Structural optimization method for working mechanism of hydraulic aerial cage driven by digital intelligence

Abstract

The invention provides a structure optimization method for a working mechanism of a digital-intelligent-driven overhead working truck, relates to the technical field of machinery, and effectively breaks through the limitation of a traditional design method through deep fusion of a data-driven modeling technology and a multi-target intelligent optimization algorithm. Cooperative optimization of the working mechanism on multi-target performance such as strength, rigidity and light weight is achieved, the design efficiency and precision are remarkably improved, and therefore the comprehensive requirements of modern engineering for light weight, high strength and high stability of equipment are met.

Description

Technical Field

[0001] This invention relates to the field of mechanical technology, and in particular to a method for optimizing the working mechanism structure of a digitally driven aerial work platform. Background Technology

[0002] Aerial work platforms are indispensable equipment in modern engineering construction, widely used in municipal facility maintenance, high-rise building construction, and industrial equipment repair. The working mechanism, as the core moving component of the aerial work platform, typically includes key components such as the basic boom, telescopic boom, articulated boom, and hydraulic cylinders. The strength, rigidity, stability, and lightweight nature of these components directly determine the overall machine's operational performance, safety, reliability, service life, and energy efficiency. With the acceleration of urbanization and the development of industrialized construction, the market is placing higher demands on aerial work platforms, including higher operating efficiency, superior lightweight design, and greater adaptability to complex working conditions.

[0003] Traditional design methods rely primarily on engineers' empirical formulas and classic single-objective optimization algorithms, making it difficult to achieve an effective balance between multi-dimensional performance indicators such as structural strength, stiffness, and lightweighting. While finite element analysis-based simulation technology can provide high-precision structural response predictions, its computational cost is extremely high, especially in optimization applications involving multiple parameters and objectives, where its low computational efficiency severely restricts design iteration speed and global optimization efficiency. Data-driven modeling techniques can, to some extent, replace expensive finite element analysis and alleviate computational burden, but their performance is highly dependent on sample distribution and model assumptions. In complex multi-objective optimization problems, single data-driven models often suffer from insufficient prediction accuracy and limited generalization ability, failing to meet the stringent requirements of high reliability and robustness in engineering practice. Therefore, more intelligent and efficient model construction and optimization solution strategies are urgently needed. Summary of the Invention

[0004] To address the problems of low computational efficiency, difficulty in achieving multi-objective equilibrium, and insufficient accuracy and generalization ability of single data-driven models in existing technologies, this invention provides a digitally driven method for optimizing the working mechanism structure of aerial work platforms. This method effectively overcomes the limitations of traditional design methods by deeply integrating data-driven modeling technology with multi-objective intelligent optimization algorithms. It achieves synergistic optimization of the working mechanism in terms of multi-objective performance such as strength, stiffness, and lightweighting, significantly improving design efficiency and accuracy, thereby meeting the comprehensive requirements of modern engineering for lightweight, high-strength, and high-stability equipment.

[0005] The present invention achieves the above-mentioned technical objectives through the following technical means.

[0006] A method for optimizing the working mechanism structure of a digitally driven aerial work platform includes the following steps:

[0007] Step S1: Establish a kinematic and mechanical analysis model of the working mechanism of the aerial work vehicle, and perform kinematic position solution and mechanical calculation for the working conditions to obtain the load data of each hinge point of the working mechanism.

[0008] Step S2: Determine the key structural parameters that affect structural performance as design variables, sample within the design variable space, and generate a design variable sample set covering the design space;

[0009] Step S3: Establish the parametric three-dimensional geometric model and finite element analysis model of the working mechanism of the aerial work vehicle; based on the hinge load data obtained in step S1, apply corresponding boundary conditions and loads to each design variable sample in the design variable sample set generated in step S2, carry out finite element analysis, and extract and record the calculation results of each structural performance.

[0010] Step S4: Integrate the design variable combination sample set from step S2 with the structural performance calculation results obtained from step S3 to form the original dataset. Perform data preprocessing on the original dataset, and divide the preprocessed dataset into a training dataset and a test dataset.

[0011] Step S5: Based on the training dataset, for each structure performance that needs to be optimized, multiple different types of data-driven models are selected in parallel for data-driven modeling, and hyperparameter tuning and structure design are performed for each model.

[0012] Step S6: Using the test dataset, rigorously evaluate all candidate data-driven models trained in Step S5 for each structural performance; for each structural performance index, select the data-driven model with the best prediction performance and the strongest robustness as the fast calculation model for the structural performance index in subsequent optimization design; thus, a multi-data-driven model system composed of multiple independent optimal models is constructed.

[0013] Step S7: Based on the multi-data-driven model system selected in Step S6, construct a multi-objective optimization mathematical model for the working mechanism of the aerial work vehicle;

[0014] Step S8: Use a multi-objective intelligent optimization algorithm to solve the multi-objective optimization model established in step S7, and find a set of optimal solutions that can achieve effective balance among multiple objectives;

[0015] Step S9: Perform non-dominated sorting and filtering on the optimized solution set obtained in step S8, and combine it with multi-attribute decision-making methods to select the best design scheme.

[0016] In the above scheme, the force analysis of each hinge point of the working mechanism in step S1 refers to the force and torque on each hinge point in the working mechanism of the aerial work vehicle, which is related to the working conditions of the aerial work vehicle. During the analysis, the working mechanism is divided into multiple rod groups, and then a corresponding equilibrium equation is established for each rod group. The force condition of the corresponding hinge point can be calculated, and the result is recorded as the basis for the application of boundary conditions and loads in subsequent finite element analysis.

[0017] In the above scheme, step S2 uses Latin hypercube sampling.

[0018] In the above scheme, the calculation results of the structural performance indicators in step S3 include the maximum stress, maximum deformation, and structural mass.

[0019] In the above scheme, step S4, data preprocessing includes data cleaning and data normalization / standardization; the dataset is divided into training dataset and test dataset using hold-out method or K-fold cross-validation.

[0020] In the above scheme, in step S5, the data-driven model includes a neural network model, a support vector machine model, a linear regression model, a Gaussian process regression model, and a random forest model; hyperparameter tuning includes grid search or Bayesian optimization, which are used to improve the fitting accuracy and generalization ability of each model.

[0021] In the above scheme, step S7, the multi-objective optimization mathematical model includes establishing the objective function, designing variables and constraints;

[0022] The objective functions are: strength, maximum stress δ, stiffness, maximum deformation ε, lightweight, and structural mass M; the optimization direction is to minimize all of them.

[0023] (2) The design variable is the dimensions of the bar group structure;

[0024] (3) Constraints:

[0025] 1) The maximum predicted stress value is less than or equal to the allowable stress of the material, i.e., δ ≤ 200 MPa;

[0026] 2) The predicted maximum deformation value is less than or equal to the maximum allowable displacement, i.e., ε ≤ 10 mm.

[0027] In the above scheme, step S9, the design scheme selection mechanism based on the multi-attribute decision method—ideal solution, has the following specific calculation process:

[0028] (1) Establish a multi-attribute decision matrix D = (d ij ) n×m , element d ij Let n be the objective function value of each solution, n be the size of the solution set, and m be the dimension of the objective function.

[0029] (2) Establish a standardized decision matrix R = (r ij ) n×m ,in

[0030]

[0031] (3) Establish a weighted decision matrix V = (v ij ) n×m Let the weight vector be w = (w j ) 1×m = (1 / 3, 1 / 3, 1 / 3), where

[0032] v ij =w j r ij

[0033] (4) Determine the positive ideal solution A + and negative ideal solution A - :

[0034]

[0035] In the formula v j + v j - —Calculate the minimum and maximum values ​​in the j-th column of the weighted decision matrix V, respectively;

[0036] (5) Calculate the Euclidean distance between each design scheme and the positive and negative ideal solutions:

[0037]

[0038] (6) Calculate the relative closeness of each design scheme to the positive and negative ideal solutions:

[0039]

[0040] In the formula, 0≤C i ≤1, C i The larger the value, the further the design is from the negative ideal solution, and the closer it is to the positive ideal solution, indicating a better design.

[0041] Compared with the prior art, the beneficial effects of the present invention are:

[0042] (1) Efficiently overcome computational bottlenecks. Constructing multiple data-driven models to replace expensive finite element simulations greatly shortens the time required for performance evaluation, solves the efficiency problem of repeatedly calling simulations in traditional optimization, makes complex multi-parameter and multi-objective optimization feasible within the engineering cycle, and significantly accelerates design iteration.

[0043] (2) Achieve multi-objective collaborative optimization. Innovatively integrate data-driven modeling and multi-objective intelligent optimization algorithms to effectively solve the trade-off problem between multiple conflicting objectives such as strength, stiffness and lightweighting, and significantly improve the overall performance of the working mechanism.

[0044] (3) Improve model accuracy and robustness. By adopting independent modeling for each performance index and a global optimization strategy, the optimal prediction model is selected for each performance index, effectively avoiding the problems of insufficient accuracy and weak generalization of a single model, and ensuring high accuracy and high robustness of the prediction system.

[0045] (4) Intelligent closed-loop ensures project implementation. A complete intelligent closed-loop process of "parametric modeling - data generation - model training - optimization solution - solution decision" is constructed, which greatly reduces the reliance on experience trial and error. The final solution is verified by multi-attribute decision-making and engineering experience to ensure that it meets performance targets and process, cost and safety specifications at the same time, which significantly improves practical value and product competitiveness. Attached Figure Description

[0046] Figure 1 This invention relates to a flowchart of a method for optimizing the working mechanism structure of a digitally driven aerial work platform.

[0047] Figure 2 This is a schematic diagram of a 3D model of an aerial work platform vehicle.

[0048] Figure 3 A simplified diagram of the working mechanism of an aerial work platform vehicle;

[0049] Figure 4 This is a schematic diagram of the basic arm structure parameters;

[0050] Figure 5 This is a graph showing the results of Latin hypercube sampling;

[0051] Figure 6 A finite element analysis cloud diagram of the basic arm;

[0052] Figure 7 A distribution chart of multidimensional structural performance index data;

[0053] Figure 8 The response map is trained for each data point driving the model based on the maximum stress data.

[0054] Figure 9 For each data point of the maximum deformation data, train the response map of the model.

[0055] Figure 10 Response maps are trained for each data-driven model based on structural quality data.

[0056] Figure 11 The test response diagrams for each data-driven model based on the maximum stress data;

[0057] Figure 12 The test response diagrams for each data-driven model based on the maximum deformation data;

[0058] Figure 13 Test response diagrams for each data-driven model based on structural quality data;

[0059] Figure 14 Here is the flowchart for the NSDE algorithm;

[0060] Figure 15 To optimize the distribution of the solution set in the target space;

[0061] Figure 16 To optimize the distribution of the solution set in the variable space;

[0062] Figure 17 The result is the relative proximity calculation of the non-dominated solution set;

[0063] Figure 18 The finite element simulation results were used to verify the optimized scheme.

[0064] 1-Turntable; 2-Basic connecting rod; 3-Basic boom; 4-Connecting plate; 5-Extension boom connecting rod; 6-Luffing hydraulic cylinder; 7-First extension boom; 8-Second extension boom; 9-Boom cover; 10-Folding boom hydraulic cylinder; 11-Folding boom connecting rod; 12-Connecting piece; 13-Working platform. Detailed Implementation

[0065] Embodiments of the present invention are described in detail below, examples of which are illustrated in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and intended to explain the present invention, and should not be construed as limiting the present invention.

[0066] A method for optimizing the working mechanism structure of a digitally-driven aerial work platform includes the following steps:

[0067] Step S1: Establish a kinematic and mechanical analysis model of the working mechanism of the aerial work vehicle, and perform kinematic position solution and mechanical calculation for typical working conditions to obtain load data at each hinge point of the working mechanism.

[0068] Step S2: Determine the key structural parameters that affect structural performance as design variables, and use advanced experimental design methods to uniformly and effectively sample within the design variable space to generate a sample set of design variable combinations covering the design space;

[0069] Step S3: Establish a parametric three-dimensional geometric model and finite element analysis model of the working mechanism of the aerial work vehicle; based on the hinge load data obtained in step S1, apply corresponding boundary conditions and loads to each design variable sample combination generated in step S2, carry out finite element analysis, extract and record the calculation results of key structural performance indicators, including but not limited to: maximum stress, maximum deformation (displacement), and structural mass.

[0070] Step S4: Integrate the design variable sample set from Step S2 with the corresponding multidimensional performance index results calculated in Step S3 to form the original dataset; perform data preprocessing on the original dataset, including but not limited to: data cleaning, data normalization / standardization, and dividing the dataset into training set and independent test set to ensure data quality and consistency;

[0071] Step S5: Based on the partitioned training dataset, for each structural performance metric that needs to be optimized, multiple different types of data-driven models are selected in parallel for independent training. Hyperparameter tuning techniques such as grid search and Bayesian optimization, as well as model structure optimization methods, are combined to maximize the fitting accuracy and generalization ability of each model.

[0072] Step S6: Using the reserved test dataset, rigorously evaluate all candidate data-driven models trained in Step S5 for each performance metric; select appropriate evaluation metrics to quantify their prediction accuracy and generalization performance; for each structural performance metric, select the data-driven model with the best prediction effect and the strongest robustness as the fast calculation model for that performance metric in subsequent optimization design; thus, construct a multi-data-driven model system composed of multiple independent optimal models, which can predict different aerial work platform structural performance metrics with high accuracy and high efficiency.

[0073] Step S7: Based on the high-precision multi-data driven model system selected in Step S6, construct a multi-objective optimization mathematical model for the working mechanism of the aerial work vehicle, including establishing the objective function, designing variables and constraints;

[0074] Step S8: Use an advanced multi-objective intelligent optimization algorithm to solve the multi-objective optimization model established in step S7, and find a set of optimal solutions that can achieve effective balance among multiple objectives;

[0075] Step S9: Perform non-dominated sorting and screening on the optimized solution set obtained in step S8, and combine multi-attribute decision-making methods and engineering experience to select the design scheme with the best comprehensive performance or the one that best meets the specific engineering requirements.

[0076] In step S1, the force analysis of each hinge point of the working mechanism refers to the force and torque on each hinge point in the working mechanism of the aerial work vehicle. These are related to the working conditions of the aerial work vehicle. During the analysis, the working mechanism is divided into multiple rod groups, and then a corresponding equilibrium equation is established for each rod group. This allows the force condition of the corresponding hinge point to be calculated. The result is recorded and used as the main basis for the application of boundary conditions and loads in subsequent finite element analysis.

[0077] In step S2, to ensure the accuracy and reliability of the data source for subsequent data-driven modeling, an advanced experimental design method, represented by Latin hypercube sampling, is selected for sampling. Latin hypercube sampling is a powerful and efficient experimental design method for solving the sampling problem of high-dimensional and expensive simulation models. Its core principle is to independently divide each design variable into equally probable intervals, randomly select a sample from each interval, and then randomly arrange and combine these sample values ​​into multidimensional sample points. This ensures a uniform distribution of sample points in each variable dimension, thereby achieving a better, unbiased, and well-filled coverage of the design space. Compared to simple random sampling, this method enables a more uniform distribution of sample points within the design space, improving the training and learning effect of the data-driven model.

[0078] In step S3, structural performance index data can be obtained either through self-programmed calculations or using CAD / CAE technology. A virtual prototype model of the aerial work platform is established, and finite element simulation analysis is performed to output performance index data. Based on the parametric modeling scheme, the working mechanism under different combinations of structural parameters can be easily analyzed and calculated, improving the efficiency and standardization of data acquisition. Furthermore, to avoid the curse of dimensionality caused by too many parameters, sensitivity analysis can be performed on the previously selected structural parameters and structural performance indexes, selecting the most critical structural parameters as design variables.

[0079] In step S4, the dataset is preprocessed, including but not limited to data cleaning, denoising, and normalization, to ensure data quality and consistency. Data partitioning methods such as hold-out and K-fold cross-validation are used to ensure the uniformity and consistency between the training and testing datasets, improving the reliability of model evaluation.

[0080] In step S5, data-driven modeling is a relatively new modeling technique in current engineering design. It is a method of constructing a system behavior model based on observed data rather than first principles. It accelerates the design process by replacing complex and time-consuming simulation calculations or high-precision models with a simplified model that requires less computation and solves quickly. This invention mainly uses a hybrid modeling approach employing commonly used data-driven models such as neural networks, support vector machines, linear regression, Gaussian process regression, and random forests. Furthermore, it optimizes the hyperparameters and designs the structure of each model as much as possible to fully realize its performance potential.

[0081] In step S6, given the diversity of design problems and the differences in the characteristics and applicable scenarios of data-driven models, validation datasets must be used sequentially during and after model training, and finally, a comprehensive test dataset must be used. By selecting appropriate evaluation metrics, such as root mean square error (RMSE), coefficient of determination (R²), maximum absolute error (MAE), and mean absolute percentage error (MAPE), the model is evaluated to select the best data-driven model with high prediction accuracy and strong generalization ability, ensuring high accuracy and reliability of its prediction results.

[0082] In step S7, the objective function of the multi-objective optimization mathematical model can be defined as a set of structural performance indicators that need to be optimized collaboratively, including but not limited to the structural mass, maximum stress value, and maximum deformation calculated based on the multi-data-driven model system in step S6 above; the design variables are mainly the key structural parameters determined in step S2, and their feasible value range is clearly defined; the constraints cover performance constraints (such as maximum allowable stress, maximum design allowable displacement, buckling safety factor, etc.), geometric constraints (such as dimensional boundary limits), and manufacturing process constraints.

[0083] In step S8, a suitable intelligent optimization algorithm is selected based on the mathematical characteristics of the optimization model established in step S7, and the corresponding algorithm parameters are set for solving. If the dimension of the optimization objective is greater than or equal to four, a multi-objective optimization algorithm should be given priority; if the dimension of the design variables is high, a large-scale optimization algorithm should be selected; if the scale of the constraints is large, an optimization algorithm with the ability to handle multiple constraints should be selected.

[0084] In step S9, the optimal solution set obtained in step S8 is first sorted and graded by non-dominated ordering to select all non-dominated solutions for further sorting. Then, the weight coefficients of each objective are set based on actual engineering requirements or data distribution characteristics. The comprehensive evaluation value of each solution is then calculated by a multi-attribute decision algorithm and manually verified in combination with factors such as process feasibility and cost. Finally, the design scheme with the best comprehensive performance and that meets specific engineering constraints is selected.

[0085] The following will, in conjunction with the accompanying drawings, take the structural parameter optimization design of the basic boom of a certain type of 20-meter aerial work platform as an example to clearly and completely describe the technical solution in the embodiment of the present invention.

[0086] Combined with appendix Figure 1 As shown in the figure, this embodiment provides a method for optimizing the working mechanism structure of a digitally driven aerial work platform.

[0087] Specifically, in step S1, in conjunction with the appendix Figure 2 and Figure 3 As shown, a parametric three-dimensional model and a mechanical model of the aerial work platform's working mechanism are first established. The rated load of the work platform under the typical working condition—maximum working radius condition—is decomposed along the x-axis and y-axis into w. x and w y Then, by establishing and solving the force balance and torque balance equations for the basic arm, telescopic arm, and folding arm, the forces at each hinge point of the working mechanism under this working condition can be calculated, serving as the load basis for subsequent finite element analysis.

[0088] In step S2, combined with the attached Figure 4 As shown, six structural parameter variables, all in mm, are selected for data-driven modeling, including the wall thicknesses P1 and P2 on the left and right sides of the basic arm, the wall thicknesses P3 and P4 on the top and bottom sides, and the inner wall length and width P5 and P6. The design variable values ​​are set to [18, 22], [18, 22], [45, 55], [45, 55], [270, 330], and [90, 110], respectively. 200 sets of design variable samples are generated using the Latin hypercube sampling method. The sampling results are shown below. Figure 5 As shown.

[0089] In step S3, to perform finite element analysis on the basic arm and obtain multidimensional structural performance index data, it is necessary to assign material properties and mesh the basic arm based on its parametric model, set boundary conditions, and add the forces calculated in the aforementioned typical working conditions as loads to establish the finite element analysis model. Then, the parameter variable data sampled by the Latin hypercube is imported into the parametric model, and the solution is run one by one. The results are then analyzed and processed, combined with the attached... Figure 6As shown. In this embodiment, the basic arm structure material parameters are set to Q345 steel, and the tetrahedral mesh patch conformal method is used for meshing. Loads and gravity are added for finite element analysis. After obtaining stress and deformation cloud maps, the maximum stress value, maximum deformation, and structural mass data are collected, as shown below. Figure 7 As shown.

[0090] In step S4, the design variable sample set and the calculated corresponding multidimensional performance index results are integrated to form the original dataset, which is then preprocessed. 70% of the data is randomly selected as the training set using the hold-out method, and the remaining 30% is used as the test set. A 10-fold cross-validation method is then used to split the training dataset. The generalization performance of the model is evaluated by repeatedly splitting the dataset, avoiding bias caused by a single random split.

[0091] In step S5, based on the aforementioned training set, the data-driven model is trained for each of the three performance indicators: maximum stress, maximum deformation, and structural quality. This embodiment selects five mainstream models for training analysis: linear regression, Gaussian process regression, random forest, support vector machine, and neural network. The support vector machine and neural network models are optimized using Bayesian optimization. The performance evaluation results of each model on the training set are shown in Tables 1, 2, and 3. The tables use root mean square error, coefficient of determination, maximum relative error, and maximum absolute error as indicators to verify model accuracy. The corresponding calculation formulas are as follows:

[0092] Root Mean Square Error (RMSE)

[0093]

[0094] Where N is the number of samples, y i For the true value, These are predicted values.

[0095] Coefficient of determination R 2

[0096]

[0097] in, The mean of the true values ​​is given. The coefficient of determination ranges from 0 to 1, with values ​​closer to 1 indicating a better model fit and values ​​closer to 0 indicating a worse model fit.

[0098] Maximum Absolute Error (MAE)

[0099]

[0100] Mean Absolute Percentage Error (MAPE)

[0101]

[0102] Among them, except R 2 In addition, smaller values ​​for RMSE, MAE, and MAPE indicate better fit.

[0103] Table 1. Performance evaluation results of each model on the maximum stress training set.

[0104]

[0105]

[0106] Table 2. Performance evaluation results of each model on the maximum deformation training set.

[0107] Model RMSE <![CDATA[R 2 ]]> MAE MAPE Linear Regression 0.1076 0.9566 0.0884 1.74％ Gaussian process regression 0.0776 0.9775 0.0602 1.17％ Random Forest 0.2068 0.8399 0.1610 3.16％ Support Vector Machine 0.1127 0.9524 0.0896 1.75％ Neural Networks 0.1674 0.8951 0.1342 2.67％

[0108] Table 3. Performance evaluation results of each model on the structural quality training set.

[0109] Model RMSE <![CDATA[R 2 ]]> MAE MAPE Linear Regression 4.49 0.9981 3.51 0.24％ Gaussian process regression 8.184 0.9939 5.36 0.38％ Random Forest 46.20 0.8021 37.53 2.61％ Support Vector Machine 3.33 0.9990 2.64 0.18％ Neural Networks 0.2078 1.0000 0.1447 0.01％

[0110] In step S6, the trained data-driven models are tested based on the aforementioned test sets. The performance evaluation results of each model on the test sets are shown in Tables 4, 5 and 6.

[0111] Table 4. Performance evaluation results of each model on the maximum stress test set.

[0112] Model RMSE <![CDATA[R 2 ]]> MAE MAPE Linear Regression 7.00 0.9143 5.94 5.01％ Gaussian process regression 4.95 0.9571 3.84 3.06％ Random Forest 4.63 0.9624 3.33 2.78％ Support Vector Machine 4.16 0.9696 3.20 2.65％ Neural Networks 4.43 0.9657 3.26 2.74％

[0113] Table 5. Performance evaluation results of each model on the maximum deformation test set.

[0114]

[0115]

[0116] Table 6. Performance evaluation results of each model on the structural quality test set.

[0117] Model RMSE <![CDATA[R 2 ]]> MAE MAPE Linear Regression 4.15 0.9980 3.25 0.22％ Gaussian process regression 6.48 0.9951 4.24 0.28％ Random Forest 41.45 0.7973 32.72 2.17％ Support Vector Machine 2.91 0.9990 2.45 0.17％ Neural Networks 0.1287 1.0000 0.1010 0.007％

[0118] Based on the evaluation metrics of the above data-driven models on the training and test sets, the support vector machine model is the best predictor of the maximum stress of the basic arm under this working condition. It has the smallest training error and the best generalization performance, with the highest R-value for both training and test data. 2All index values ​​are greater than 0.95, fully meeting the high-precision and high-reliability prediction requirements of practical engineering. Similarly, the Gaussian process regression model is the best prediction model for maximum deformation, and the neural network model is the best prediction model for structural quality. The above data also fully verify the necessity and advantages of constructing a multi-data-driven model system: by independently selecting the optimal model for different performance indicators, the limitations of a single model in achieving high accuracy and strong generalization ability across all indicators are effectively overcome, significantly improving the reliability and robustness of the overall prediction system.

[0119] In step S7, based on the selected multi-data-driven model system, a multi-objective optimization mathematical model for the basic arm is constructed using the support vector machine model for predicting maximum stress, the Gaussian process regression model for predicting maximum deformation, and the neural network model for predicting structural quality.

[0120] (1) The objective functions are: strength, maximum stress δ, stiffness, maximum deformation ε, lightweight, and structural mass M; the optimization direction is to minimize all of them.

[0121] (2) The design variables are: P1, P2, P3, P4, P5, P6, and the value range is the same as described in step S2.

[0122] (3) Constraints:

[0123] 1) The maximum predicted stress value is less than or equal to the allowable stress of the material, i.e., δ≤200MPa.

[0124] 2) The predicted maximum deformation value is less than or equal to the maximum allowable displacement, i.e., ε ≤ 10 mm.

[0125] In step S8, given that the optimization model in this embodiment has three optimization objectives, six design variables, and two constraints, it belongs to a constrained multi-objective optimization problem. Therefore, the Non-dominated Sorting Differential Evolution Algorithm (NSDE), a constrained multi-objective evolutionary algorithm, is selected as the optimization solver. The overall flow of the NSDE algorithm is as follows: Figure 14 As shown in Table 6, by setting parameters such as population size, number of iterations, and algorithm control parameters, optimization calculations were performed, ultimately yielding an optimized solution set of size 300. Figure 15 and Figure 16 It represents different trade-offs between strength, stiffness, and lightweight objectives.

[0126] Table 6 NSDE Algorithm Parameter Settings

[0127] Maximum number of generations 100 Population size 300 Tournament Scale 2 Crossover probability 0.9 Mutation probability 0.1 scaling factor 0.5 Swap probability 0.3 polynomial coefficient of variation 20

[0128] In step S9, the optimized solution set output by the NSDE algorithm is sorted and filtered for non-dominated solutions, retaining only the non-dominated solutions. In this embodiment, all 300 optimized solutions obtained are non-dominated solutions. Then, the weights of each objective are set according to the actual engineering situation. Here, it is assumed that the objective weights are equal. In this embodiment, the ideal solution method (Technique for Order Preference by Similarity to Ideal Solution, TOPSIS) from the multi-attribute decision method is used to calculate the comprehensive evaluation value of each non-dominated solution. The specific calculation process is as follows:

[0129] (1) Establish a multi-attribute decision matrix D = (d ij ) n×m , element d ij Let n be the objective function value of each solution, n be the size of the solution set (n = 300), and m be the dimension of the objective function (m = 3).

[0130] (2) Establish a standardized decision matrix R = (r ij ) n×m ,in

[0131]

[0132] (3) Establish a weighted decision matrix V = (v ij ) n×m The weight vector can be determined according to the designer's preference. For ease of calculation, this section sets the weight vector as w = (w j ) 1×m = (1 / 3, 1 / 3, 1 / 3), where

[0133] v ij =w j r ij

[0134] (4) Determine the positive ideal solution A + and negative ideal solution A - :

[0135]

[0136] In the formula v j + v j - —Calculate the minimum and maximum values ​​in the j-th column of the weighted decision matrix V, respectively.

[0137] (5) Calculate the Euclidean distance between each solution and the positive and negative ideal solutions:

[0138]

[0139] (6) Calculate the relative proximity of each scheme to the positive and negative ideal solutions:

[0140]

[0141] In the formula, 0≤C i ≤1, C i The larger the value, the farther the solution is from the negative ideal solution, and the closer it is to the positive ideal solution, indicating a better solution. The relative proximity calculation results for the non-dominated solution set are as follows: Figure 17 As shown, the maximum relative closeness is 0.7378, and the corresponding most satisfactory solutions are P1 = 21.59 mm, P2 = 20.68 mm, P3 = 49.23 mm, P4 = 46.72 mm, P5 = 295.21 mm, and P6 = 105 mm. High-precision finite element simulations were performed based on this solution. Figure 18 Finally, the comparison results of the basic arm parameter schemes and performance indicators before and after optimization are shown in Table 7.

[0142] Table 7 Comparative Analysis of Basic Arm Parameter Schemes Before and After Optimization

[0143]

[0144] As shown in Table 7, the basic boom design scheme optimized using the method described in this invention successfully achieved a synergistic improvement in multiple performance objectives while ensuring structural safety: the maximum stress was significantly reduced by 7.83%, the maximum deformation was reduced by 0.53%, and the structural mass was effectively reduced by 5.67%. These results fully verify the efficiency, feasibility, and significant benefits of the method described in this invention in the structural optimization design of aerial work platform mechanisms.

[0145] In the description of this specification, references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0146] Although embodiments of the present invention have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting the present invention. Those skilled in the art can make changes, modifications, substitutions and variations to the above embodiments within the scope of the present invention without departing from the principles and spirit of the present invention.

Claims

1. A method for optimizing the working mechanism structure of a digitally driven aerial work platform, characterized in that, Includes the following steps: Step S1: Establish a kinematic and mechanical analysis model of the working mechanism of the aerial work vehicle, and perform kinematic position solution and mechanical calculation for the working conditions to obtain the load data of each hinge point of the working mechanism. Step S2: Determine the key structural parameters that affect structural performance as design variables, sample within the design variable space, and generate a design variable sample set covering the design space; Step S3: Establish the parametric three-dimensional geometric model and finite element analysis model of the working mechanism of the aerial work vehicle; based on the hinge load data obtained in step S1, apply corresponding boundary conditions and loads to each design variable sample in the design variable sample set generated in step S2, carry out finite element analysis, and extract and record the calculation results of each structural performance. Step S4: Integrate the design variable combination sample set from step S2 with the structural performance calculation results obtained from step S3 to form the original dataset. Perform data preprocessing on the original dataset, and divide the preprocessed dataset into a training dataset and a test dataset. Step S5: Based on the training dataset, for each structure performance that needs to be optimized, multiple different types of data-driven models are selected in parallel for data-driven modeling, and hyperparameter tuning and structure design are performed for each model. Step S6: Using the test dataset, rigorously evaluate all candidate data-driven models trained in Step S5 for each structural performance; for each structural performance index, select the data-driven model with the best prediction performance and the strongest robustness as the fast calculation model for the structural performance index in subsequent optimization design; thus, a multi-data-driven model system composed of multiple independent optimal models is constructed. Step S7: Based on the multi-data-driven model system selected in Step S6, construct a multi-objective optimization mathematical model for the working mechanism of the aerial work vehicle; Step S8: Use a multi-objective intelligent optimization algorithm to solve the multi-objective optimization model established in step S7, and find a set of optimal solutions that can achieve effective balance among multiple objectives; Step S9: Perform non-dominated sorting on the optimized solution set obtained in step S8, and combine it with multi-attribute decision-making methods to select the best design scheme.

2. The method for optimizing the working mechanism structure of a digitally driven aerial work platform according to claim 1, characterized in that, In step S1, the force analysis of each hinge point of the working mechanism refers to the force and torque on each hinge point in the working mechanism of the aerial work vehicle, which is related to the working conditions of the aerial work vehicle. During the analysis, the working mechanism is divided into multiple rod groups, and a corresponding equilibrium equation is established for each rod group. The force condition of the corresponding hinge point can be calculated, and the result is recorded as the basis for the application of boundary conditions and loads in subsequent finite element analysis.

3. The method for optimizing the working mechanism structure of a digitally driven aerial work platform according to claim 1, characterized in that, In step S2, Latin hypercube sampling is performed.

4. The method for optimizing the working mechanism structure of a digitally driven aerial work platform according to claim 1, characterized in that, In step S3, the calculation results of the structural performance indicators include maximum stress, maximum deformation, and structural mass.

5. The method for optimizing the working mechanism structure of a digitally driven aerial work platform according to claim 1, characterized in that, In step S4, data preprocessing includes data cleaning and data normalization / standardization; the dataset is divided into training and testing datasets using hold-out or K-fold cross-validation.

6. The method for optimizing the working mechanism structure of a digitally driven aerial work platform according to claim 1, characterized in that, In step S5, the data-driven models include neural network models, support vector machine models, linear regression models, Gaussian process regression models, and random forest models; hyperparameter tuning includes grid search or Bayesian optimization to improve the fitting accuracy and generalization ability of each model.

7. The method for optimizing the working mechanism structure of a digitally driven aerial work platform according to claim 2, characterized in that, In step S7, the multi-objective optimization mathematical model includes establishing the objective function, designing variables and constraints; The objective functions are: strength, maximum stress δ, stiffness, maximum deformation ε, lightweight, and structural mass M; The optimization direction is always minimization; (2) The design variable is the dimensions of the bar group structure; (3) Constraints: 1) The maximum predicted stress value is less than or equal to the allowable stress of the material, i.e., δ ≤ 200 MPa; 2) The predicted maximum deformation value is less than or equal to the maximum allowable displacement, i.e., ε ≤ 10 mm.

8. The method for optimizing the working mechanism structure of a digitally driven aerial work platform according to claim 1, characterized in that, In step S9, the design scheme selection mechanism based on the multi-attribute decision-making method—ideal solution—has the following specific calculation process: (1) Establish a multi-attribute decision matrix D = (d ij ) n×m , element d ij Let n be the objective function value of each solution, n be the size of the solution set, and m be the dimension of the objective function. (2) Establish a standardized decision matrix R = (r ij ) n×m ,in (3) Establish a weighted decision matrix V = (v ij ) n×m Let the weight vector be w = (w j ) 1×m = (1 / 3, 1 / 3, 1 / 3), where v ij =w j r ij (4) Determine the positive ideal solution A + And negative ideal solution A - : In the formula v j + v j - —Calculate the minimum and maximum values ​​in the j-th column of the weighted decision matrix V, respectively; (5) Calculate the Euclidean distance between each design scheme and the positive and negative ideal solutions: (6) Calculate the relative closeness of each design scheme to the positive and negative ideal solutions: In the formula, 0≤C i ≤1, C i The larger the value, the further the design is from the negative ideal solution, and the closer it is to the positive ideal solution, indicating a better design.

Images