Method for analyzing gradual change reliability and global sensitivity of cabin door lifting mechanism

Abstract

The invention provides a cabin door lifting mechanism gradual change reliability and global sensitivity analysis method, and relates to the technical field of reliability and sensitivity analysis. Dynamic modeling, gradual change failure domain description, Kriging proxy modeling based on active learning and global sensitivity analysis are organically fused; the performance degradation characteristics and the reliability evolution process of the cabin door lifting mechanism in the whole life cycle can be accurately described in a multi-level and multi-dimension mode. By introducing a gradual change failure membership function, continuous evolution of structural performance from safety to failure is completely reflected, the failure probability and the variation coefficient thereof are quantitatively calculated through Monte Carlo simulation, and clear redundancy margin is reserved for safety design. Meanwhile, by means of an iterative sampling strategy of active learning-Kriging and calculation of a global sensitivity index, the local and global prediction precision and convergence efficiency of the proxy model are remarkably improved, and the contribution degree and interaction effect of each influence factor to the overall reliability are revealed.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of reliability and sensitivity analysis, in particular to a hatch lifting mechanism gradual reliability and global sensitivity analysis method. BACKGROUND

[0002] The hatch lifting mechanism of a large amphibious aircraft is the core component for load transfer and motion control during the opening and closing of the hatch, and its working state directly affects the sealing performance and operational safety of the aircraft. With long-term service, frequent opening and closing, and the effects of complex environments, the performance of the components in the lifting mechanism, such as rods, hinges, and transmission devices, may degrade, such as reduced stiffness, increased clearance, motion lag, or incoordination, which may further lead to sealing failure or control failure.

[0003] In the prior art, the research on the reliability of the lifting mechanism mainly focuses on kinematic modeling and control strategy optimization, and the reliability evaluation of the performance degradation of the components during actual service is less involved. In addition, traditional reliability analysis often uses a static binary criterion ("failure or not failure") to calculate the failure probability, ignoring the continuous degradation process of the structure performance over time or the number of uses, and cannot reveal the multi-stage failure modes and their mutual influences, making it difficult to provide effective support for the health state monitoring and reliability prediction of the hatch lifting mechanism throughout its life cycle. SUMMARY

[0004] The purpose of the present application is to provide a hatch lifting mechanism gradual reliability and global sensitivity analysis method to address the deficiencies in the above-mentioned technology.

[0005] To achieve the above-mentioned purpose, the technical solution adopted by the present application is as follows: The present application provides a hatch lifting mechanism gradual reliability and global sensitivity analysis method, the method comprising: S1: establishing a dynamics model of the hatch lifting mechanism to obtain the effective lifting height of the hatch lifting mechanism; S2: selecting the radii of the bolts at several connection points in the hatch lifting mechanism as original variables, and obtaining the original response values of the original variables according to the effective lifting height; S3: defining the original gradual failure domain of the original response values, and using a membership function to represent the original gradual failure domain, introducing standard normal variables into the original variables to obtain extended variables, and obtaining the extended response values of the extended variables according to the original response values and the membership function; S4: according to the joint probability density function of the extended variables, randomly extracting a sample pool of N extended variables; S5: Randomly extract several extended variables from the sample pool as input, and the corresponding extended response values of the several extended variables as output to form a training sample set, construct an initial Kriging model according to the training sample set, obtain the mean and variance of the extended response value corresponding to each extended variable in the sample pool, and obtain the learning function value of each extended variable in the sample pool based on the mean and variance; S6: When the learning function value of any extended variable is less than 2, the extended variable and the extended response value corresponding to the learning function value less than 2 are obtained and added to the training sample set, and the Kriging update model is reconstructed until the learning function value of each extended variable is greater than or equal to 2; when the learning function value of each extended variable is greater than or equal to 2, the gradual failure probability and the variation coefficient of the extended response value are calculated by using the Monte Carlo method; S7: Based on the Kriging update model and the membership function reconstructed by adding the extended variable and the extended response value corresponding to the last learning function value less than 2 to the training sample set, the global sensitivity index of the gradual failure probability is calculated, and the influence degree of the bolt radius of different connection positions in the original variable on the gradual failure probability is obtained.

[0006] Further, in S2, the calculation formula of the original response value is:

[0007] wherein, is the original response value, is the effective lifting height, R is the original variable, , R 1 to R n is the bolt radius of different connection positions, is the critical lifting height of the hatch lifting mechanism.

[0008] Further, in S3, the calculation formula of the extended response value is: .

[0009] wherein, is the extended response value, is the extended variable, , R n+1 is the standard normal variable, is the membership function, is the inverse function of the membership function, is the original gradual failure domain, is the cumulative distribution function of the standard normal variable.

[0010] Further, in S4, the Latin hypercube method or Sobol sequence method is used to extract the extended variable group with a sample size of N to form a sample pool, and the expression of the sample pool is:

[0011] wherein, S is the sample pool, N is the sample size of the extended variable, is the extended variable extracted for the N time.

[0012] Further, in S5, the calculation formula of the learning function value is:

[0013] wherein, is the learning function value of the extended variable extracted for the j time, is the mean value of the extended response value corresponding to the extended variable extracted for the j time, is the variance of the extended response value corresponding to the extended variable extracted for the j time.

[0014] Further, in S6, the calculation formula of the gradual failure probability and the coefficient of variation is:

[0015]

[0016] wherein, is the gradual failure probability, is the coefficient of variation, is the failure indicator function of the extended response value corresponding to the extended variable extracted for the j time, is the joint probability density function of the original variable, , R i is the bolt radius of the i th connection site.

[0017] Further, in S7, the calculation formula of the global sensitivity index is:

[0018] wherein, is the global sensitivity index, is the conditional gradual failure probability when the bolt radius of the i th connection site is fixed, is the membership function when the bolt radius of the i th connection site is fixed, The conditional joint probability density function of the original variable of the bolt radius of the first connection part, i The conditional joint probability density function of the original variable of the bolt radius of the first connection part, The conditional joint probability density function of the original variable of the bolt radius of the first connection part, i The conditional joint probability density function of the original variable of the bolt radius of the first connection part, j The conditional failure indication function of the extended response value corresponding to the extended variable of the second extraction, , R -i The conditional joint probability density function of the original variable of the bolt radius of the first connection part, i The conditional joint probability density function of the original variable of the bolt radius of the first connection part,

[0019] Further, in S4, the calculation formula of the joint probability density function of the extended variable is:

[0020] Wherein, The joint probability density function of the extended variable, The probability density function of the standard normal variable, The conditional joint probability density function of the original variable of the bolt radius of the first connection part, i The conditional joint probability density function of the original variable of the bolt radius of the first connection part.

[0021] Further, S1 comprises: S11: Import the parameterized entity model of the hatch lifting mechanism into the engineering analysis software, parse and extract the mass, moment of inertia, center of mass position and geometric size parameters of each component, and construct a physical property database; S12: Import the physical property database into the simulation platform, disassemble the hatch lifting mechanism into three sub-mechanisms of main lifting mechanism, auxiliary lifting mechanism and guide mechanism, and based on the topological relationship between the three sub-mechanisms, construct a coupled multi-body dynamics model; S13: Based on the dynamics model, analyze the influence of the bolt radius of different connection parts on the force and motion characteristics of the hatch lifting mechanism, and obtain the effective lifting height.

[0022] Further, in S2, the bolt radii of the plurality of connection parts at least include the lower lifting rocker joint bolt radius, the upper lifting rocker joint bolt radius, the C-type connecting rod second joint bolt radius, the C-type connecting rod first joint bolt radius, the spline shaft connecting rod first joint bolt radius and the spline shaft connecting rod second joint bolt radius in the main lifting mechanism.

[0023] The beneficial effects of the present application include: The application provides a cabin door lifting mechanism gradual reliability and global sensitivity analysis method, which can accurately depict the performance degradation characteristics and reliability evolution process of the cabin door lifting mechanism in the whole life cycle in multiple levels and multiple dimensions by organically integrating dynamic modeling, gradual failure domain depiction, Kriging proxy modeling based on active learning and global sensitivity analysis. Compared with the traditional static evaluation method which only relies on binary failure criterion, the application introduces a gradual failure membership function, which not only reflects the continuous evolution of the structure performance from health to failure, but also quantitatively calculates the failure probability and its coefficient of variation through Monte Carlo simulation, thereby leaving a clear redundancy margin for safety design. At the same time, with the help of the iteration sampling strategy of active learning-Kriging, the local and global prediction accuracy and convergence efficiency of the proxy model are significantly improved while the direct simulation calculation amount is greatly reduced. In addition, through the calculation of the global sensitivity index, the contribution and interaction effect of each influencing factor on the overall reliability can be quantitatively revealed, thereby providing reliable engineering decision support for structure optimization, key component reinforcement and maintenance strategy making. The technical path not only significantly improves the reliability analysis depth and efficiency of the cabin door lifting mechanism, but also provides a universal and generalizable solution for the simulation modeling and gradual reliability research of other large and complex mechanical and electrical systems. BRIEF DESCRIPTION OF DRAWINGS Figure 1 A flowchart of a cabin door lifting mechanism gradual reliability and global sensitivity analysis method provided by the application is provided. Figure 2 A structural schematic diagram of a cabin door lifting mechanism provided by the application is provided. Figure 3 A structural schematic diagram of a main lifting mechanism of a cabin door lifting mechanism provided by the application is provided. Figure 4 A structural schematic diagram of an assisted lifting mechanism of a cabin door lifting mechanism provided by the application is provided. Figure 5 A structural schematic diagram of a self-lubricating joint bearing of a cabin door lifting mechanism provided by the application is provided. Figure 6 A mechanical model of a tension-compression spring of a cabin door lifting mechanism provided by the application is provided. Figure 7 A structural schematic diagram of a guide mechanism of a cabin door lifting mechanism provided by the application is provided. Figure 8 A force change curve of an assisted spring in a simulation process of a cabin door lifting mechanism provided by the application is provided, wherein the red solid line corresponds to an auxiliary assisted spring, and the blue dashed line corresponds to a main assisted spring. Figure 9The curves showing the change in torque on the main transmission rod of a door lifting mechanism provided in this application under the conditions of having a power assist spring and not having a power assist spring are shown. The solid red line corresponds to having a power assist spring and the dashed blue line corresponds to not having a power assist spring. Figure 10 The curve showing the change in torque on the main transmission rod under varying frictional forces in a hatch lifting mechanism provided in this application is shown, where the solid red line corresponds to a friction coefficient of 0.01 and the dashed blue line corresponds to a friction coefficient of 0.02. Figure 11(a) is a schematic diagram of the lower critical lifting height of a door lifting mechanism provided in this application; Figure 11(b) is a schematic diagram of the upper critical lifting height of a door lifting mechanism provided in this application; Figure 12 The indicator function for the gradient failure domain in the method provided in this application.

[0024] Icons: 100-Main lifting mechanism; 1-Splined shaft; 2-Bracket; 3-First linkage shaft rocker arm; 4-Splined shaft connecting rod; 5-Splined shaft connecting rod first joint; 6-Splined shaft connecting rod second joint; 7-Splined shaft rocker arm; 8-C-type connecting rod; 9-C-type connecting rod first joint; 10-C-type connecting rod second joint; 11-Second linkage shaft rocker arm; 12-Upper lifting rocker arm first joint; 13-Upper lifting rocker arm; 14-Linkage shaft; 15-Upper lifting rocker arm limit bolt; 16-Upper lifting rocker arm second joint; 17-Upper cross hinge; 18-First torque tube; 19-Lower lifting rocker arm; 20-Lower cross hinge; 21-Lower lifting rocker arm joint; 200-Assisted lifting mechanism; 22-First support; 23-Assisted spring guide rod; 24-Assisted spring; 25-Assisted spring sleeve; 26-Assisted spring sleeve connector; 27-Double-headed rocker arm; 28-Support sleeve; 29-Second support; 30-Curved connecting rod; 31-Assisted lifting rocker arm; 32-Second torque tube; 33-Third torque tube; 34-Lower rocker arm; 35-Third support; 36-Auxiliary spring sleeve; 37-Auxiliary spring; 38-Auxiliary spring guide rod; 39-Third linkage shaft rocker arm; 40-First connecting piece; 41-Second connecting piece; 42-Inner ring of spherical bearing; 43-Connecting bolt; 44-Outer ring of spherical bearing; 300 - Guide mechanism; 45 - Guide groove; 46 - Guide roller. Detailed Implementation

[0025] This application provides a gradual reliability and global sensitivity analysis method for the full life-cycle performance degradation characteristics of aircraft door lifting mechanisms. It can reveal the influence mechanism of key component parameters on system reliability from multiple levels and dimensions, including dynamic modeling, failure domain characterization, surrogate model construction, and sensitivity index evaluation. For example... Figure 1 As shown, the specific steps include: S1: a cabin door lifting mechanism dynamics model is established, and an effective lifting height of the cabin door lifting mechanism is obtained, specifically including the following steps: S11: a parameterized entity model of the cabin door lifting mechanism is imported into an engineering analysis software from a three-dimensional modeling software, and mass, moment of inertia, mass center position and geometric size parameters of each component are parsed and extracted to construct a physical property database.

[0026] Specifically, a three-dimensional modeling software UG is used to construct the parameterized entity model of the cabin door lifting mechanism. After the parameterized entity model of the cabin door lifting mechanism is constructed, the parameterized entity model of the cabin door lifting mechanism is exported to Parasolid intermediate format. The cabin door lifting mechanism parameterized entity model in Parasolid intermediate format is imported into the engineering analysis software CATIA. In the import process, the CATIA software automatically parses the intermediate format file, identifies and restores the geometric information and assembly relationship of each component of the model, and constructs the physical property database. After the import is completed, the three-dimensional model of the cabin door lifting mechanism is checked in the CATIA software to ensure that the model is imported accurately. Therefore, S11 can realize the standardized acquisition and cross-platform transmission of the physical property data of the cabin door lifting mechanism, and provide accurate initial parameter set for establishing the cabin door lifting mechanism dynamics model.

[0027] S12: The physical property database is imported into the ADAMS simulation platform, the cabin door lifting mechanism is disassembled into a main lifting mechanism 100, an auxiliary lifting mechanism 200 and a guide mechanism 300, and a coupled multi-body dynamics model is constructed based on the topological relationship among the three sub-mechanisms.

[0028] Specifically, as Figure 2 and Figure 3As shown, the main lifting mechanism 100 includes a spline shaft 1, a bracket 2, a first linkage shaft rocker arm 3, a spline shaft connecting rod 4, a spline shaft rocker arm 7, a C-shaped connecting rod 8, a second linkage shaft rocker arm 11, an upper lifting rocker arm 13, a linkage shaft 14, an upper lifting rocker arm limit bolt 15, an upper cross hinge 17, a first torsion tube 18, a lower lifting rocker arm 19 and a lower cross hinge 20. The spline shaft 1 is fixedly connected with the cabin door inner handle and the cabin door outer handle driving connecting rod. When the cabin door inner and outer handles are rotated, the movement is transmitted to the C-shaped connecting rod 8 through the spline shaft 1, the spline shaft connecting rod 4, the linkage shaft 14, the first linkage shaft rocker arm 3 and the second linkage shaft rocker arm 11. The C-shaped connecting rod 8 rotates with the linkage shaft 14 and generates a pressing force on the upper lifting rocker arm 13. With the upper cross hinge 17 and the lower cross hinge 20 on the main hinge as the fulcrum, the cabin door is driven to make upward lifting movement. The bolts are used for hinging between the components in the main lifting mechanism 100, and the friction and gap between the hinged parts are the key to affect the performance of the cabin door lifting mechanism. Therefore, the bolt is taken as the research object for parameterization processing, and the rotary pair constraint between the bolt and the connecting piece is replaced with contact and plane pair constraint. When the multi-body kinematics simulation of the main lifting mechanism 100 is carried out, the constraint relationship between the components of the main lifting mechanism 100 is shown in Table 1.

[0029] Table 1 Constraint relationship of kinematic pairs of the main lifting mechanism 100

[0030] As Figure 4As shown, the power-assisted lifting mechanism 200 includes a power-assisted spring lifting mechanism and an auxiliary spring lifting mechanism. The power-assisted spring lifting mechanism includes a first support 22, a power-assisted spring guide rod 23, a power-assisted spring 24, a power-assisted spring guide rod sleeve 25, a double-end rocker arm 27, a support sleeve 28, a second support 29, a curved link 30, a power-assisted lifting rocker arm 31, a second torsion tube 32, a third torsion tube 33, and a lower lifting rocker arm 34. The power-assisted lifting mechanism 200 is used to provide power assistance for the lifting movement of the hatch door. Compared with the power-assisted spring lifting mechanism, the auxiliary spring lifting mechanism is relatively simple and includes a third support 35, an auxiliary spring sleeve 36, an auxiliary spring 37, an auxiliary spring guide rod 38, and a third linkage shaft rocker arm 39. When the hatch door starts to move upward under the action of the main lifting mechanism 100, the rotation of the linkage shaft rocker arm compresses the auxiliary spring 37 and stores energy for the triggering of the power-assisted lifting mechanism 200 and the lifting movement of the hatch door. During the upward movement of the hatch door, the lifting of the tail of the lower lifting rocker arm 19 drives the movement of the torsion tube, further drives the rotation of the power-assisted lifting rocker arm 31 fixedly connected to the torsion tube, and transmits the rotation trend of the power-assisted lifting rocker arm 31 to the power-assisted spring 24 through the curved link 30, the double-end rocker arm 27, and the power-assisted spring guide rod 23, thereby smoothly activating the power-assisted spring 24 in the compressed state and making the energy stored in the power-assisted spring 24 serve the lifting movement of the hatch door without reservation. Compared with the power-assisted spring lifting mechanism, although the auxiliary spring lifting mechanism is relatively simple in structure, it can prevent the power-assisted spring lifting mechanism from being triggered by mistake and effectively improve the safety of the auxiliary lifting mechanism, because the power-assisted spring lifting mechanism can only act after the auxiliary spring lifting mechanism is completely triggered. Figure 5 As shown, the constraint mode of the joint bearing connecting two components is referred to as a joint bearing, and the specific kinematic pair of the self-lubricating joint bearing is shown in Table 2, and the kinematic constraint relationship between the components of the power-assisted lifting mechanism 200 is shown in Table 3.

[0031] Table 2 Kinematic pair constraint relationship of self-lubricating joint bearing

[0032] Table 3 Kinematic constraint relationship between components of power-assisted lifting mechanism 200

[0033] The contact constraint between the first connecting member 40 and the second connecting member 41 can effectively prevent the mold penetration phenomenon caused by excessive movement amplitude, and restore the actual movement process to the greatest extent. The parameters of the booster spring 24 and the auxiliary spring 37, such as the elastic coefficient and the preload, need to be determined when performing multi-body dynamics simulation as key components. In the View interface of the ADAMS software, the spring is established based on the points on the two parts, and the spring force and damping force are calculated based on the distance between the positions of the two parts and the rate of change thereof. The positive action force of the spring tends to push the two objects apart, while the negative action force tends to pull the two objects closer together, so the positive and negative of the spring force represent different working states of the spring. When the spring force is positive, it indicates that the spring is in a compression state, and when the spring force is negative, it indicates that the spring is in a stretching state. The mechanical model of the tension-compression spring damper is as shown in Figure 6 . By using the tension-compression spring damper tool, a pair of spring forces with resistance is applied between the two components, and the solver Solver of the ADAMS software calculates the spring force by the following formula:

[0034] wherein, P is the spring force, C is the viscous damping coefficient, dQ / dt is the relative velocity of the two ends of the spring, Q is the initial length of the spring, K is the spring stiffness coefficient, Q is the length of the spring under the pre-action force, P is the pre-action force of the spring. In the booster lifting mechanism 200, the function of the spring is to provide an action force to the connecting objects at both ends, so the viscous damping coefficient is set to C =0, that is, the spring damper is regarded as a pure spring without damping. According to the provided parameter information of the booster spring and the auxiliary spring, the elastic stiffness coefficient, the pre-action force, and the length information under the preload of SPRING_1 and SPRING_2 are shown in Table 4.

[0035] Table 4 Spring parameters

[0036] The guide mechanism 300 includes a guide groove 45 and a guide roller 46, as shown in Figure 7The guide rollers 46 are located on both sides of the door frame, and the guide grooves 45 are fixed on the fuselage and do not move with the door. During the lifting of the door, the guide rollers 46 move along the predetermined track in the guide grooves 45 until the guide rollers 46 leave the guide grooves 45, the door rotates transversely along the main hinge under the condition of transverse force, and the final process of opening the door is completed. Therefore, the guide mechanism 300 in the door lifting mechanism has three functions: first, when the door is in the closed state, the guide rollers 46 connected with the door are located in the guide grooves 45 fixed on the fuselage, thereby limiting the transverse displacement of the door and fixing the door; second, when the door is in the lifting state, the guide mechanism 300 limits the movement direction of the door, so that the door can only move along the guide grooves 45. Third, when the aircraft is in the high-altitude flight state, the aerodynamic load of the door is transmitted to the fuselage through the guide mechanism 300. The longitudinal distance between the initial position of the guide roller 46 in the guide mechanism 300 and the slot of the guide groove 45 is defined as the critical lifting height of the door lifting mechanism. When the effective lifting height of the door is lower than the critical lifting height, the guide roller 46 cannot smoothly move out of or slide into the guide groove 45, that is, the door lifting or closing cannot be normally performed, and the door is stuck.

[0037] S12, by adopting the modular modeling idea, integrating multiple sub-mechanisms in a unified dynamic framework, and then realizing overall collaborative analysis through joint simulation, the limitations of the prior art on collaborative simulation of complex systems can be broken through, unified construction of a multi-mechanism coupled dynamic model can be realized, the collaborative motion behavior of each sub-mechanism under actual working conditions can be accurately simulated, the construction efficiency and accuracy of the model are improved, and the motion characteristics and mutual coupling relationship of each sub-mechanism can be finely described.

[0038] S13: Multi-mechanism coupled motion characteristic analysis: a coupling correlation model of the bolt radius of the connection position in the door lifting mechanism and the force characteristic of the main force transmission rod of the door lifting mechanism is constructed, and the effective lifting height of the door lifting mechanism is obtained.

[0039] In addition, a coupling correlation model of the elastic force change of the assisting spring 24 and the friction coefficient on the force characteristic of the main force transmission rod is also needed to be constructed, and comprehensive influencing factors are needed to be considered to accurately obtain the effective lifting height of the door lifting mechanism.

[0040] The motion law of the door lifting mechanism simulation model is established according to the motion parameters provided in the CATIA three-dimensional entity model and the working stroke requirement of the pin. Since the actual door opening process is relatively fast, about two seconds, the acceleration of each component of the door lifting mechanism causes the torque acting on the main force transmission rod to deviate, therefore, the simulation time of the model is pulled to be long enough, so that the influence caused by the acceleration of each component is effectively removed.

[0041] Two assist springs are configured in the hatch lifting mechanism, which are main assist spring and auxiliary assist spring. The spring forces provided by the main assist spring and the auxiliary assist spring during the mechanism lifting process are shown in Figure 8 The main assist spring is used to trigger the working state of the auxiliary assist spring, and its own output is small, while the auxiliary assist spring provides the main lifting force during the mechanism movement, which can lift the hatch and effectively maintain the stability of the hatch in a certain height range.

[0042] Figure 9 The force moment changes of the main force transmission rod with and without assist springs are shown. The red solid line represents the force moment curve when the spring is configured, and the blue dotted line corresponds to the case without spring. When the simulation time reaches 70 seconds (corresponding to the door handle rotating to 70 degrees), the assist spring starts to work, and the main force transmission rod moment reaches the peak value at about 90 degrees, i.e. the maximum human power required for opening operation, and the assist effect is most significant after about 120 degrees, and the door body can maintain the open state. In the case without assist spring, as the hatch is lifted, the main force transmission rod moment continues to rise, and reaches the maximum value (about 37000 N•mm) at a certain height. If the external applied force is removed at this time, the hatch will fall due to its own gravity, and the opening and maintaining functions cannot be realized.

[0043] In addition, the friction at each hinge position and the sequence controller in the hatch lifting mechanism also has a significant impact on the motion performance. Figure 10 The force moment changes of the main force transmission rod under different friction coefficients (0.01 and 0.02) are shown. The results show that as the friction coefficient increases, the force moment of the main force transmission rod increases significantly, i.e. the operator needs to exert more torque to complete the hatch lifting, which is highly consistent with the actual operation experience. This shows that the friction has an important influence on the performance of the main force transmission rod, and needs to be fully considered in the design and analysis to ensure that the model truly reflects the actual motion behavior of the hatch lifting mechanism.

[0044] By the above kinetic model and the analysis of the motion characteristics, the effective lifting height of the hatch lifting mechanism can be accurately calculated. The effective lifting height should ensure that the minimum requirement height of the hatch opening and closing under extreme conditions - the critical lifting height of the hatch lifting mechanism. As shown in FIG. 11(a), during the closing process of the hatch, when the center point position of the guide roller 46 is lower than the highest position of the guide groove 45, that is, the effective lifting height is lower than the lower critical lifting height, the hatch cannot be normally closed, that is, the hatch lifting mechanism fails. During the opening or closing process of the hatch, when the center point position of the guide roller 46 is at the position shown in FIG. 11(b), at this time, the lowest point position of the guide roller 46 is higher than the wall surface of the narrowest part of the guide groove 45, that is, the effective lifting height is greater than the upper critical lifting height, the hatch can be normally opened and closed, that is, the hatch lifting mechanism is effective. When the guide roller 46 is between the two, there is a certain possibility that the hatch cannot be closed or opened, that is, the mechanism is between failure and effectiveness, so it is necessary to describe this problem by using the gradual reliability.

[0045] S2: Select the bolt radius of several connection positions in the hatch lifting mechanism as the original variable, and obtain the original response value of the original variable according to the effective lifting height.

[0046] As can be seen from S1, the bolts are mainly used for hinging between the components in the hatch lifting mechanism, and the friction and gap between the hinges are the key factors affecting the effective lifting height. Among them, the bolt radius is the key factor affecting the friction and gap between the hinges, therefore, the friction and gap size at the bolt can be characterized by the size of the bolt radius. The bolt radius of several connection positions in the hatch lifting mechanism is selected as the original variable, and the original response value (i.e. the effective lifting height) corresponding to each original variable is obtained by combining the dynamic simulation results, to study the influence of the bolt radius on the gradual reliability of the hatch lifting mechanism. S2 not only determines the functional relationship between the parameters and the performance response, but also lays a quantitative foundation for subsequent construction of the gradual failure domain. The calculation formula of the original response value is:

[0047] Among them, is the original response value, is the effective lifting height, which is obtained by ADAMS simulation, R is the original variable, , R 1 to R n are the bolt radii of different connection positions, is the critical lifting height of the hatch lifting mechanism, specifically the lower critical lifting height, which is given by ADAMS modeling.

[0048] S3: Define the original gradual failure domain of the original response value, and use the membership function to represent the original gradual failure domain. Introduce the standard normal variable in the original variable to obtain the extended variable. According to the original response value and the membership function, the extended response value of the extended variable is obtained.

[0049] In view of the stiffness attenuation and the increase of the gap between the fittings that may occur in long-term service, the gradual failure domain of the original response value is defined and represented by a membership function. Specifically, the original variable is mapped to the standard normal space by normal transformation to obtain the extended variable. Then, according to the design response and the pre-set threshold interval, the membership function is constructed to upgrade the traditional binary failure criterion to a continuous gradual failure measurement. Through the fuzzy representation, the transition process and uncertainty of the structure performance from safety to failure can be more truly reflected. The threshold interval of the membership function can be automatically adjusted based on the engineering tolerance requirements or the pre-set safety margin, and piecewise linear, cubic Hermite interpolation or Logistic S-shaped function can be selected to balance the smoothness and boundary sensitivity of the model. The calculation formula of the extended response value is: .

[0050] wherein, is the extended response value, is the extended variable, , R n+1 is the standard normal variable, is the membership function, is the inverse function of the membership function, is the original gradual failure domain, is the cumulative distribution function of the standard normal variable.

[0051] S4: According to the joint probability density function of the extended variable, a sample pool with a sample capacity of N is randomly extracted.

[0052] Specifically, after defining the extended variables and characterizing the gradual failure domain, a large number of samples can be randomly drawn based on the joint probability density function of these extended variables using at least one of the Latin hypersolution method and the Sobol sequence method. This constructs a sample pool that can fully represent the input uncertainty and is efficiently used for subsequent surrogate model training. Since the Latin hypersolution method emphasizes marginal uniformity and the Sobol sequence method strengthens multidimensional low-discrepancy, these two sampling methods can be flexibly selected or combined according to problem requirements to further reduce sampling redundancy in high-dimensional space. They also improve the training efficiency and prediction accuracy of the surrogate model while ensuring sample uniformity. Based on these high-quality sample pools, subsequent Kriging iterative sampling and Monte Carlo reliability analysis can achieve high-precision assessment of the full life-cycle performance degradation and failure probability of the hatch lifting mechanism with minimal computational cost, providing solid data support for structural optimization and maintenance decisions. The formula for calculating the joint probability density function of the extended variables is:

[0053] in, For the joint probability density function of the extended variables, Let be the probability density function of a standard normal variable. For the first i The edge density function of the bolt radius at each connection point.

[0054] The expression for the sample pool is:

[0055] in, S For the sample pool, N To expand the sample size of the variables, For the first N The expanded variables extracted in this step.

[0056] S5: Randomly select several extended variables from the sample pool as inputs and use the extended response values ​​corresponding to these extended variables as outputs to form a training sample set. Construct the initial Kriging model based on the training sample set, obtain the mean and variance of the extended response values ​​corresponding to each extended variable in the sample pool, and obtain the learning function value of each extended variable in the sample pool based on the mean and variance.

[0057] After the completion of the sample pool construction, the method enters the existing data-based surrogate model training phase. First, a number of extension variables are randomly selected from the sample pool as inputs, and the extension response values of these extension variables obtained in the real dynamics simulation or existing data are simultaneously obtained as the corresponding outputs to form a training sample set. In specific implementation, through a simple random sampling function (such as numpy.random.choice in Python), 1 N 1 N 1 N ) extension variables are selected, and their input-output pairs are arranged into a training data format for subsequent model calibration. Through this design, the diversity of the training sample set is ensured, and sufficient response information is provided for the surrogate model to cover different uncertainty situations. The expression of the training sample set is:

[0058] wherein, D T is the training sample set, A is the 1 N 1 extension variable, , B is the A corresponding extension response value, .

[0059] Based on the training sample set, a Kriging initial model is constructed. The Kriging initial model can not only output the predicted mean value for any extension variable, but also give the corresponding prediction variance, thereby fully characterizing the confidence level of the model for different inputs. Specifically, at the software implementation level, a mature library (such as GaussianProcessRegressor in scikit-learn of Python or fitrgp of MATLAB) is called for hyperparameter optimization and model training. After training, each extension variable in the sample pool is input into the Kriging initial model to quickly obtain the mean and variance of the corresponding extension response value. Subsequently, based on the mean and variance, the learning function value of each extension variable in the sample pool is constructed. The calculation formula of the learning function value is:

[0060] wherein, is the learning function value of the j th extension variable, is the mean value of the extension response value corresponding to the j th extension variable, is the variance of the extension response value corresponding to the j th extension variable.

[0061] By combining the Kriging surrogate model and the learning function, the application can greatly reduce the dependence on expensive original simulation while ensuring global prediction accuracy. By sorting the sample pool using the learning function value, the key points with the most uncertainty or the most likely to approach the failure threshold can be accurately identified, providing direction for subsequent iteration and sample updating. This strategy not only significantly improves the local and global convergence speed of the surrogate model, but also ensures efficient coverage of the limit state region during reliability assessment, thereby accurately depicting the performance degradation characteristics and failure probability of the hatch lifting mechanism within the limited computational budget.

[0062] S6: When the learning function value of any extended variable is less than 2, the extended variable and the extended response value corresponding to the learning function value less than 2 are obtained and added to the training sample set, and the Kriging update model is reconstructed until the learning function value of each extended variable is greater than or equal to 2; when the learning function value of each extended variable is greater than or equal to 2, the gradual failure probability and the coefficient of variation of the extended response value are calculated using the Monte Carlo method.

[0063] In the active learning-Kriging cycle sampling stage, the model first calculates the learning function value of each extended variable in the sample pool, and discriminates the learning function value. When the learning function value of any extended variable is less than 2, it indicates that the prediction mean and the failure threshold are close or the prediction uncertainty is too large, and the reliability of the model in this area is not enough. At this time, the extended variable and its extended response value corresponding to the learning function value less than 2 are extracted and added to the training sample set; then, the Kriging update model is reconstructed based on the updated training set to incorporate the latest high-value information. By iteratively sampling the most representative "difficult" points, the mean and variance estimates of the surrogate model in its prediction interval are continuously optimized until the learning function value of all extended variables is greater than 2, indicating that the model's prediction of the area near the failure threshold has reached sufficient confidence.

[0064] After completing the active learning iteration, a Kriging surrogate model with high accuracy and reliability for the entire extended variable space is obtained. At this time, enter the Monte Carlo simulation stage, randomly sample a large number of extended variables, and quickly predict the extended response values of each sample point through the surrogate model. Based on these extended response values, the gradual failure probability is obtained by statistical failure membership function, and the coefficient of variation is calculated to measure the stability and dispersion of the probability estimate. This two-step method avoids blind multiple calls to the original high-cost simulation, ensuring the accuracy of failure probability evaluation while significantly reducing the overall computational load. The formulas for calculating the gradual failure probability and the coefficient of variation are:

[0065]

[0066] wherein, is the gradual failure probability, is the coefficient of variation, is the failure indicator function of the extended response value corresponding to the extended variable of the j is the joint probability density function of the original variables, is the failure indicator function of the extended response value corresponding to the extended variable of the , R i is the bolt radius of the i is the bolt radius of the

[0067] By the ordered iterative sampling and Monte Carlo fusion strategy, the gradual failure characteristics of the hatch lifting mechanism in the whole life cycle can be accurately described under the condition of limited simulation resources, and the failure probability and coefficient of variation with statistical confidence can be obtained. This technology not only ensures the efficient convergence of the surrogate model in the global range, but also provides high-quality input data for subsequent global sensitivity analysis, thereby providing solid and reliable quantitative support for structural optimization, safety margin design and maintenance decision.

[0068] S7: Based on the Kriging update model and the membership function reconstructed by adding the extended variable and the extended response value corresponding to the last learning function value less than 2 to the training sample set, the global sensitivity index of the gradual failure probability is calculated, and the influence degree of the bolt radius of different connection positions in the original variables on the gradual failure probability is obtained.

[0069] The calculation formula of the global sensitivity index is:

[0070] wherein, is the global sensitivity index, is the conditional gradual failure probability when the bolt radius of the i is the bolt radius of the is the membership function when the bolt radius of the i is the bolt radius of the is the conditional joint probability density function of the original variables without the bolt radius of the i is the bolt radius of the is the conditional failure indicator function of the extended response value corresponding to the extended variable of the i is the bolt radius of the j is the bolt radius of the , R -i is the bolt radius of the i is the bolt radius of the

[0071] Through sensitivity analysis, the influence degree of the bolt radius of each connection position on the gradual failure probability can be clearly sorted, and the key sensitive variables and their interaction mechanism can be identified, which provides a quantitative basis for structure optimization and design redundancy. Further, these global sensitivity indicators not only support reinforcement or tolerance tightening for the most sensitive bolts, but also can be used to develop periodic detection and maintenance strategies, allocating limited resources to components that contribute most to system reliability, thereby significantly improving the safety and reliability of the cabin door lifting mechanism under complex service conditions.

[0072] In this embodiment, first, according to the dynamic model and simulation results of the cabin door lifting mechanism, the quantitative corresponding relationship between the effective lifting height and the reliability of the door action can be determined: as shown in the following table, when the effective lifting height is lower than the lower critical lifting height Figure 12 =27.251mm, the door cannot be normally closed; when the effective lifting height is higher than the upper critical lifting height =35.2mm, the door can be normally opened and closed; when the effective lifting height is between the two, there is a certain possibility that the door cannot be closed or opened. The variability in this interval is the focus of the gradual reliability analysis, and by improving the traditional binary failure to continuous gradual failure criterion, the failure probability of the mechanism under the tolerance fluctuation and service degradation can be more truly reflected.

[0073] For the above risk interval, this embodiment selects the bolt radius of six key connection positions in the cabin door lifting mechanism as the original random variables: the bolt radius of the lower lifting rocker joint 21, the bolt radius of the upper lifting rocker second joint 16, the bolt radius of the C-type connecting rod second joint 10, the bolt radius of the C-type connecting rod first joint 9, the bolt radius of the spline shaft connecting rod first joint 5, and the bolt radius of the spline shaft connecting rod second joint 6 as the original variables. The corresponding original response value is represented as:

[0074] wherein, R 1 is the bolt radius of the lower lifting rocker joint 21, R 2 is the bolt radius of the upper lifting rocker second joint 16, R 3 is the bolt radius of the C-type connecting rod second joint 10, R 4 is the bolt radius of the C-type connecting rod first joint 9, R 5 is the bolt radius of the spline shaft connecting rod first joint 5, R 6 is the bolt radius of the spline shaft connecting rod second joint 6.

[0075] ​Regarding the uniform distribution characteristics of manufacturing tolerances, the six bolt radii are regarded as independent uniform distribution random variables, whose distribution parameters are shown in Table 5, and the extended variable space is constructed according to the distribution parameters of the six bolt radii. Subsequently, the steps of Latin hypercube or Sobol sequence sampling, active learning sampling and reconstruction of Kriging surrogate model, and Monte Carlo sampling to estimate the progressive failure probability and coefficient of variation are completed in turn. The ADAMS dynamic simulation model is called 918 times in the whole process, effectively balancing the calculation accuracy and simulation overhead.

[0076] Table 5 Distribution parameters of different bolt radii

[0077] The calculated progressive failure probability of the hatch lifting mechanism is 5.44×10 -5 , and the corresponding coefficient of variation is 4.29%. Table 6 shows the results of global sensitivity analysis of the progressive reliability of the hatch lifting mechanism. It can be seen that, in the global sensitivity analysis, the bolt radii of the second joint 10 of the C-shaped connecting rod and the first joint 9 of the C-shaped connecting rod have the most significant influence on the progressive failure probability of the hatch lifting mechanism, indicating that the gap at this connection position plays a dominant role in the reliability of the hatch operation; the bolt radii of the second joint 6 of the spline shaft connecting rod, the first joint 5 of the spline shaft connecting rod, and the second joint 16 of the upper lifting rocker arm have a lesser influence on the progressive failure probability of the hatch lifting system, while the bolt radius of the lower lifting rocker arm joint 21 has the least influence on the progressive failure probability of the hatch lifting system, indicating that the gap at this position has little effect on the reliability of the hatch lifting mechanism.

[0078] Table 6 Results of global sensitivity analysis of the progressive reliability of the hatch lifting mechanism

[0079] This embodiment not only verifies the feasibility and efficiency of the proposed method in real engineering cases, but also provides clear quantitative basis for structure optimization and tolerance control through quantitative failure probability and global sensitivity indicators. For high sensitivity positions (such as the C-shaped connecting rod joint), appropriate tightening of tolerances or addition of detection redundancy can be made in actual design to maximize the sealing performance and operational safety of the hatch lifting mechanism under long-term service and complex working conditions.

[0080] The above only describes the preferred embodiments of the present application and is not intended to limit the present application. Those skilled in the art can make various modifications and changes to the present application. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present application shall be included in the protection scope of the present application.

Claims

1. A method for analyzing the gradual reliability and global sensitivity of a hatch lifting mechanism, characterized in that, The method includes: S1: Establish a dynamic model of the hatch lifting mechanism and obtain the effective lifting height of the hatch lifting mechanism; S2: Select the bolt radii of several connecting parts in the door lifting mechanism as the original variables, and obtain the original response values ​​of the original variables according to the effective lifting height; S3: Define the original gradual failure domain of the original response value, and use the membership function to characterize the original gradual failure domain. Introduce a standard normal variable into the original variable to obtain the extended variable. Based on the original response value and the membership function, obtain the extended response value of the extended variable. S4: Based on the joint probability density function of the extended variables, randomly select extended variables with a sample size of N to form a sample pool; S5: Randomly select several extended variables from the sample pool as inputs and take the extended response values ​​corresponding to the several extended variables as outputs to form a training sample set. Construct the Kriging initial model based on the training sample set, obtain the mean and variance of the extended response values ​​corresponding to each extended variable in the sample pool, and obtain the learning function value of each extended variable in the sample pool based on the mean and variance. S6: When the learning function value of any extended variable in the sample pool is less than 2, obtain and add the extended variable and extended response value corresponding to the learning function value less than 2 to the training sample set, and reconstruct the Kriging update model until the learning function value of each extended variable is greater than or equal to 2; when the learning function value of each extended variable is greater than or equal to 2, use the Monte Carlo method to calculate the gradual failure probability and coefficient of variation of the extended response value. S7: Add the extended variable corresponding to the last learning function value less than 2 and the extended response value to the Kriging update model and the membership function reconstructed in the training sample set, calculate the global sensitivity index of the gradual failure probability, and obtain the degree of influence of the bolt radius of different connection parts in the original variables on the gradual failure probability.

2. The method according to claim 1, characterized in that, In step S2, the formula for calculating the original response value is: in, The original response value, For the aforementioned effective height increase, R For original variables, , R 1 to R n These represent the bolt radii for different connection points. This is the critical lifting height of the door lifting mechanism.

3. The method according to claim 2, characterized in that, In step S3, the formula for calculating the extended response value is: 。 in, For the extended response value, To expand variables, , R n+1 For a standard normal variable, The membership function is... It is the inverse function of the membership function. This refers to the original gradient failure domain. This is the cumulative distribution function of a standard normal variable.

4. The method according to claim 3, characterized in that, In step S4, an expanded variable of size N is drawn to form a sample pool using the Latin hypercube method or the Sobol sequence method. The expression for the sample pool is: in, S For the sample pool, N To expand the sample size of the variables, For the first N The expanded variables extracted in this step.

5. The method according to claim 4, characterized in that, In S5, the formula for calculating the learning function value is: in, For the first j The learning function value of the expanded variable extracted in the second step. For the first j The mean of the expanded response values ​​corresponding to the expanded variables extracted each time. For the first j The variance of the expanded response values ​​corresponding to the expanded variables extracted each time.

6. The method according to claim 5, characterized in that, In step S6, the formulas for calculating the gradual failure probability and the coefficient of variation are as follows: in, The gradient failure probability is... The coefficient of variation is... For the first j The failure indication function for the extended response value corresponding to the next extracted extended variable. Let be the joint probability density function of the original variables. , R i For the first i Bolt radius at each connection point.

7. The method according to claim 6, characterized in that, In step S7, the formula for calculating the global sensitivity index is: in, This refers to the global sensitivity index. For the first i The probability of gradual failure under fixed bolt radii at a connection point. For the first i The membership function when the bolt radius of each connection part is fixed. For excluding the first i The conditional joint probability density function of the original variables of the bolt radius at each connection point. For the first i When the bolt radius of the first connection part is fixed, the first j The conditional failure indicator function for the extended response value corresponding to the next extracted extended variable. , R -i For the - i Bolt radius at each connection point.

8. The method according to claim 6, characterized in that, In S4, the formula for calculating the joint probability density function of the extended variables is: in, For the joint probability density function of the extended variables, Let be the probability density function of a standard normal variable. For the first i The edge density function of the bolt radius at each connection point.

9. The method according to any one of claims 1 to 8, characterized in that, S1 includes: S11: Import the parametric solid model of the door lifting mechanism into the engineering analysis software, analyze and extract the mass, moment of inertia, center of mass position and geometric dimension parameters of each component, and construct a physical property database; S12: Import the physical property database into the simulation platform, disassemble the hatch lifting mechanism into three sub-mechanisms: main lifting mechanism, auxiliary lifting mechanism, and guide mechanism, and construct a coupled multibody dynamics model based on the topological relationship between the three sub-mechanisms; S13: Based on the dynamic model, analyze the influence of bolt radii at different connection points on the force and motion characteristics of the hatch lifting mechanism, and obtain the effective lifting height.

10. The method according to any one of claims 1 to 8, characterized in that, In S2, the bolt radii of the plurality of connecting parts include at least the lower lifting rocker arm joint bolt radius, the upper lifting rocker arm joint bolt radius, the C-type connecting rod second joint bolt radius, the C-type connecting rod first joint bolt radius, the spline shaft connecting rod first joint bolt radius, and the spline shaft connecting rod second joint bolt radius in the main lifting mechanism.

Images