Anti-vibration intelligent optimization design method for transportation of large skid-mounted equipment

Abstract

The invention relates to the technical field of structural design and computational simulation, and discloses a large-scale skid-mounted equipment transportation anti-vibration intelligent optimization design method, which comprises the following steps: firstly, establishing a three-dimensional structural model of the skid-mounted equipment, simulating the vibration response of the equipment in different transportation environments based on random vibration simulation analysis, and then, through an active learning optimization algorithm, calculating the anti-vibration intelligent optimization design of the skid-mounted equipment. And the supporting points of the equipment and the positions of the damping devices are optimized, and the optimal anti-vibration design scheme is determined step by step. According to the optimization method, the finite element simulation and the active learning technology are combined, the vibration response under different design variables can be quickly and accurately predicted, the design scheme is automatically adjusted according to the optimization target, the stability and the safety of the skid-mounted equipment in the transportation process are ensured, and finally, the optimization efficiency of the skid-mounted equipment is improved. By means of the optimized design, vibration damage to equipment in the transportation process can be reduced, the weight and cost can be reduced, and the transportation efficiency is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of structural design and computational simulation, and specifically to an intelligent optimization design method for vibration resistance during the transportation of large skid-mounted equipment. Background Technique

[0002] Skid-mounted equipment is widely used in industries such as oil, natural gas, chemical engineering, and hydrogen energy. As an integrated equipment device, it has the characteristics of modularization, easy transportation, and installation. However, during the transportation of skid-mounted equipment, affected by factors such as uneven roads and traffic bumps, it often encounters strong vibration shocks. These vibrations not only affect the structural stability of the equipment but may also cause wear, damage, or accuracy deviation of the equipment's components. In severe cases, it may even lead to equipment failures, increasing maintenance costs and affecting production safety.

[0003] Existing vibration resistance designs mostly rely on experience and conventional design methods. Usually, additional shock absorbers or support points are added to the equipment frame to reduce the vibration impact. However, this method often cannot be optimized according to specific transportation conditions and is prone to design redundancy and material waste, unable to meet the multiple requirements of safety, reliability, weight, and cost in the transportation of modern skid-mounted equipment.

[0004] In view of the deficiencies of the existing technology, the present invention provides an intelligent optimization design method for vibration resistance during the transportation of large skid-mounted equipment. Summary of the Invention

[0005] In view of the deficiencies of the existing technology, the present invention provides an intelligent optimization design method for vibration resistance during the transportation of large skid-mounted equipment, solving the problems that existing vibration resistance designs mostly rely on experience and conventional design methods, cannot be optimized according to specific transportation conditions, and are prone to design redundancy and material waste.

[0006] To achieve the above objectives, the present invention is realized through the following technical solutions: An intelligent optimization design method for vibration resistance during the transportation of large skid-mounted equipment, including the following steps:

[0007] S1. Establish a vibration resistance design model for large skid-mounted equipment;

[0008] S2. Conduct random vibration simulation analysis;

[0009] S3. Define design variables and optimization objectives, and form the design variables and optimization objectives into parametric codes;

[0010] S4. Adopt an active learning method, under the Bayesian optimization framework, perform intelligent optimization design according to the design variables and the vibration resistance design model.

[0011] Preferably, the specific implementation method of step S1 is:

[0012] S11. Determine the geometric configuration and the distribution of key components of the skid-mounted equipment according to its structural characteristics and functional requirements, including pipelines, pipe clamps, bases, and supports, etc. The model includes Figure 2 as shown;

[0013] Conduct detailed modeling for each component, and the parameters include geometric dimensions such as length, diameter, thickness, mass distribution, and material properties such as elastic modulus, density, and Poisson's ratio;

[0014] S12. To simulate the vibrations in the actual transportation environment, construct a power spectral density (PSD) model of random vibration input based on road types including Class A or Class B road surfaces and vehicle transportation characteristics;

[0015] The PSD curve describes the distribution characteristics of vibration frequency and amplitude, and is usually normalized with the gravitational acceleration 9.81 m / s 2 as the reference magnitude. The generation of PSD is based on the statistical characteristics of road surface unevenness Gq(n0), and the basis is the 8-level classification standard for describing road surface unevenness formulated by the International Organization for Standardization after actual measurement;

[0016] The standard recommends that the road surface power spectral density be fitted with the following formula:

[0017]

[0018] n is the spatial frequency m-1, which is the reciprocal of the wavelength and corresponds to the time frequency Hz. n0 is the reference spatial frequency, generally n0 = 0.1, and G q (n0) is the road surface power spectral density value at the reference spatial frequency n0, also known as the road surface unevenness coefficient. W is the curve frequency exponent, generally taken as 2;

[0019] Generate corresponding vibration loading conditions for different working conditions and define them as the excitation input received by the skid-mounted equipment during transportation;

[0020] S13. After the geometric modeling is completed, discretize the skid-mounted equipment using a high-quality finite element mesh generation method to ensure the balance between model accuracy and computational efficiency;

[0021] Use a moderately sparse mesh for relatively rigid components such as bases and pipelines, and use a fine mesh for relatively flexible components such as pipe clamps and connecting components;

[0022] Construct the overall finite element model of the skid-mounted equipment by reasonably setting contact conditions and connection relationships including bonding, coplanar and co-nodal;

[0023] On this basis, set boundary conditions and support conditions to approximate the actual working conditions.

[0024] Preferably, the specific implementation method of step S2 is as follows:

[0025] S21. Use the power spectral density (PSD) curve defined in step S1 as the random vibration input and load it into the finite element model of the skid-mounted equipment;

[0026] In the model, set the excitation direction of random vibration according to the actual transportation conditions, including the vertical Y direction, and combine the support and constraint conditions of the equipment to ensure that the excitation conforms to the actual transportation situation;

[0027] The input PSD curve should be normalized and divided into multiple frequency bands according to the frequency range for convenient random vibration simulation analysis;

[0028] S22. Create a job to perform random vibration response simulation analysis and calculation on the skid-mounted equipment;

[0029] Establish the corresponding analysis steps required for simulating large skid-mounted equipment: static analysis step, frequency analysis step, and random vibration analysis step;

[0030] During the analysis process, calculate the displacement response at each unit of the overall model and the stress response at key parts, especially pay attention to the junction parts of pipes and pipe clamps, base supports, and other areas vulnerable to vibration;

[0031] Through post-processing analysis, extract the stress values of the units corresponding to the maximum stress in frequency over the full frequency range as the output curve to obtain the stress-frequency broken line graph of the full frequency band;

[0032] By comparing the maximum values of the broken line graph, extract the frequency corresponding to the maximum stress and the unit positions on the cloud diagram.

[0033] Preferably, the specific implementation method of step S3 is as follows:

[0034] S31. Based on the results of random vibration analysis in step S2, combined with stress and displacement response data, select the key parts in the skid-mounted equipment: For the selected pipe clamps, use parametric modeling technology to define the optimization variables: The moving distance of the pipe clamps at the position on the pipeline is used as the variable;

[0035] S32. After generating the result file based on the results of random vibration analysis in step S2, combined with stress and displacement response data, output the response data of stress and displacement in the model to the corresponding results. Some programs are as follows Figure 5 shown;

[0036] S33. According to the output response data results, write a C++ program to extract the maximum stress in the output data results.

[0037] Preferably, the specific implementation method of step S4 is as follows:

[0038] S41. Initialize the data by Latin hypercube sampling according to the requirements of the variables and the design domain;

[0039] S42. Take the initialized data as the modeling parameters for parametric modeling in step S2, and use the modulus prediction program in step S3 to calculate its modulus as a data set to train the surrogate model;

[0040] The surrogate model adopts a Gaussian regression surrogate model:

[0041] y(x)~N(μ(x),k(x,x′))

[0042] Select the radial basis as the kernel function, that is:

[0043]

[0044] where σ 2 is the hyperparameter and l is the length parameter;

[0045] Determine the expression of the prediction point according to the posterior probability, that is:

[0046]

[0047] Get:

[0048] y * |y~N(K * K -1 y,Cov(x))

[0049] where K is the covariance matrix, K is the covariance between the sample point x and the prediction point x * K * is the covariance matrix of the prediction point x * The expression includes the following:

[0050]

[0051] K * =[k[x1,x * ,k[x2,x * ,k[x3,x * ,…,k[x n ,x *

[0052] K ** =[k[x * ,x *

[0053] S43. Use the grey wolf algorithm with introduced constraint conditions to optimize the double addition point criterion, and add the obtained optimal point as a new point to the sample set;

[0054] ​​In the Grey Wolf Optimizer (GWO), the hunting behavior of grey wolves consists of: the encircling phase, the pursuing phase, and the attacking phase. During the encircling phase, the grey wolf population updates its position according to the following equation:

[0055]

[0056] where D is the distance between the current individual and the prey, t is the iteration number, is the coefficient vector, and X p is the position of the prey;

[0057] X w is the position of the current grey wolf individual, and the update formulas for the direction vectors and are as follows:

[0058]

[0059] where the value of a linearly decreases from 2 to 0, and r2 is a random number uniformly distributed on [0, 1];

[0060] During the pursuing phase, the α, β, and δ wolves in the Grey Wolf Optimizer lead other grey wolf individuals to move towards the area where the prey is located in the search space. Therefore, the selection of the α, β, and δ wolves determines whether the algorithm can find the optimal value. When optimizing and solving, each grey wolf corresponds to a target value. Thus, in the optimization selection, the best wolf, the second-best wolf, and the third-best wolf in the current grey wolf population are selected as the α, β, and δ wolves for the current iteration step;

[0061] During the hunting phase, the position of the δ wolf is updated based on the positions of the α, β, and δ wolves. The position update formula for the grey wolf individual during the prey pursuit phase is as follows:

[0062] D α = |C1·X α - X|

[0063] D β = |C2·X β - X|

[0064] D δ = |C3·X δ - X|

[0065]

[0066] where C1, C2, and C3 are the three components of C, X is the position of the current solution, X α is the position of the best solution α wolf in the population, X β is the position of the second-best solution β wolf in the population, and X δ is the position of the third-best solution δ wolf in the population;

[0067] In the attack phase, the gray wolf population completes the process of attacking prey according to the change of the value of a. The initial value of a is 2. During the process of a decreasing to 0, according to the formula A = 2·a·r1 - a, the corresponding value also changes within the interval [-a, a].

[0068]

[0069] It can be seen that the value range of shrinks gradually. When a decreases from 2 to 1, the search range of gray wolf individuals is relatively large, and at this time the algorithm mainly conducts global search. When a decreases from 1 to 0, the search range of gray wolf individuals gradually shrinks, and at this time the local search ability of the algorithm is enhanced;

[0070] S44. Judge whether the stored optimal value converges. If it converges, end the optimization process;

[0071] Use the gray wolf algorithm with introduced constraint conditions to optimize the double - adding point criterion, and take the obtained optimal point as a new point to be added to the sample set;

[0072] In the invention, a double - point adding point criterion is adopted. The expression of the adding point criterion includes the following:

[0073]

[0074] In the formula, k is a control coefficient. When optimizing the maximum value, k takes 1, and when optimizing the minimum value, k takes - 1;

[0075] Through this adding point criterion, two points will be generated in each round of iteration during the optimization process. One point represents the point closest to the extreme point under the current surrogate model, and one point represents the point with the highest uncertainty under the current surrogate model. In this way, it meets the consideration of the development ability and search ability of the optimization algorithm. However, during the optimization calculation process, the same points may be obtained through the optimization adding point criterion at different iteration steps. Calculating the same sample point not only increases the calculation cost but also does not improve the accuracy of the surrogate model. Therefore, in the Bayesian algorithm under the double - point adding point criterion, such repeated points need to be removed to avoid repeated calculation and addition to the dataset.

[0076] Preferably, the step S4 includes S45: Repeat steps S42 - S44 until the iteration converges;

[0077] Use the Bayesian optimization algorithm to optimize the parameters of the pipe clamp's position on the pipeline. First, generate 10 initial sample points through Latin hypercube sampling, calculate the objective function values corresponding to the sample points through finite - element simulation, and establish an initial surrogate model, and set the number of Bayesian iterations to 100 times;

[0078] The gray wolf algorithm is used to optimize the dotting criterion. The population size of the gray wolf algorithm is set to 100, and the number of iterations is set to 500.

[0079] Preferably, the method can adaptively optimize the anti-vibration design of the skid-mounted equipment by combining the Bayesian optimization algorithm, the gray wolf algorithm and the active learning technology, automatically select the optimal design parameters, and reduce the time cost of manual intervention and repeated trial and error.

[0080] Preferably, the optimized design can significantly reduce the impact of vibration during transportation on the equipment. The optimized design not only improves the anti-vibration performance, but also reduces the equipment weight and transportation cost. During the optimization process, the optimized movement of the pipe clamp position is 0.16954 meters, and the maximum stress value after optimization is effectively reduced by 59.86 Mpa.

[0081] The present invention provides an intelligent optimization design method for anti-vibration during the transportation of large skid-mounted equipment. It has the following beneficial effects:

[0082] 1. By combining the ABAQUS simulation and the active learning optimization algorithm, the present invention provides an efficient anti-vibration design method, achieving a significant improvement in the anti-vibration performance of large skid-mounted equipment during transportation. Compared with the traditional design method, the present invention can accurately predict and avoid potential vibration problems at the initial stage of design, avoiding the adverse effects of vibration that cannot be accurately foreseen when using empirical and conventional design methods.

[0083] 2. The present invention uses the ABAQUS simulation technology to establish an accurate vibration model and simulate the vibration response under different pipe clamp positions. This high-precision simulation method can more accurately identify the vibration transmission path and optimize the design compared with the traditional empirical calculation method, reducing the occurrence of resonance phenomenon and greatly improving the stability and safety of the equipment.

[0084] 3. Through the active learning optimization algorithm, the present invention can intelligently adjust the design scheme according to the feedback data after each simulation and automatically select the optimal design parameters. Different from the traditional method of manually adjusting the design scheme, the present invention reduces manual intervention and trial and error time, accelerates the design process, and improves work efficiency and design accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0085] Figure 1 is a flowchart of an intelligent optimization method for anti-vibration design of large skid-mounted equipment during transportation according to the present invention;

[0086] Figure 2 is a geometric schematic diagram of the large skid-mounted equipment according to the present invention;

[0087] Figure 3 is a schematic diagram of the road surface unevenness standard of the large skid-mounted equipment in the present invention;

[0088] Figure 4 Schematic diagram of partial code for variable parameterization of large-scale skid-mounted equipment in the present invention;

[0089] Figure 5 Schematic diagram of partial code for extracting and processing data parameters of large-scale skid-mounted equipment in the present invention;

[0090] Figure 6 Schematic diagram of partial code for reading data in the present invention;

[0091] Figure 7 Flow chart of Latin hypercube sampling in the present invention;

[0092] Figure 8 Flow chart of grey wolf algorithm in the present invention;

[0093] Figure 9 Flow chart of double addition point criterion in the present invention;

[0094] Figure 10 Schematic diagram of the iteration process in the present invention;

[0095] Figure 11 Schematic diagram of the optimization result in the present invention;

[0096] Figure 12 Schematic diagram of stress comparison before and after optimization in the present invention. Detailed implementation manners

[0097] Next, in combination with the accompanying drawings of the present invention specification, the technical solutions in the embodiments of the present invention will be clearly and completely described. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments in the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0098] Please refer to the attached Figure 1 , an embodiment of the present invention provides a transportation anti-vibration intelligent optimization design method for large-scale skid-mounted equipment, including the following steps:

[0099] S1. Establish an anti-vibration design model for large-scale skid-mounted equipment;

[0100] S2. Conduct random vibration simulation analysis;

[0101] S3. Define design variables and optimization objectives, and form the design variables and optimization objectives into parameterized codes;

[0102] S4. Adopt an active learning method, and under the Bayesian optimization framework, conduct intelligent optimization design according to the design variables and the anti-vibration design model.

[0103] The specific implementation method of the step S1 is as follows:

[0104] S11. According to the structural characteristics and functional requirements of the skid-mounted equipment, determine its geometric configuration and the distribution of key components, including pipelines, pipe clamps, bases, and supports, etc. The model includes Figure 2 as shown;

[0105] Conduct detailed modeling for each component. The parameters include geometric dimensions such as length, diameter, thickness, mass distribution, and material properties such as elastic modulus, density, and Poisson's ratio;

[0106] S12. To simulate the vibration in the actual transportation environment, based on the road types including Class A or Class B pavements and the vehicle transportation characteristics, construct a power spectral density (PSD) model of random vibration input;

[0107] The PSD curve describes the distribution characteristics of vibration frequency and amplitude, and is usually normalized with the gravitational acceleration 9.81 m / s 2 as the reference magnitude. The generation of PSD is based on the statistical characteristics of road surface unevenness Gq(n0), and the basis is the eight-level classification standard for describing road surface unevenness formulated by the International Organization for Standardization after actual measurement;

[0108] The standard recommends that the road surface power spectral density be fitted with the following formula:

[0109]

[0110] n is the spatial frequency m-1, which is the reciprocal of the wavelength and corresponds to the time frequency Hz. n0 is the reference spatial frequency, generally n0 = 0.1, G q (n0) is the road surface power spectral density value at the reference spatial frequency n0, also known as the road surface unevenness coefficient. W is the curve frequency exponent, generally taken as 2;

[0111] Generate corresponding vibration loading conditions for different working conditions and define them as the excitation input received by the skid-mounted equipment during transportation;

[0112] S13. After the geometric modeling is completed, use a high-quality finite element mesh generation method to discretize the skid-mounted equipment to ensure the balance between model accuracy and calculation efficiency;

[0113] Adopt a moderately sparse mesh generation for the components with relatively large rigidity, including the base and pipelines, and a fine mesh generation for the components with relatively large flexibility, including pipe clamps and connecting components;

[0114] By reasonably setting the contact conditions and connection relationships, including bonding, coplanar and co-nodal, construct the overall finite element model of the skid-mounted equipment;

[0115] On this basis, set the boundary conditions and support conditions to approach the actual working conditions.

[0116] The specific implementation method of step S2 is as follows:

[0117] S21. Take the power spectral density (PSD) curve defined in step S1 as the random vibration input and load it into the finite element model of the skid-mounted equipment;

[0118] In the model, set the excitation direction of random vibration according to the actual transportation conditions, including the vertical Y direction, and combine the support and constraint conditions of the equipment to ensure that the excitation conforms to the actual transportation situation;

[0119] The input PSD curve should be normalized and divided into multiple frequency bands according to the frequency range for convenient random vibration simulation analysis;

[0120] S22. Create a job to perform random vibration response simulation analysis and calculation on the skid-mounted equipment;

[0121] Establish the corresponding analysis steps required for simulating large skid-mounted equipment: static analysis step, frequency analysis step, and random vibration analysis step;

[0122] During the analysis process, calculate the displacement response at each unit of the overall model and the stress response at key parts, especially paying attention to areas vulnerable to vibration such as the junction of pipes and pipe clamps, and the base support;

[0123] Through post-processing analysis, extract the stress values of the units corresponding to the maximum stress in frequency over the full frequency range as the output curve to obtain the stress-frequency broken line graph for the full frequency band;

[0124] By comparing the maximum values of the broken line graph, extract the frequency corresponding to the maximum stress and the unit positions on the contour map.

[0125] The specific implementation method of step S3 is as follows:

[0126] S31. Based on the results of random vibration analysis in step S2, combined with stress and displacement response data, select the key parts in the skid-mounted equipment: pipe clamps. For the selected pipe clamps, use parametric modeling technology to define the optimization variables: the moving distance of the pipe clamps at the position on the pipe as the variable, as shown in some programs such as Figure 4 shown;

[0127] S32. After generating the result file based on the results of random vibration analysis in step S2, combined with stress and displacement response data, output the response data of stress and displacement in the model to the corresponding results. Some programs include Figure 5 shown;

[0128] S33. According to the output response data results, write a C++ program to extract the maximum stress in the output data results, as shown in some programs such as Figure 6 shown.

[0129] The specific implementation method of step S4 is as follows:

[0130] S41. Initialize the data through Latin Hypercube Sampling according to the requirements of variables and the design domain. The Latin Hypercube Sampling process is as Figure 7 ;

[0131] S42. Use the initialized data as the modeling parameters for parametric modeling in step S2, and calculate its modulus as a data set using the modulus prediction program in step S3 to train the surrogate model;

[0132] The surrogate model adopts a Gaussian regression surrogate model:

[0133] y(x)~N(μ(x),k(x,x′))

[0134] Select the radial basis as the kernel function, that is:

[0135]

[0136] where σ 2 is the hyperparameter and l is the length parameter;

[0137] Determine the expression of the prediction point according to the posterior probability, that is:

[0138]

[0139] We get:

[0140] y * |y~N(K * K -1 y,Cov(x))

[0141] where K is the covariance matrix, K is the covariance between the sample point x and the prediction point x * and K * is the covariance matrix of the prediction point x * . The expression includes the following:

[0142]

[0143] K * =[k[x1,x * ,k[x2,x * ,k[x3,x * ,…,k[x n ,x *

[0144] K ** =[k[x * ,x *

[0145] ​​S43. The grey wolf algorithm with introduced constraint conditions is used to optimize the double addition point criterion, and the obtained optimal point is added to the sample set as a new point. The principle of the grey wolf algorithm is as follows Figure 8 ;

[0146] In the grey wolf algorithm, the hunting behavior of grey wolves includes: the encirclement stage, the pursuit stage, and the attack stage; in the encirclement stage, the grey wolf group updates its position according to the formula:

[0147]

[0148] where D is the distance between the current individual and the prey, t is the number of iterations, is the coefficient vector, X p is the position of the prey;

[0149] X w is the position of the current grey wolf individual, and the update formulas of the direction vectors and include the following:

[0150]

[0151] where the value of a linearly decreases from 2 to 0, and r2 is a random number uniformly distributed on [0,1];

[0152] In the pursuit stage, the α, β, and δ wolves in the grey wolf optimization algorithm lead other grey wolf individuals to move towards the area close to the prey in the search space. Therefore, the selection of the α, β, and δ wolves determines whether the algorithm can find the optimal value. In the optimization solution, each grey wolf corresponds to a target value. Therefore, in the optimization selection, the optimal wolf, the second-best wolf, and the third-best wolf in the current grey wolf population are selected as the α, β, and δ wolves in the current iteration step;

[0153] In the hunting stage, the position of the δ wolf is updated according to the positions of the α, β, and δ wolves. The position update formula of the grey wolf individual in the prey pursuit stage is as follows:

[0154] D α =|C1·X α -X|

[0155] D β =|C2·X β -X|

[0156] D δ =|C3|X δ -X|

[0157]

[0158] where C1, C2, and C3 are the three components of C, X is the position of the current solution, and X αis the position of the alpha wolf, the optimal solution of the population, X β is the position of the beta wolf, the second-best solution of the population, X δ is the position of the delta wolf, the third-best solution of the population;

[0159] In the attack phase, the grey wolf population completes the process of attacking prey according to the change of the value of a. The initial value of a is 2. During the process of a decreasing from 2 to 0, according to the formula A = 2·a·r1 - a, its corresponding value also changes within the interval [-a, a],

[0160]

[0161] It can be seen that the value range of shrinks gradually. When a decreases from 2 to 1, the search range of grey wolf individuals is relatively large. At this time, the algorithm mainly conducts global search. When a decreases from 1 to 0, the search range of grey wolf individuals gradually shrinks, and at this time, the local search ability of the algorithm is enhanced;

[0162] S44. Judge whether the stored optimal value converges. If it converges, end the optimization process;

[0163] The grey wolf algorithm with introduced constraint conditions is used to optimize the double-point addition criterion, and the obtained optimal point is added to the sample set as a new point, as Figure 9 shown;

[0164] In the invention, a double-point addition criterion is adopted, and the expression of the addition criterion includes the following:

[0165]

[0166] In the formula, k is a control coefficient. When optimizing the maximum value, k takes 1, and when optimizing the minimum value, k takes -1;

[0167] Through this addition criterion, two points will be generated in each round of iteration in the optimization process. One point represents the point closest to the extreme point under the current surrogate model, and one point represents the point with the highest uncertainty under the current surrogate model. In this way, the optimization algorithm takes into account both the exploitation ability and the search ability. However, in the optimization calculation process, the same point may be obtained through the optimization addition criterion at different iteration steps. Calculating the same sample point not only increases the calculation cost, but also does not improve the accuracy of the surrogate model. Therefore, in the Bayesian algorithm under the double-point addition criterion, this kind of repeated point needs to be removed to avoid repeated calculation and addition to the dataset.

[0168] The step S4 includes S45: Repeat steps S42 - S44 until the iteration converges;

[0169] The Bayesian optimization algorithm is used to optimize the parameters of the pipe clamp's position on the pipeline. First, 10 initial sample points are generated through Latin hypercube sampling. The objective function values corresponding to the sample points are obtained through finite element simulation calculations, and an initial surrogate model is established. The number of Bayesian iterations is set to 100 times;

[0170] For the optimization of the sampling criterion, the Grey Wolf algorithm is used. The population size of the Grey Wolf algorithm is set to 100, and the number of iterations is set to 500. Figure 10 This is the schematic diagram at the beginning of the iteration.

[0171] In the finally found optimization solution, the position of the pipe clamp has moved forward by 0.16954 meters relative to the initial structure, as Figure 11 shown. Figure 12 This is the comparison of the stress nephograms near the pipe clamp before and after optimization. Compared with the stress nephogram before optimization, the maximum stress has effectively decreased by 59.86 Mpa, reduced to 73 Mpa, and the anti-vibration performance has been improved by 44%.

[0172] The method can adaptively optimize the anti-vibration design of skid-mounted equipment through the Bayesian optimization algorithm, combined with the Grey Wolf algorithm and active learning technology, automatically select the optimal design parameters, and reduce the time cost of manual intervention and repeated trial and error.

[0173] The optimized design can significantly reduce the impact of vibration during transportation on the equipment. The optimized design not only improves the anti-vibration performance but also reduces the equipment weight and transportation cost. During the optimization process, the position of the pipe clamp has been optimized and moved by 0.16954 meters, and the maximum stress after optimization has effectively decreased by 59.86 Mpa.

[0174] First of all, simulation is an important part. The three-dimensional structure of the skid-mounted equipment is accurately modeled using the ABAQUS simulation software, considering the geometric characteristics, material properties, etc. of each component to ensure the accuracy of the simulation. These components include the base, pipeline, pipe clamp, support points, etc. They may be affected by vibration during transportation. Through finite element analysis of these components, the simulation can simulate the vibration responses under different conditions. These simulation results provide basic data for subsequent optimization. Especially in terms of vibration frequency and amplitude, the simulation can help us identify and predict which parts are likely to be affected by vibration during transportation and adjust the design scheme in a timely manner.

[0175] Traditional anti-vibration design methods usually rely on experience to determine the positions of support points and the layout of shock-absorbing devices. However, these methods often ignore the specific impacts under different conditions during transportation. By using ABAQUS simulation, it is possible to simulate the vibration environment at the initial stage of design and input the simulation data into an active learning optimization algorithm. The core of the active learning algorithm is its ability to automatically adjust design variables based on the feedback after each simulation. After each simulation, the algorithm will intelligently update the positions of pipe clamps, the layout of support points, and the optimal positions of shock-absorbing devices according to the new data, thereby optimizing the design and reducing the vibration amplitude.

[0176] The introduction of the active learning optimization algorithm gives this invention more advantages in dealing with complex design tasks. Traditional optimization methods rely on manual experience or fixed design processes, which are often inefficient and difficult to cope with complex and changeable transportation environments. In contrast, the optimization algorithm of this invention can adaptively adjust and obtain the optimal solution through repeated trials, reducing the design cycle and manual intervention. The simulation results after each optimization are used to adjust the model, and the algorithm can automatically select the most suitable design scheme after each simulation. This not only improves the design accuracy but also greatly saves time and costs.

[0177] In the design process, structural modeling is required first, including the modeling of various key components such as pipes, bases, supports, and shock absorbers. The geometric dimensions, mass distribution, and material properties of each component, such as elastic modulus, density, etc., need to be defined in detail. During the simulation process, the vibration responses of all components will be accurately simulated, especially the parts that are easily affected by vibration, such as the connection between pipes and pipe clamps. According to the simulation results, we can extract the vibration data of these parts and make targeted adjustments during the optimization process to ensure that the vibration is effectively reduced.

[0178] The main goal of the optimized design is to reduce the vibration amplitude of the equipment during transportation by adjusting the support points, pipe clamp positions, and the layout of shock-absorbing devices. Through the active learning algorithm, designers do not need to repeatedly try and error. The algorithm will automatically select the optimal design scheme. After each optimization, the simulation results will verify the effectiveness of the design. If the optimization effect does not meet the expectations, the algorithm will adjust again according to the new feedback data until the optimal vibration suppression effect is achieved.

[0179] In practical applications, it can effectively reduce the vibration amplitude of the equipment, making it less likely to be damaged during transportation and reducing the maintenance cost. Moreover, the optimized design not only improves the anti-vibration performance but also can reduce the weight of the equipment, lower the transportation cost, and further improve the transportation efficiency. It is applicable to heavy equipment that requires long-distance transportation and faces large vibration impacts, such as the transportation of large equipment in industries like hydrogen energy, petroleum, and chemical engineering.

[0180] Although embodiments of the present invention have been shown and described, those of ordinary skill in the art will appreciate that various changes, modifications, substitutions, and variations can be made to these embodiments without departing from the principles and spirit of the present invention. The scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A vibration-resistant intelligent optimization design method for transporting large-scale skid-mounted equipment, characterized in that It includes the following steps: S1. Establish an anti-vibration design model for large-scale skid-mounted equipment; S2. Conduct random vibration simulation analysis; S3. Define design variables and optimization objectives, and form the design variables and optimization objectives into parametric codes; S4. Adopt an active learning method, under the Bayesian optimization framework, conduct intelligent optimization design according to the design variables and the anti-vibration design model.

2. The intelligent optimization design method for vibration resistance during transportation of a large-scale skid-mounted equipment according to claim 1, wherein, The specific implementation method of step S1 is as follows: S11. According to the structural characteristics and functional requirements of the skid-mounted equipment, determine its geometric configuration and the distribution of key components, including pipelines, pipe clamps, bases and supports; Conduct detailed modeling for each component, and the parameters include geometric dimensions such as length, diameter, thickness, mass distribution and material properties such as elastic modulus, density and Poisson's ratio; S12. To simulate the vibration in the actual transportation environment, based on road types including Class A or Class B roads and vehicle transportation characteristics, construct a power spectral density (PSD) model for random vibration input; The PSD curve describes the distribution characteristics of vibration frequency and amplitude, usually normalized with the gravitational acceleration 9.81 m / s 2 as the reference magnitude. The generation of PSD is based on the statistical characteristics of road surface unevenness Gq(n0), according to the 8-level classification standard for describing road surface unevenness formulated by the International Organization for Standardization after actual measurement; The power spectral density of the standard recommended road surface is fitted by the following formula: n is the spatial frequency in m⁻¹, which is the reciprocal of the wavelength and corresponds to the temporal frequency in Hz. n₀ is the reference spatial frequency. Generally, n₀ = 0.1, G q G(n₀) is the road surface power spectral density value at the reference spatial frequency n₀, also known as the road surface unevenness coefficient. W is the curve frequency index, generally taken as 2; Generate corresponding vibration loading conditions for different working conditions, and define them as the excitation input received by the skid-mounted equipment during transportation; S13. After the geometric modeling is completed, use a high-quality finite element mesh generation method to discretize the skid-mounted equipment to ensure the balance between model accuracy and calculation efficiency; Adopt a moderately sparse mesh generation for components with relatively high rigidity, including bases and pipelines, and adopt a fine mesh generation for components with relatively high flexibility, including pipe clamps and connecting components; By reasonably setting contact conditions and connection relationships including bonding, coplanar and co-nodal, construct the overall finite element model of the skid-mounted equipment; On this basis, set boundary conditions and support conditions to approximate the actual working conditions.

3. A method for intelligent optimization design of anti-vibration in the transportation of large-scale skid-mounted equipment according to claim 1, characterized in that The specific implementation method of step S2 is as follows: S21. Use the PSD curve defined in step S1 as the random vibration input and load it into the finite element model of the skid-mounted equipment; In the model, set the excitation direction of random vibration according to the actual transportation working conditions, including the vertical Y direction, and combine the support and constraint conditions of the equipment to ensure that the excitation conforms to the actual transportation situation; The input PSD curve should be normalized and divided into multiple frequency bands according to the frequency range for convenient random vibration simulation analysis; S22. Create a job to conduct random vibration response simulation analysis and calculation on the skid-mounted equipment; Establish the corresponding analysis steps required for simulating the large-scale skid-mounted equipment: static analysis step, frequency analysis step, random vibration analysis step; During the analysis process, calculate the displacement response at each unit of the overall model and the stress response at key parts, especially pay attention to the junction parts of pipelines and pipe clamps, and the areas of the base supports that are vulnerable to vibration; Through post-processing analysis, extract the stress values of the units corresponding to the maximum stress in frequency over the full frequency as the output curve to obtain the stress-frequency broken line graph of the full frequency band; By comparing the maximum values of the broken line graph, extract the frequency corresponding to the maximum stress and the unit positions on the nephogram.

4. A method for intelligent optimization design of anti-vibration in the transportation of large-scale skid-mounted equipment according to claim 1, characterized in that The specific implementation method of step S3 is as follows: S31. Based on the results of the random vibration analysis in step S2 and combined with the stress and displacement response data, identify the key parts in the skid-mounted equipment: pipe clamps. For the selected pipe clamps, use parametric modeling technology to define the optimization variables: the moving distance of the pipe clamps at their positions on the pipeline is taken as a variable; S32. After generating the result file based on the results of the random vibration analysis in step S2 and combined with the stress and displacement response data, output the stress and displacement response data in the model to the corresponding results. Part of the program is shown in Figure 5; S'33. According to the output response data results, write a C++ program to extract the maximum stress in the output data results.

5. A method for intelligent optimization design of vibration resistance in the transportation of large-scale skid-mounted equipment according to claim 1, characterized in that, The specific implementation method of step S4 is as follows: S41. Initialize the data through Latin hypercube sampling according to the requirements of the variables and the design domain; S42. Use the initialized data as the modeling parameters for the parametric modeling in step S2, and use the modulus prediction program in step S3 to calculate its modulus as a data set to train the surrogate model; The surrogate model adopts a Gaussian regression surrogate model: y(x)~N(μ(x),k(x,x′)) Select the radial basis as the kernel function, that is: where σ 2 is a hyperparameter and l is a length parameter; Determine the expression of the prediction point according to the posterior probability, that is: Obtain: y * | y ∼ N(K * K -1 y, Cov(x)) where K is the covariance matrix, and K is the covariance between the sample point x and the prediction point x * ; K * is the covariance matrix of the prediction point x * , and the expression includes the following: K * = [k[x1, x * , k[x2, x * , k[x3, x * , …, k[x n , x * ​ K ** = [k[x * , x * ​ S43. Use the grey wolf algorithm with introduced constraint conditions to optimize the double-point addition criterion, and add the obtained optimal point as a new point to the sample set; In the grey wolf algorithm, the hunting behavior of grey wolves includes: the encirclement stage, the pursuit stage, and the attack stage; in the encirclement stage, the grey wolf group updates its position according to the formula: where D is the distance between the current individual and the prey, and t is the number of iterations. is the coefficient vector, and X p is the position of the prey. X w is the position of the current gray wolf individual, and the update formulas for the direction vectors and are as follows: Among them, the value of a linearly decreases from 2 to 0, and r2 is a random number uniformly distributed on [0,1]; In the pursuit stage, the α, β, and δ wolves in the grey wolf optimization algorithm lead other grey wolf individuals to move towards the area close to the prey in the search space. Therefore, the selection of the α, β, and δ wolves determines whether the algorithm can find the optimal value. When optimizing and solving, each grey wolf corresponds to a target value. Therefore, in the optimization selection, the optimal wolf, the second-best wolf, and the third-best wolf in the current grey wolf population are selected as the α, β, and δ wolves in the current iteration step; In the hunting stage, update the position of the δ wolf according to the positions of the α, β, and δ wolves. The position update formula for the grey wolf individual in the prey pursuit stage is as follows: D α = |C1·X α - X| D β = |C2·X β - X| D δ = |C3·X δ - X| Among them, C1, C2, and C3 are the three components of C, X is the position of the current solution, and X α is the position of the alpha wolf, the optimal solution in the population, and X β is the position of the beta wolf, the second-best solution in the population, and X δ is the position of the delta wolf, the third-best solution in the population; S44. Determine whether the stored optimal value converges. If it converges, end the optimization process; Use the grey wolf algorithm with introduced constraint conditions to optimize the double-point addition criterion, and add the obtained optimal point as a new point to the sample set; In the invention, a double-point addition criterion is adopted. The expression of the addition criterion is as follows: In the formula, k is a control coefficient. When optimizing the maximum value, k takes 1, and when optimizing the minimum value, k takes -1.

6. The intelligent optimization design method for vibration resistance during transportation of a large-scale skid-mounted equipment according to claim 1, characterized in that Step S4 includes S45: Repeat steps S42 to S44 until the iteration converges; Use the Bayesian optimization algorithm to optimize the parameters of the position of the pipe clamp on the pipeline. First, generate 10 initial sample points through Latin hypercube sampling, obtain the objective function values corresponding to the sample points through finite element simulation calculation, and establish an initial surrogate model. Set the number of Bayesian iterations to 100 times; For the optimization of the addition criterion, use the grey wolf algorithm. The population size of the grey wolf algorithm is set to 100, and the number of iterations is set to 500.

7. A method for intelligent optimization design of vibration resistance during transportation of a large-scale skid-mounted equipment according to claim 1, characterized in that, The method can adaptively optimize the anti-vibration design of skid-mounted equipment through the Bayesian optimization algorithm, combined with the grey wolf algorithm and active learning technology, automatically select the optimal design parameters, and reduce the time cost of manual intervention and repeated trial and error.

8. A method for intelligent optimization design of vibration resistance during transportation of a large-scale skid-mounted equipment according to claim 1, characterized in that, The optimized design can significantly reduce the impact of vibration during transportation on the equipment. The optimized design not only improves the anti-vibration performance but also reduces the equipment weight and transportation cost. During the optimization process, the optimized movement of the pipe clamp position is 0.16954 meters, and the maximum stress after optimization is effectively reduced by 59.86 Mpa.

Images