrinkage of the calibration parameter domain by constructing a parameter residual causal graph, dividing the load action layer and crack action layer according to the load transfer residual and crack evolution residual, and implementing double residual constraints on cross-layer parameters, thereby improving the interpretability of finite element simulation parameter calibration; by extracting parameter identification feature vectors to identify confused parameter groups, and combining response separation degree and evidence observability to generate target supplementary test conditions, it achieves collaborative evaluation of load response and crack propagation uncertainty, thereby improving the credibility of blade cut risk assessment.

[0106] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A simulation evaluation method for UAV propeller blade cuts based on experimental calibration, characterized in that, include: A blade laceration simulation model was established based on blade laceration test data. The dynamic parameters involved in the collision and the bionic skin damage parameters were set as parameters to be calibrated. Explicit dynamic finite element calculations were performed and finite element load response characteristics and finite element crack evolution characteristics were extracted. The time reference and spatial coordinate system were unified with the blade laceration test data to obtain load transfer residuals and crack evolution residuals. Local perturbations are applied to the calibration parameters. A cause-effect graph of parameter residuals is constructed based on the perturbation response of load transfer residuals and crack evolution residuals. The load action layer and crack action layer are divided according to the dominant residuals. The parameters of the two layers are calibrated in sequence. The cross-layer parameters are adjusted by double residual constraints to generate the initial calibration parameter domain. The calibration parameters are subjected to local re-perturbation, the parameter identification feature vector is extracted, and the confused parameter group is identified according to the degree of overlap of the parameter identification feature vector and the sensitivity degradation state. At the same time, candidate supplementary test conditions are generated within the executable working condition range. Finite element pre-run is performed using the blade cut simulation model to calculate the response separation degree and evidence observability of the confused parameter group and generate the target supplementary test condition. According to the target test conditions, the blade cut test was carried out. The test data was mapped to the blade cut test data to a unified double residual space. The initial calibration parameter domain was subjected to multi-condition cross-constraint. The initial calibration parameter domain was shrunk according to the difference in double residual response to generate a reliable parameter domain. The reliable parameter domain was loaded into the blade cut simulation model, and the parameter envelope finite element calculation was performed to generate the blade cut response envelope.

2. The method for simulating and evaluating UAV blade cuts based on experimental calibration as described in claim 1, characterized in that, The steps for performing explicit dynamic finite element calculations and extracting finite element load response characteristics and finite element crack evolution characteristics are as follows: Based on the blade cut test data, the blade and bionic skin were discretized using the finite element method. The blade was set as a rigid finite element body, and the bionic skin was set as a hyperelastic failureable finite element body. A reference point was established at the blade rotation center. The blade motion boundary, the bionic skin constraint boundary, and the contact relationship between the blade and the bionic skin were loaded into the finite element discretized structure to generate a blade cut simulation model. The dynamic parameters involved in the collision and the biomimetic skin damage parameters were extracted from the blade cut test data. The two types of parameters were set as parameters to be calibrated, and explicit dynamic finite element calculations were performed to generate finite element response sequences. The support reaction time history is extracted from the finite element response sequence, and the formation time and adjacent expansion relationship of the damaged area are identified according to the bionic skin failure criterion. The support reaction time history is organized into finite element load response characteristics, and the spatial location and expansion time sequence of the damaged area are organized into finite element crack evolution characteristics.

3. The method for simulating and evaluating UAV blade cuts based on experimental calibration as described in claim 2, characterized in that, The steps for obtaining the load transfer residual and the crack evolution residual are as follows: Using the moment of first contact of the blade as the test time reference, the test support reaction time history is extracted from the blade laceration test data, and the bionic skin surface coordinates are established. The leading edge of the crack in the continuous test images is mapped to the bionic skin surface coordinates according to the crack initiation position and the progression order, generating test load response characteristics and test crack evolution characteristics. The finite element load response characteristics and the experimental load response characteristics are time-aligned according to the experimental time base. The finite element crack evolution characteristics and the experimental crack evolution characteristics are projected onto the bionic skin surface coordinates. The differences in load amplitude and transmission timing, as well as the differences in crack leading edge position and expansion process, are calculated respectively to obtain the load transmission residual and crack evolution residual.

4. The method for simulating and evaluating UAV blade cuts based on experimental calibration as described in claim 3, characterized in that, The steps for constructing the parameter residual cause-effect graph are as follows: Forward and reverse disturbances are constructed around the current values ​​of each parameter to be calibrated. They are then loaded into the blade cut simulation model according to the parameter isolation principle. Explicit dynamic finite element calculations are performed while keeping the non-disturbance parameters and collision conditions unchanged to generate a parameter disturbance simulation set. The parameter disturbance simulation set and the blade cut test data are integrated into a time base and spatial coordinate system. The load transfer residual and crack evolution residual under each combination of disturbance parameters are calculated. The two types of residuals corresponding to the current value of the parameter to be calibrated are used as the disturbance reference. The time series change and amplitude change of the residuals are extracted to generate a parameter dual residual response set. By using the parameter dual residual response set, a directed interaction connection is established between the parameter to be calibrated and two types of residuals. The causal direction is determined according to the order of the residual responses, and the interaction weights are set according to the stability of the response direction and the contribution of residual changes, thus constructing a parameter residual causal graph.

5. The method for simulating and evaluating UAV blade cuts based on experimental calibration as described in claim 4, characterized in that, The steps for generating the initial calibration parameter domain are as follows: Based on the influence weight of each parameter to be calibrated on the two types of residuals in the parameter residual causality diagram, the degree of dominance of the parameter to be calibrated on the load transfer residual and the crack evolution residual is calculated, the load action layer and the crack action layer are divided, and the parameter to be calibrated that has a directed action connection on both the load transfer residual and the crack evolution residual is determined as the cross-layer parameter, and a layered parameter set is generated. The load-bearing layer parameters are traversed through the layered parameter set and parameter combinations whose load transfer residuals are within a preset limit are selected. Within the selected parameter combinations, the crack-bearing layer parameters are locally searched, and the cross-layer parameters are corrected based on the same direction of change of the two types of residuals. The intersection constraints are applied to parameter combinations whose residuals are both within the limit range to generate the initial calibration parameter domain.

6. The method for simulating and evaluating UAV blade cuts based on experimental calibration as described in claim 1, characterized in that, The steps for extracting parameters to identify feature vectors are as follows: Within the initial calibration parameter domain, the parameter combination with the minimum normalized sum of load transfer residual and crack evolution residual is selected as the re-perturbation reference. Around the values ​​of each parameter to be calibrated in the re-perturbation reference, positive local re-perturbation and reverse local re-perturbation are constructed according to multiple perturbation amplitudes and loaded into the blade cut simulation model to perform explicit dynamic finite element calculation, generating a parameter re-perturbation simulation set. By integrating the parameter re-perturbation simulation set with the blade cut test data into a unified time and spatial coordinate system, the temporal amplitude response of the load transfer residual and the crack evolution residual relative to the local re-perturbation is extracted and normalized to generate parameter identification feature vectors.

7. The method for simulating and evaluating UAV blade cuts based on experimental calibration as described in claim 6, characterized in that, The steps for generating candidate test cases are as follows: The distance between the parameter identification feature vectors is calculated based on the directional angle between them and the temporal overlap relationship. The response sensitivity is calculated based on the ratio of the change in the double residual to the local reperturbation amplitude. The parameters to be calibrated that are closest to each other and whose response sensitivity decreases with the increase of the local reperturbation amplitude are merged to generate a confused parameter group. Within the executable operating conditions, the collision motion states are discretely combined, and each discrete combination is associated with a set of confusion parameters to generate candidate supplementary test conditions and corresponding finite element pre-simulation task sets.

8. The method for simulating and evaluating UAV blade cuts based on experimental calibration as described in claim 7, characterized in that, The steps for generating the target supplementary measurement condition are as follows: Execute the finite element pre-simulation task set, calculate the response separation degree based on the parameter identification feature vector distance of the confused parameter group under each candidate supplementary test condition, and calculate the evidence observability based on the coverage of the finite element load response characteristics and finite element crack evolution characteristics within the test acquisition range, and generate the candidate test condition evaluation matrix. The candidate working condition evaluation matrix is ​​sorted by dual objectives and non-dominated order. Candidate supplementary working conditions are extracted from the first non-dominated layer. The response separation degree and evidence observability are positively normalized. The maximum value of the two normalized indicators is used to form the dual objective ideal point. The deviation distance between each candidate supplementary working condition and the dual objective ideal point is calculated. The candidate supplementary working condition with the smallest deviation distance is determined as the target supplementary working condition.

9. The method for simulating and evaluating UAV blade cuts based on experimental calibration as described in claim 8, characterized in that, The steps for generating the trusted parameter domain are as follows: According to the target test conditions, conduct blade cut test, obtain test data, load the parameter combination in the initial calibration parameter domain into the blade cut simulation model to perform explicit dynamic finite element calculation, and integrate the time reference and spatial coordinate system with the test data to obtain the test load residual and the test crack residual. The residuals of the supplementary load and the residuals of the supplementary crack are normalized and mapped to the residuals of load transfer and crack evolution to generate a multi-condition dual residual matrix. The matrix is ​​then reorganized according to the parameter combination identifier and the condition index. The consistency of the change direction of the two types of residuals under different conditions and the difference of normalized amplitude are calculated for the same parameter combination to generate a cross-condition residual difference spectrum. Using the cross-condition residual difference spectrum, the direction of change of the double residuals of the same parameter combination under different conditions is encoded as a constraint vector. Parameter combinations with the same direction of change of double residuals are selected according to the constraint vector, and non-dominated sorting is performed according to the normalized amplitude difference. Parameter combinations that are not simultaneously dominated by other parameter combinations are extracted to generate a cross-compatible parameter sequence. Connect the cross-compatible parameter sequences according to the adjacency relationship of the parameter values, calculate the double residual gradient direction between adjacent parameter combinations, take the parameter combination with the minimum sum of normalized double residuals as the core node, extract the parameter region that is connected to the core node and has a continuous gradient direction, and perform intersection constraint with the initial calibration parameter domain to generate a reliable parameter domain.

10. The method for simulating and evaluating UAV blade cuts based on experimental calibration as described in claim 1 or 9, characterized in that, The process of generating the blade cut response envelope involves extracting the boundary parameter combination and the internal representative parameter combination from the reliable parameter domain and loading them into the blade cut simulation model to perform parameter envelope finite element calculation, generating the parameter envelope response sequence of each parameter combination, calculating the upper and lower bounds of the load response according to a unified time scale, and reconstructing the spatial envelope of the crack leading edge trajectory to generate the blade cut response envelope.

Images