Wear prediction method for brake pad of elevator drum brake

Abstract

The abrasion prediction method for the brake pad of the elevator drum brake comprises the steps that a brake test bed is built, the brake duration and the brake distance under different brake pressure and brake rotating speed working conditions are obtained through a brake test, and the average friction coefficient and the average abrasion coefficient are calculated; least square polynomial fitting is carried out on the average friction coefficient and the average wear coefficient, Latin hypercube sampling is carried out on the braking parameters, and simulation working condition data are generated; establishing a brake finite element model to carry out heat-stress-wear coupling simulation analysis on each simulation working condition to obtain the maximum wear depth of the surface of the brake pad under each simulation working condition; constructing a brake pad surface maximum wear depth prediction model based on a PSO-BP neural network, and training the prediction model by adopting the simulation working condition data and the brake pad surface maximum wear depth; and adopting the trained PSO-BP neural network-based brake pad surface maximum wear depth prediction model to predict the brake pad surface maximum wear depth under a specific working condition.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of wear prediction of brake pads, and relates to a wear prediction method for brake pads of an elevator drum brake. BACKGROUND

[0002] The research on the friction and wear of brake pads in the braking process in the prior art is mainly directed to vehicle brake pads and train brake pads, and the research on elevator drum brake pads is less. The main research direction for vehicle brake pads is the macro analysis of dynamics and thermodynamics in the braking process and the research on the micro wear mechanism, and the methods used can be divided into two categories: analytical calculation and experimental method. The analytical method generally does not consider the mutual relationship between wear and contact stress, simplifies the wear model, and models by comprehensively using the theories of thermodynamics, materials science and kinematics to obtain the analytical solution of the wear amount. The experimental method is divided into using specific experimental devices and finite element simulation. The method of using specific test bench braking can accurately obtain the wear characteristics of brake pads under specific working conditions, but it has certain requirements for the device, and it is high in cost and inconvenient when studying the whole life cycle of brake pads. The finite element simulation method solves the problem of large calculation error caused by model simplification in the traditional analytical method, and can effectively calculate complex boundary conditions, coupling problems and nonlinear problems. Through the simulation experiment method, some data that are not easy to measure in real experiments, such as temperature, force distribution and its change trend, can be obtained, but since there are a large number of steps that need to be defined artificially in the simulation setting, and the requirement for the computing resources of the equipment is also high, the accuracy needs to be corrected simultaneously with the real test. There are limitations in using only one of the methods.

[0003] Chinese patent CN116341147A is a finite element simulation analysis method for optimizing the design of brake structure parameters. Before the simulation experiment, the friction coefficient and wear rate between the brake pad and the brake wheel are not considered to change due to different working conditions in the actual braking process when the braking condition changes. Moreover, when the parameters of the brake wheel and the brake pad are changed at the same time, if the initial braking load and the initial speed are unchanged, the braking performance provided by the brake will certainly change. In this case, although the influence of structural changes on wear can be evaluated accordingly, there are certain problems in extending these to the concept of "optimization", and the overall judgment needs to be combined with the braking performance.

[0004] A method for predicting the service life of brake pads using an inertia bench only considers the maximum wear under the current brake pad. In actual braking, the thickness of the brake pad decreases with the increase of the number of uses, and different thicknesses will change the structure under the action of external force, and the maximum equivalent stress and temperature will also change. Since wear involves the coupling factors of force and temperature, the change in thickness will inevitably change the wear trend, and further research on the wear trend in the whole life cycle is needed. SUMMARY

[0005] To solve the above technical problems, the purpose of the present application is to provide a wear prediction method for elevator drum brake pads.

[0006] The present application provides a wear prediction method for elevator drum brake pads, comprising:

[0007] Step 1: Build a brake test bench, obtain the brake duration and brake distance under different brake pressure and brake speed conditions through bench brake test, and calculate the average friction coefficient and average wear coefficient;

[0008] Step 2: Least square polynomial fitting is performed on the average friction coefficient and the average wear coefficient, and Latin hypercube sampling is performed on the brake parameters to generate simulation condition data;

[0009] Step 3: Establish a brake finite element model to carry out thermal-stress-wear coupling simulation analysis on each simulation condition to obtain the maximum wear depth of the brake pad under each simulation condition;

[0010] Step 4: Construct a brake pad surface maximum wear depth prediction model based on PSO-BP neural network, and train and test the prediction model using simulation condition data and brake pad maximum wear depth;

[0011] Step 5: Use the trained brake pad surface maximum wear depth prediction model based on PSO-BP neural network to predict the wear depth of the brake pad under specific conditions.

[0012] The wear prediction method for elevator drum brake pads of the present application has the following

[0013] Advantages:

[0014] The technical scheme provided by the present application combines a brake test bench, a finite element simulation experiment and a neural network prediction to realize the prediction of the wear depth of the brake pad. First, the test bench is used to obtain the average friction coefficient and the average wear coefficient under different initial braking pressures and initial speeds. Then, based on the average friction coefficient and the average wear coefficient obtained by the bench brake test, a fitting polynomial is obtained, the extracted brake parameters are brought into the polynomial to calculate the average friction coefficient and the average wear coefficient under each group of brake parameter samples, and then the working condition data for subsequent simulation are formed. Then, the maximum wear depth of the brake pad under each simulation working condition is obtained through finite element simulation. Then, a maximum wear depth prediction model of the brake pad surface based on a PSO-BP neural network is constructed, and the simulation working condition data and the maximum wear depth of the brake pad are used to train the prediction model, and finally the maximum wear depth is predicted through the trained prediction model. The present application reduces the cost of bench test research on the service life of the brake pad, and has higher accuracy compared with pure simulation experiment. BRIEF DESCRIPTION OF DRAWINGS

[0015] Figure 1 is a flowchart of a brake pad wear prediction method of an elevator drum brake of the present application;

[0016] Figure 2 is a structural schematic diagram of a brake test bench of an embodiment of the present application;

[0017] Figure 3 is a flowchart of a thermal-stress-wear coupling simulation solution of an embodiment of the present application;

[0018] Figure 4 is a change curve of the fitness of a prediction model in the training process of an embodiment of the present application;

[0019] Figure 5 is a comparison result diagram of the maximum wear depth prediction value and the true value of the brake pad surface of an embodiment of the present application.

[0020] 1-rotating main shaft, 2-servo motor, 3-clutch, 4-brake wheel, 5-brake pad fixing clamp, 6-hydraulic pump, 7-base. DETAILED DESCRIPTION

[0021] As shown in Figure 1 , a brake pad wear prediction method of an elevator drum brake of the present application, comprising:

[0022] Step 1: build a brake test bench, obtain the braking time and braking distance under different braking pressures and braking speeds through bench brake test, and calculate the average friction coefficient and the average wear coefficient.

[0023] The brake test bench is as shown in Figure 2As shown, including: installed on the base 7 on the brake wheel 4, rotating main shaft 1, servo motor 2, clutch 3 and brake pad fixed fixture 5. Brake wheel 4 through the rotating main shaft 1 connects the clutch 3 and servo motor 2, brake pad fixed on the brake pad fixed fixture 5.

[0024] In order to ensure that the brake pad and the contact area of the brake wheel is more than 85%, before the formal test, first run three brake wheel and brake pad operation. Run-in condition is brake wheel speed 24 rad / s, brake pressure 23 kN.

[0025] In specific implementation, by Figure 2 As shown in the brake test bench, under the conditions listed in table 1, bench brake test. In the brake test, the use of control system to open servo motor 2, open the clutch 3 after servo motor 2 speed rise, through the transmission of rotating main shaft 1 drive brake wheel 4 rotation, when the speed reaches the requirements and keep stable, control system drive hydraulic pump 6 with brake pad fixed fixture 5 makes brake pad and brake wheel 4 contact. Record the braking time and braking distance.

[0026] Table 1 brake parameters under different conditions

[0027]

[0028] In specific implementation, step 1 in the calculation of the average friction coefficient is specific for:

[0029] (1) in the brake pad fixed fixture installation force sensor, measured by the vertical instantaneous friction force due to brake friction.

[0030] (2) by the following formula to calculate the instantaneous friction coefficient:

[0031]

[0032] In the formula, μ is the instantaneous friction coefficient between the brake pad and the brake wheel, f is the instantaneous friction force measured by the cantilever sensor, F is the brake pressure applied by the brake pad on the brake wheel.

[0033] (3) under the same conditions, repeat the test 3 times, take the average value of the average friction coefficient under the brake condition.

[0034] (4) repeat the above steps to obtain the average friction coefficient under the different brake pressure and brake speed conditions set.

[0035] In specific implementation, step 1 in the calculation of the average wear coefficient is specific for:

[0036] (1) with the precision of not less than 0.1 g weighing instrument to measure the mass loss of brake pad after brake test.

[0037] (2) The wear coefficient of the brake pad is calculated according to the following formula:

[0038]

[0039] In the formula, q is the wear coefficient of the brake pad, m1 is the weight of the brake pad before the brake test, m2 is the weight of the brake pad after the brake test, H is the hardness of the brake pad material, F is the brake pressure applied to the brake wheel by the brake pad, p is the density of the brake pad material, and s is the brake distance during the entire brake process.

[0040] (3) After repeating the test 3 times under the same conditions, the average wear coefficient under the brake condition is obtained by taking the average value.

[0041] (4) Repeat the above steps to obtain the average wear coefficient under different brake pressure and brake speed conditions.

[0042] Step 2: Perform least squares polynomial fitting on the average friction coefficient and the average wear coefficient, and perform Latin hypercube sampling on the brake parameters to generate simulation condition data, specifically:

[0043] Step 2.1: For the brake speed-average friction coefficient and brake speed-average wear coefficient data under multiple brake pressures in the bench test, use the brake speed as the independent variable, the average friction coefficient and the average wear coefficient as the dependent variable, and use the multi-order least squares polynomial regression to establish a polynomial fitting model; when the average relative error of the established model in the fitting interval of each pressure level is not more than 10%, the polynomial can be used for interpolation prediction of the average friction coefficient and the average wear coefficient at any speed point under the same pressure.

[0044] In specific implementation, the average friction coefficient and the average wear coefficient under brake pressures of 15 kN, 20 kN and 25 kN are modeled by 3-order polynomial regression using the least squares method, with the average relative error controlled to be not more than 10%.

[0045] Step 2.2: Within the preset range of each brake parameter, a plurality of parameter combination samples are sampled using the Latin hypercube test design method, and the average friction coefficient and the average wear coefficient under each parameter combination sample are calculated using the polynomial model in step 2.1 as the subsequent simulation condition data; the brake parameters include brake speed, brake pressure and brake pad thickness.

[0046] In specific implementation, within the preset range of each brake parameter shown in Table 2, 120 samples are sampled using the Latin hypercube test design method, and the average friction coefficient and the average wear coefficient of each sample are calculated using the polynomial fitting model to generate simulation condition data.

[0047] Table 2: Brake parameter value range

[0048]

[0049] Step 3: Establish a finite element model of the brake to carry out thermal-stress-wear coupling simulation analysis under each simulation condition, and obtain the maximum wear depth of the brake pad under each simulation condition. Specifically:

[0050] Step 3.1: Only keep the key components of the drum brake, i.e. the brake pad and the brake wheel, and establish a three-dimensional model of the brake based on the structural parameters of the key components. The structural parameters of the key components include: the inner radius of the brake wheel, the outer radius of the brake wheel, the thickness of the brake wheel, the width of the brake wheel, the width of the brake pad, the wrap angle of the brake pad, and the thickness of the brake pad.

[0051] In specific implementation, the inner radius of the brake wheel is 390 mm, the outer radius of the brake wheel is 510 mm, the thickness of the brake wheel is 120 mm, and the width of the brake wheel is 130 mm. The width of the brake pad is 130 mm, the wrap angle is 30°, and the thickness of the brake pad is 10 mm.

[0052] Step 3.2: Import the established three-dimensional model of the brake into the finite element analysis software and assign material properties. The material properties include: the elastic modulus, density, thermal conductivity, Poisson's ratio, thermal expansion coefficient, and specific heat capacity of the brake pad and the brake wheel. The material parameters of the brake wheel and the brake pad are shown in Table 3.

[0053] Table 3 Material parameters of brake pad and brake wheel

[0054]

[0055] Step 3.3: Set the contact conditions and boundary conditions, apply the load, and perform meshing to build the finite element model of the brake.

[0056] Set the contact conditions and boundary conditions for the finite element model and apply the load. Specifically, define the rotational degrees of freedom of the brake wheel by rigid body constraints, constrain the degrees of freedom of the brake pad except the radial direction, add contact pairs to the brake wheel and the brake pad, and add convective heat transfer boundary conditions to the brake pad and the brake wheel. The convective heat transfer coefficient is calculated by the following formula:

[0057]

[0058] where λ a represents the thermal conductivity of air, D represents the outer diameter of the brake wheel, R e represents the Reynolds number.

[0059] Step 3.4: Perform thermal-stress-wear coupling simulation for each set of simulation condition data obtained in Step 2.2, such as Figure 3The entire braking process is discretized into B incremental steps during the simulation, and the thermal-stress coupling analysis is performed in each incremental step to output the normal contact pressure and contact slip.

[0060] Step 3.5: Assuming that the contact nodes of the brake wheel and the brake pad in the finite element grid are a and the incremental step is b, the wear increment of each contact node of the brake pad is calculated:

[0061]

[0062] In the formula, Ah(a, b) is the wear increment of the contact node a in the bth incremental step, Q is the average wear coefficient calculated in step 2.2, H is the hardness of the brake pad surface, p(a, b) is the normal contact pressure borne by the contact node a in the bth incremental step, and As(a, b) is the contact slip of the contact node a in the bth incremental step.

[0063] During braking, the brake pad is considered to be stationary, so As is the sliding displacement of the brake wheel relative to the brake pad, and the cumulative maximum wear depth is:

[0064]

[0065] Step 3.6: Update the grid of the wear surface using the ALE algorithm, and loop steps 3.4-3.5 until the simulation time reaches the preset endpoint to obtain the maximum wear depth of the brake pad under each group of simulation conditions.

[0066] Step 4: Construct a brake pad surface maximum wear depth prediction model based on a PSO-BP neural network, train the prediction model using simulation condition data and brake pad maximum wear depth, and specifically:

[0067] Step 4.1: The simulation condition data obtained in step 2.2 and the brake pad surface maximum wear depth obtained in step 3 under each group of simulation conditions constitute a data set, which is normalized and divided into a training set and a test set according to a ratio of 7:3. The normalization method uses the Mapminmax function:

[0068]

[0069] In the formula, y, x are the normalized and non-normalized data respectively; y max = 1; y min = -1; x min , x max are the minimum and maximum values in the input parameters respectively.

[0070] Step 4.2: Establish a BP neural network with a defined network structure, transfer function, and learning rules. Use the braking speed, braking pressure, average friction coefficient, and brake pad thickness as inputs to the BP neural network, and the maximum wear depth as output. Specifically:

[0071] The input layer parameters of the BP neural network are the braking speed, braking pressure, average friction coefficient, and brake pad thickness. The output layer parameter is the maximum wear depth of the brake pad during a single braking simulation. The number of neurons in the hidden layer is determined according to the following formula:

[0072]

[0073] Where W BP is the number of neurons in the hidden layer, Q in is the number of neurons in the input layer, Q out is the number of neurons in the output layer, Q η To adjust the parameters, the value range is 0 to 10, the BP neural network training times is 1000, the learning rate is 0.1, and the training error is 0.0001.

[0074] The calculation formula of the hidden layer is as follows:

[0075]

[0076] Where b d is the output value of the dth neuron in the hidden layer, is the input value of the cth neuron in the input layer, ξ cd and ψ d are the weight and threshold between the cth neuron in the input layer and the dth neuron in the hidden layer, respectively.

[0077] The calculation formula of the output layer is as follows:

[0078]

[0079] Where y k is the output value of the kth neuron, ξ dk and ψ k are the weight and threshold between the cth neuron in the hidden layer and the kth neuron in the output layer, respectively.

[0080] Step 4.3: Use the particle swarm optimization (PSO) to optimize the weights and thresholds of the BP neural network to obtain a prediction model for the maximum wear depth of the brake pad surface based on the PSO-BP neural network. Specifically:

[0081] Step 4.3.1: Select the parameters of the PSO algorithm, including the particle swarm size of 40, the maximum number of iterations of 500, the inertia weight ω of 0.8, the individual learning factor c1 of 1.6, and the group learning factor c2 of 1.8; initialize the speed and position of each particle in the population, and each particle is composed of a set of BP neural network weights and thresholds.

[0082] Step 4.3.2: Take the loss function value as the fitness index, and the fitness function is:

[0083]

[0084] where N is the number of training set samples, M is the number of neural network output nodes, y j,i and y j are the actual value and expected value of the jth output node in the ith sample, w is the weight coefficient of the jth output node,

[0085] Step 4.3.3: Calculate the particle fitness value based on the fitness function, update the individual extreme value and global extreme value of each particle; and adjust the speed and position of the particle accordingly until convergence or the maximum number of iterations is reached.

[0086] Step 4.3.4: Load the final global extreme value, i.e., the optimal weight and threshold, into the BP neural network, and use the training set for back propagation learning until the error or iteration termination condition is met, to obtain the final prediction model.

[0087] The smaller the fitness, the smaller the model prediction error. After 30 iterations, the model fitness value change curve is as shown in Figure 4 .

[0088] Step 4.4: Import the training set into the brake pad surface maximum wear depth prediction model based on PSO-BP neural network for training, and input the brake rotation speed, brake pressure, average friction coefficient, and brake pad thickness in the test set into the trained network to obtain the prediction result of the brake pad surface maximum wear depth, and compare it with the true result corresponding to the test set.

[0089] Step 4.5: Select the best hidden layer node number by calculating the mean square error, and use the mean square error, mean absolute error, determination coefficient, and average relative error as indicators to evaluate the BP neural network prediction model.

[0090] RMSE, MAE, R 2 , and MRE are 4.31e-6, 4.42e-6, 0.9835, and 10.06%, respectively, which meet the expected effect. The comparison result graph of the maximum wear depth prediction value and the true value is as shown in Figure 5 .

[0091] Step 5: The trained PSO-BP neural network-based brake pad surface maximum wear depth prediction model is used to predict the brake pad wear depth under specific working conditions.

[0092] The key point of the application: Since the properties related to friction and wear cannot be studied only by simulation experiments, a brake test bench is built to obtain relevant data for subsequent simulation experiments and neural network prediction.

[0093] 1. In the initial simulation experiment condition, the friction coefficient and wear coefficient need to be set artificially. In classical friction theory, the friction coefficient is only related to the two materials in contact. However, experiments conducted in a large number of literature reflect that different braking loads and sharp changes in temperature can cause changes in the friction coefficient and wear coefficient. Before the simulation experiment, the friction coefficient and wear coefficient need to be determined through specific experiments. Because it is impossible to obtain the friction coefficient and wear coefficient of each point at every moment in reality, the average value is used to determine the friction coefficient and wear coefficient.

[0094] Therefore, the method of the application is mainly based on two assumptions: ① the main factors affecting the friction coefficient and wear coefficient are material, temperature and force; ② according to the wear law of mechanical parts, the change of material properties between two adjacent braking processes can be ignored in the stable friction stage, and the research of the application is aimed at the stable wear stage.

[0095] In addition, the braking time and braking mileage are different under different braking speeds and braking loads, and the braking time needs to be confirmed artificially in the simulation experiment. Compared with the analytical method derived by simplifying the model, the accuracy of using specific bench test is higher.

[0096] 2. Based on the cost problem, for the specific brake pad to be tested, too many bench brake tests not only affect the device, but also reduce the actual service life of the brake pad, so the data set is obtained in the form of simulation.

[0097] It is relatively easy to change the thickness of the brake pad in the simulation experiment. Here is based on the assumption that the influence of thickness is indirectly affected by the wear amount through the influence of stress and temperature, and although the influence of the friction coefficient and wear coefficient also considers the two factors of stress and temperature, the main influence of temperature and stress is still the load and initial speed, and the possible influence of the change of thickness on the friction coefficient and wear coefficient is ignored. At present, there is a lack of specific theoretical derivation of the influence of brake pad thickness on stress and temperature in the research status, and even in some simulation-related literature, the main influence is the influence of materials, etc. Therefore, the application considers it and studies it by using the simulation method.

[0098] In the simulation experiment, the change trend of temperature, force, wear and tear and the like of a certain point can be obtained, and the distribution can be directly observed, which is difficult to obtain in the actual experiment.

[0099] The above merely describes the preferred embodiments of the present application, and is not intended to limit the idea of the present application. Any modification, equivalent replacement, improvement and the like made within the spirit and principle of the present application shall be included in the protection scope of the present application.

Claims

1. A method for predicting wear of brake pads of an elevator drum brake, characterized in that: include: Step 1: Build a brake test bench and conduct bench braking tests to obtain the braking time and distance under different braking pressures and braking speeds, and calculate the average friction coefficient and average wear coefficient; Step 2: Perform least squares polynomial fitting on the average friction coefficient and the average wear coefficient, and perform Latin hypercube sampling on the braking parameters to generate simulation working condition data; Step 3: Establish a brake finite element model and conduct thermal-stress-wear coupling simulation analysis for each simulation condition to obtain the maximum wear depth of the brake pad surface under each simulation condition; Step 4: Construct a prediction model for the maximum wear depth of the brake pad surface based on the PSO-BP neural network, and use the simulation working condition data and the maximum wear depth of the brake pad to train and test the prediction model; Step 5: Use the trained brake pad surface maximum wear depth prediction model based on the PSO-BP neural network to predict the maximum wear depth of the brake pad surface under specific working conditions.

2. The wear prediction method for brake pads of an elevator drum brake according to claim 1, wherein: The brake test bench includes: a brake wheel installed on a base, a rotating main shaft, a servo motor, a clutch and a brake pad fixing fixture; the brake wheel is connected to the clutch and the servo motor through the rotating main shaft, and the brake pad is fixed on the brake pad fixing fixture; during the brake test, the control system is used to turn on the servo motor. After the clutch is opened, the speed of the servo motor increases, and the brake wheel is driven to rotate through the transmission action of the rotating main shaft. After the speed reaches the requirement and remains stable, the control system drives the hydraulic pump to drive the brake pad fixing fixture to make the brake pad contact with the brake wheel.

3. The wear prediction method for brake pads of an elevator drum brake according to claim 1, wherein: The calculation of the average friction coefficient in step 1 is specifically as follows: (1) Install a force sensor on the brake pad fixture to measure the vertical upward instantaneous friction force generated by braking friction; (2) Calculate the instantaneous friction coefficient using the following formula: Where μ is the instantaneous friction coefficient between the brake pad and the brake wheel, f is the instantaneous friction force measured by the cantilever sensor, and F is the braking pressure applied by the brake pad on the brake wheel; (3) Repeat the test three times under the same conditions and take the average value to obtain the average friction coefficient under the braking condition; (4) Repeat the above steps to obtain the average friction coefficient under different braking pressure and braking speed conditions.

4. The wear prediction method for brake pads of an elevator drum brake according to claim 1, wherein: The calculation of the average wear coefficient in step 1 is specifically as follows: (1) Use a weighing instrument with an accuracy of not less than 0.1g to measure the mass loss of the brake pad after the braking test; (2) Calculate the wear coefficient of the brake pad according to the following formula: Where q is the brake pad wear coefficient, m1 is the brake pad weight before the braking test, m2 is the brake pad weight after the braking test, H is the hardness of the brake pad material, F is the braking pressure applied by the brake pad on the brake wheel, ρ is the density of the brake pad material, and s is the braking distance of the entire braking process; (3) Repeat the test three times under the same conditions and take the average value to obtain the average wear coefficient under the braking condition; (4) Repeat the above steps to obtain the average wear coefficient under different braking pressure and braking speed conditions.

5. The wear prediction method for brake pads of an elevator drum brake according to claim 1, wherein: The step 2 is specifically as follows: Step 2.1: Based on the braking speed-average friction coefficient and braking speed-average wear coefficient data at various braking pressures in the bench test, a polynomial fitting model is established using multi-order least squares polynomial regression, with braking speed as the independent variable and average friction coefficient and average wear coefficient as the dependent variables. When the average relative error of the established model within the fitting interval of each pressure level does not exceed 10%, the polynomial can be used to interpolate and predict the average friction coefficient and average wear coefficient at any speed point under the same pressure. Step 2.2: Within the preset range of each braking parameter, use the Latin hypercube experimental design method to sample multiple groups of parameter combination samples. Using the polynomial model in step 2.1, calculate the average friction coefficient and average wear coefficient under each group of parameter combination samples as the subsequent simulation working condition data; the braking parameters include braking speed, braking pressure and brake pad thickness.

6. The wear prediction method for brake pads of an elevator drum brake according to claim 5, characterized in that: The step 3 is specifically as follows: Step 3.1: Retain only the key components of the drum brake, the brake pad and brake wheel, and build a 3D brake model based on the structural parameters of the key components. The key component structural parameters include: brake wheel inner radius, brake wheel outer radius, brake wheel thickness, brake wheel width, brake pad width, brake pad wrap angle, and brake pad thickness; Step 3.2: Import the established 3D brake model into the finite element analysis software and assign material properties, including the elastic modulus, density, thermal conductivity, Poisson's ratio, thermal expansion coefficient, and specific heat capacity of the brake pad and brake wheel; Step 3.3: Set contact and boundary conditions, apply loads, and perform meshing to complete the brake finite element model. Step 3.4: Perform a thermal-stress-wear coupling simulation for each set of simulation data obtained in step 2.

2. During the simulation, the entire braking process is discretized into B incremental steps. In each incremental step, a thermal-stress coupling analysis is first performed to output the normal contact pressure and contact slip. Step 3.5: Assuming that the contact node between the brake wheel and the brake pad in the finite element mesh is a and the incremental step is b, calculate the wear increment of each contact node of the brake pad: Where Δh(a,b) is the wear increment of contact node a at the bth incremental step, and Q is The average wear coefficient calculated in step 2.2, H is the material hardness of the brake pad surface, p(a, b) is the normal contact pressure on contact node a in the bth increment, and Δs(a, b) is the contact slip of contact node a in the bth increment; During the braking process, the brake pad is considered to be stationary, so Δs is the sliding displacement of the brake wheel relative to the brake pad, and the cumulative maximum wear depth is: Step 3.6: Use the ALE algorithm to update the mesh of the worn surface. Repeat steps 3.4–3.5 until the simulation time reaches the preset end point to obtain the maximum wear depth of the brake pad surface under each set of simulation conditions.

7. The wear prediction method for brake pads of an elevator drum brake according to claim 6, wherein: The step 4 is specifically as follows: Step 4.1: The data sets of each set of simulation conditions obtained in step 2.2 and the maximum wear depth of the brake pad surface under each set of simulation conditions obtained in step 3 are used to form a data set, and the data sets are divided into a training set and a test set after normalization. Step 4.2: Establish a BP neural network with a defined network structure, transfer function, and learning rules. The braking speed, braking pressure, average friction coefficient, and brake pad thickness are used as inputs of the BP neural network, and the maximum wear depth is used as output. Step 4.3: Use the particle swarm optimization (PSO) to optimize the weights and thresholds of the BP neural network to obtain a prediction model for the maximum wear depth of the brake pad surface based on the PSO-BP neural network. Step 4.4: The training set is imported into the PSO-BP neural network-based brake pad surface maximum wear depth prediction model for training. The braking speed, braking pressure, average friction coefficient, and brake pad thickness in the test set are input into the trained network to obtain the predicted maximum wear depth of the brake pad surface. The predicted results are then compared with the actual results corresponding to the test set. Step 4.5: Filter out the optimal number of hidden layer nodes by calculating the mean square error, and use the mean square error, mean absolute error, determination coefficient and mean relative error as evaluation indicators to evaluate the BP neural network prediction model.

8. The wear prediction method for brake pads of an elevator drum brake according to claim 7, wherein: The step 4.2 is specifically as follows: The input layer parameters of the BP neural network are the braking speed, braking pressure, average friction coefficient, and brake pad thickness. The output layer parameter is the maximum wear depth of the brake pad during a single braking simulation. The number of neurons in the hidden layer is determined according to the following formula: Where W BP is the number of hidden layer neurons, Q in is the number of neurons in the input layer, Q out is the number of neurons in the output layer, Q η To adjust the parameters, the value range is 0 to 10, the BP neural network training times is 1000, the learning rate is 0.1, and the training error is 0.0001; The calculation formula of the hidden layer is as follows: Where b d is the output value of the dth neuron in the hidden layer, is the input value of the cth neuron in the input layer, ξ cd and ψ d are the weight and threshold between the cth neuron in the input layer and the dth neuron in the hidden layer respectively; The calculation formula of the output layer is as follows: Where y k is the output value of the kth neuron, ξ dk and ψ k are the weight and threshold between the cth neuron in the hidden layer and the kth neuron in the output layer, respectively.

9. The wear prediction method for brake pads of an elevator drum brake according to claim 7, wherein: The step 4.3 is specifically as follows: Step 4.3.1: Select the parameters of the PSO algorithm, including the particle swarm size, the maximum number of iterations, the inertia weight ω, the individual learning factor c1, and the group learning factor c2; initialize the speed and position of each particle in the swarm. Each particle consists of a set of BP neural network weights and thresholds; Step 4.3.2: Use the loss function value as the fitness indicator. The fitness function is: Among them, N is the number of training set samples; M is the number of neural network output nodes; y j,i and are the actual value and expected value of the jth output node in the i-th sample, respectively, w j is the weight coefficient of the j-th output node, Step 4.3.3: Calculate the particle fitness value based on the fitness function, update the individual extreme value and global extreme value of each particle, and adjust the particle speed and position accordingly until convergence or the maximum number of iterations is reached; Step 4.3.4: Load the final global extreme value, that is, the optimal weight and threshold, into the BP neural network, and use the training set for backpropagation learning until the error or iteration termination condition is met to obtain the final prediction model.

Images