A model-driven harness development and optimization method

Abstract

The application provides a model-driven wire harness development and optimization method, relates to the technical field of wire harness design and optimization, and accurately identifies potential stress concentration areas by collecting actual operation parameters to construct a digital twin model and subdividing the model into a large number of node units for accurate stress analysis. In this way, the problem that the existing wiring scheme is prone to causing stress concentration and leading to wire harness damage is solved.

Description

Technical Field

[0001] This invention relates to the field of wire harness design and optimization technology, and more specifically, to a model-driven method for wire harness development and optimization. Background Technology

[0002] In recent years, with the rapid development of industries such as automotive and aerospace, wire harnesses, as a key component of electrical systems, have received increasing attention for their design and optimization. Traditional wire harness design primarily relies on experience and two-dimensional drawing tools. While this can meet the needs of certain simple applications, with the increasing complexity of modern industrial systems, this design approach has gradually revealed its inability to cope with multi-dimensional performance requirements. To address this issue, model-driven design methods have gradually become a new direction in wire harness development in recent years. Model-driven methods simulate the geometry, physical characteristics, and operating states of wire harnesses by constructing digital models, providing powerful tools for performance analysis, optimization design, and fault prediction. Among these, the introduction of digital twin technology has further promoted the innovative development of wire harness technology. Digital twin technology utilizes data acquisition, physical modeling, and dynamic simulation to map the actual operating state of the wire harness into a virtual model, achieving full lifecycle management of the wire harness from design to operation. Theoretically, this method can improve the accuracy of wire harness design, optimize wiring paths, and enhance the operational reliability of wire harnesses under actual working conditions.

[0003] However, existing model-driven wire harness development technologies still have certain shortcomings. First, most current wire harness design methods focus only on geometric routing and space occupation, neglecting key mechanical and thermal factors such as stress distribution, vibration effects, and temperature rise during actual operation. This can lead to designs that fail to meet reliability requirements under complex operating conditions. Second, existing technologies often rely on static models for routing path optimization, lacking stress analysis and path optimization methods under dynamic operating conditions. This can result in optimization results that do not match the actual usage environment. Furthermore, existing methods largely depend on human experience for the design of fixed support positions and protective structures, lacking systematic optimization strategies and making it difficult to efficiently address stress concentration issues. These shortcomings not only limit the accuracy and efficiency of wire harness design but may also lead to fatigue failure, breakage, or electrical faults in actual operation. Summary of the Invention

[0004] To address the aforementioned technical problems, this invention is proposed. This invention provides a model-driven method for wire harness development and optimization, which to some extent solves the problem that existing wiring schemes, due to their inaccurate analysis of stress distribution under actual operating conditions, easily lead to stress concentration and wire harness damage.

[0005] According to one aspect of the present invention, a model-driven method for developing and optimizing wire harnesses is provided, comprising:

[0006] Collect the actual operating parameters of the wiring harness, and construct a digital twin model of the wiring harness based on the actual operating parameters;

[0007] The digital twin model of the wire harness is divided into N node units, and stress analysis is performed on the N node units to obtain stress distribution data for each node unit.

[0008] Based on the stress distribution data, the wiring path of the wire harness is optimized;

[0009] Based on the optimization results of the wiring path, the position of the fixing bracket of the wire harness is adjusted, and a protective structure is added at the node unit where the stress distribution data exceeds the preset stress threshold.

[0010] Furthermore, the wire harness digital twin model integrates the physical property database, geometric property database, and material property database through a data mapping algorithm.

[0011] Furthermore, the vibration transfer function of each segment of the wire harness is calculated through the physical feature sub-vector in the digital feature vector. If the amplitude of the vibration transfer function of a certain segment of the wire harness exceeds the safety limit, the stiffness parameter of that segment of the wire harness is stored in the physical property database.

[0012] The digital feature vector is obtained by fusing the actual operating parameters through a neural network, and then by orthogonal decomposition to obtain physical feature sub-vectors characterizing the physical properties of the wire harness, geometric feature sub-vectors characterizing the shape of the wire harness, and material feature sub-vectors characterizing the material of the wire harness.

[0013] Furthermore, the geometric feature sub-vectors are used to construct the three-dimensional spatial curve equation of the wire harness. When the curvature of the three-dimensional spatial curve at any point exceeds the safe curvature, the stress concentration factor at that point is calculated and stored in the geometric characteristic database of the wire harness digital twin model.

[0014] Furthermore, constructing the equation of the three-dimensional space curve includes:

[0015] The geometric feature sub-vectors are dimensionality reduced to extract the key control point coordinate sequence of the wire bundle in three-dimensional space;

[0016] When the spacing between control points is less than a preset distance d, control point filtering is performed;

[0017] Based on the control point coordinate sequence, the three-dimensional spatial curve equation is constructed.

[0018] Furthermore, obtaining the stress distribution data for each node element includes:

[0019] Based on the location information of each node unit, the corresponding boundary conditions and constraints are extracted.

[0020] After setting the boundary conditions, different constitutive equations are constructed based on the material structure of each node element.

[0021] The deformation of the nodal elements is calculated based on the displacement boundary conditions, and the strain components in each direction are calculated based on the deformation.

[0022] Substituting the strain components into the constitutive equation yields the stress components of each nodal element.

[0023] Furthermore, when the node unit is a composite material, a set of layered constitutive equations is established;

[0024] The set of constitutive equations for the layered structure determines the number of layers in the composite material and the material properties of each layer, records the layup angles, and treats the core layer as an independent layer, establishing a local coordinate system and a global coordinate system.

[0025] If the two coordinate systems do not coincide, the constitutive equations of each layer are transformed from the local coordinate system to the global coordinate system using the coordinate transformation matrix. Finally, based on the principle of equivalent stiffness superposition, the constitutive equations of each layer are combined into a global stiffness matrix, thereby obtaining a set of layered constitutive relation equations describing the mechanical behavior of composite material nodal elements.

[0026] Furthermore, the ant colony algorithm is used to optimize the wiring path of the wire harness;

[0027] The ant colony algorithm uses the entropy weight method to dynamically adjust the stress weight of different node units, and adopts the pheromone decay factor to avoid the ant colony algorithm getting trapped in local optima.

[0028] Furthermore, the entropy weighting method dynamically adjusts the stress weights of different node elements, including:

[0029] Calculate the stress percentage for each node element and calculate the information entropy based on these percentages;

[0030] The difference coefficient is obtained by subtracting the information entropy from one, and the initial weight is obtained by normalizing the difference coefficient.

[0031] The weights are dynamically adjusted based on the relationship between the stress value of the node element and a preset threshold.

[0032] Furthermore, when the optimal solution remains unchanged for multiple consecutive iterations, it is determined that it has entered a local optimum. By setting different pheromone decay factors, the pheromone differences between paths are accelerated and eliminated, thereby escaping the local optimum and finding a better solution.

[0033] Compared with existing technologies, the model-driven wire harness development and optimization method provided by this invention constructs a digital twin model by collecting actual operating parameters and subdivides the model into a large number of node units for precise stress analysis, which can accurately identify potential stress concentration areas. This solves the problem that existing wiring solutions, in their precise analysis of stress distribution under actual operating conditions, easily lead to stress concentration and wire harness damage. Attached Figure Description

[0034] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort. In the drawings:

[0035] Figure 1 This is a flowchart of a model-driven wire harness development and optimization method according to an embodiment of the present invention. Detailed Implementation

[0036] Hereinafter, exemplary embodiments according to the present invention will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are merely some embodiments of the present invention, and not all embodiments of the present invention. It should be understood that the present invention is not limited to the exemplary embodiments described herein.

[0037] Figure 1 This is a flowchart illustrating a model-driven wire harness development and optimization method according to an embodiment of the present invention. Figure 1 As shown, the model-driven approach to harness development and optimization includes:

[0038] S1: Collect the actual operating parameters of the wiring harness, including the temperature distribution parameters, vibration frequency parameters, and ambient humidity parameters of the wiring harness, and construct a digital twin model of the wiring harness using the temperature distribution parameters, the vibration frequency parameters, and the ambient humidity parameters;

[0039] Temperature sensors are installed at key nodes of the wiring harness, recording temperature distribution parameters at a preset sampling period T, where T is 1 minute. Accelerometers are installed every 20 centimeters along the wiring harness, collecting vibration frequency parameters ranging from 0-1000 Hz. Humidity sensors are installed at the wire inlet, outlet, and intermediate connection points of the wiring harness, collecting ambient humidity parameters. The temperature distribution parameters, vibration frequency parameters, and ambient humidity parameters are input into a pre-trained deep neural network, which includes three parallel LSTM layers to process these parameters respectively. After fusion processing by the deep neural network, a digital feature vector of the wiring harness is output. A digital twin model of the wiring harness is constructed based on this digital feature vector, comprising a physical property layer, a geometric property layer, and a material property layer.

[0040] The specific steps for constructing the digital twin model of the wire harness based on the digital feature vector are as follows:

[0041] The digital feature vectors are orthogonally decomposed to obtain physical feature sub-vectors characterizing the physical properties of the harness, geometric feature sub-vectors characterizing the shape of the harness, and material feature sub-vectors characterizing the material of the harness. A temperature field distribution is established using the temperature components in the physical feature sub-vectors. When the temperature field distribution model shows a local temperature gradient greater than a preset temperature threshold, the area is marked as a key temperature monitoring area. The vibration transfer function of each segment of the harness is calculated based on the vibration components in the physical feature sub-vectors. If the amplitude of the vibration transfer function of a certain segment of the harness exceeds a safety limit, the stiffness parameter of that segment of the harness is stored in the physical properties of the harness digital twin model. The system employs a database to construct a three-dimensional spatial curve equation for the wire harness using the geometric feature vectors. When the curvature of the three-dimensional spatial curve at any point exceeds the safe curvature, the stress concentration factor at that point is calculated and stored in the geometric property database of the wire harness digital twin model. A material constitutive equation is established based on the material feature vectors, and the elastic modulus and yield strength of each part of the wire harness are updated in real time. The updated material parameters are then written into the material property database of the wire harness digital twin model. Finally, a data mapping algorithm is used to integrate the physical property database, the geometric property database, and the material property database to form a complete wire harness digital twin model.

[0042] More specifically, calculating the vibration transfer function of each segment of the harness based on the vibration components in the physical feature sub-vectors includes:

[0043] The vibration components in the physical feature sub-vectors are transformed to the frequency domain using Fourier transform to obtain the frequency response function of each measurement point of the wire harness. When the amplitude of the frequency response function reaches a peak, the frequency is recorded as the natural frequency of the wire harness, and a vibration natural frequency dataset is established. The wire harness is discretized at equal intervals, dividing it into K micro-segments. A recursive algorithm is used to calculate the vibration transfer function between adjacent micro-segments. When the ratio of the vibration transfer function of the i-th segment to that of the (i+1)-th segment exceeds a preset threshold α, the location is marked as a vibration weak point. The preset threshold α is 1.5. If the number of vibration weak points detected exceeds 3, the vibration transmission characteristics of the harness are judged as abnormal, and a vibration suppression strategy is triggered. Based on the vibration transmission function of each micro-segment, the vibration transmission equation of the entire harness is fitted using the least squares method. When the goodness of fit of the vibration transmission equation is greater than 0.95, the equation is used as the global vibration transmission model of the harness. The vibration response characteristics of the harness under different working conditions are predicted by the global vibration transmission model, and the prediction results are fed back to the physical characteristic database of the harness digital twin model.

[0044] More specifically, the logic for calculating the vibration transfer function between adjacent infinitesimal segments using a recursive algorithm is as follows:

[0045] Starting from the first micro-segment at the beginning of the harness, it is designated as the reference segment, with the initial value of the vibration transfer function of the reference segment set to 1. Based on the data collected by the accelerometer of the reference segment, the vibration displacement response x1 and vibration acceleration response a1 of the micro-segment are obtained. For the second micro-segment adjacent to the reference segment, its vibration displacement response x2 and vibration acceleration response a2 are collected. The phase difference θ between the second micro-segment and the reference segment is calculated. When the phase difference θ is greater than 90 degrees, it is determined that there is a vibration coupling phenomenon between the two micro-segments. Based on the vibration displacement responses x1, x2 and a2, the vibration displacement response x2 and vibration acceleration response a2 of the reference segment are collected. The vibration acceleration responses a1 and a2 are used to establish a state-space equation system, in which the mass, stiffness, and damping of the infinitesimal segments are treated as state variables. The fourth-order Runge-Kutta method is used to solve the state-space equation system to obtain the vibration transfer function Hi,i+1 between adjacent infinitesimal segments. After the vibration transfer function between the i-th and i+1-th infinitesimal segments is calculated, the i+1-th segment is set as the new reference segment, and the above calculation process is repeated until all infinitesimal segments are traversed. If the vibration transfer function calculation result of a pair of adjacent infinitesimal segments shows a singular value, the Kalman filter algorithm is used to smooth the vibration response data at that point and then recalculate.

[0046] On the other hand, the logic for constructing the three-dimensional space curve equation of the wire bundle using the aforementioned geometric feature sub-vectors is as follows:

[0047] First, the geometric feature sub-vectors are dimensionality reduced to extract the key control point coordinate sequence of the harness in three-dimensional space. When the spacing between the control points is less than a preset distance d, control points are filtered to reduce computational complexity. Based on the control point coordinate sequence, a cubic spline interpolation method is used to construct the parametric equation of the harness centerline, where the parametric equation contains three components: x(t), y(t), and z(t). When a large curvature region is detected in the harness, the density of control points in that region is increased to improve the fitting accuracy of the curve in the large curvature region. The constructed three-dimensional space curve equation is stored in the geometric characteristic database of the harness digital twin model.

[0048] The specific process for selecting control points is as follows:

[0049] Sort all control points according to the cumulative arc length parameter along the harness direction to obtain the initial control point sequence P; calculate the Euclidean distance between adjacent control points. When the distance between any two adjacent control points is less than a preset threshold d, calculate the curvature values ​​k1 and k2 at these two control points; if the curvature difference |k1-k2| between the two control points is less than the curvature threshold ε, remove the control point with the smaller curvature from the sequence P and store its spatial location information in the candidate point set Q; when the control point density of a certain section is detected to be less than n points per meter, select the nearest control point in that section from the candidate point set Q to supplement it; for the turning sections of the harness, set a minimum control point spacing threshold dmin. When the distance between adjacent control points is less than dmin, the tangent continuity of the segment is calculated. If the tangent direction changes smoothly, only the endpoints and midpoints are retained as control points. While screening control points, it is ensured that key feature points of the harness must be retained, including the harness's fixing points, branch points, and contact points with other components. After the initial screening is completed, the maximum deviation between the curve constructed based on the screened control points and the original curve is calculated. If the maximum deviation exceeds the allowable error δ, control points are gradually added from the candidate point set Q until the accuracy requirements are met. If the number of screened control points is still too large, a curve segmentation fitting method is used to further reduce the number of control points while ensuring geometric accuracy.

[0050] Finally, the physical property database, the geometric property database, and the material property database are linked and integrated using a data mapping algorithm, as follows:

[0051] First, a master index table is established based on the geometric property database, marking each control point of the harness as a unique node ID. When the distance between the control points is greater than a preset value, a transition node is generated through an interpolation algorithm to ensure the continuity of the data mapping. For the temperature field data in the physical property database, a temperature-node mapping table is established, associating the temperature value of each node with the node ID. If the temperature data of a node is missing, a weighted average calculation is performed using the temperature values ​​of adjacent nodes. For the vibration transfer function data, a vibration-node mapping table is established, recording the vibration amplitude and phase information of each node. When the vibration data of a node is abnormal, a data verification process is triggered, and the abnormal value is corrected using historical data. For the material property database, a material-node mapping table is established, storing the material property parameters of each node. If there is a material transition zone in the harness, a material mixing rule is established at the corresponding node, and the equivalent material parameters are calculated. A relational data model is adopted, establishing the association between the three mapping tables through the node ID. When any database is updated, the associated data is automatically updated through a trigger mechanism. Finally, a unified state vector is constructed, which contains the node ID, spatial coordinates, physical parameters, and material parameters, forming a complete digital twin model data structure.

[0052] S2: Divide the digital twin model of the wire harness into N node units, perform stress analysis on the N node units, and obtain stress distribution data for each node unit, where N is an integer greater than 100;

[0053] First, a local coordinate system for each node element is established, and the position and orientation of each node element are determined based on the spatial curve equation of the wire bundle. Based on the position information of each node element, its corresponding boundary conditions and constraints are extracted. After setting the boundary conditions, a stress-strain constitutive equation is established for each node element. If the node element is a single material, Hooke's law is used to establish the constitutive equation; if the node element is a composite material, a layered constitutive equation system is established. Next, the strain field is calculated. First, the deformation of the node element is calculated based on the displacement boundary conditions; then, the strain components in each direction are calculated based on the deformation. Once the strain field is determined, the stress components are calculated by substituting them into the constitutive equation. Finally, based on each stress component, the equivalent stress is calculated using the von Meyses criterion.

[0054] Finally, stress assessment is performed, and the calculated equivalent stress is compared with the material strength. If the equivalent stress exceeds the material yield strength, it is marked as a potential failure area. When a stress concentration area is found, the mesh of the node element at that location is refined and recalculated. After completing the stress analysis of all node elements, the stress distribution data is stored in the database, and the stress concentration area is marked as a key monitoring target.

[0055] When the nodal elements are composite materials, the specific process for establishing the layered constitutive equations is as follows:

[0056] First, determine the number of layers in the composite material and the material properties of each layer, including fiber-direction elastic modulus E1, transverse elastic modulus E2, in-plane shear modulus G12, and principal Poisson's ratio ν12. When the fiber layup direction is known, record the layup angle θ of each layer. If a sandwich structure exists, treat the sandwich layer as an independent layer. Next, establish a local coordinate system for each layer, defining axis 1 as the fiber direction, axis 2 as the transverse direction perpendicular to the fiber, and axis 3 as the normal to the laminate. When global analysis is required, establish a global coordinate system xyz. If the local coordinate system does not coincide with the global coordinate system, establish a coordinate transformation matrix. Then, in the local coordinate system of each layer, establish a simplified plane stress state constitutive equation. Based on the coordinate transformation matrix, transform the constitutive equations of each layer to the global coordinate system. Finally, use the principle of equivalent stiffness superposition to combine the constitutive equations of each layer into a global stiffness matrix. After superposition, obtain the layered constitutive relation equations describing the mechanical behavior of the entire composite material nodal element.

[0057] More specifically, the specific constitutive equations are shown below:

[0058] ;

[0059] in, The normal stress is in the x-direction. The normal stress is in the y-direction. For shear stress in the xy plane, These are the stiffness matrix elements after transformation to the global coordinate system. For strain in the x-direction, Strain in the y-direction For shear strain, The coefficient of thermal expansion is... The change in temperature Let be the ply angle influence function of the k-th layer. Let be the ply angle of the k-th layer.

[0060] The ply angle influence function of the k-th layer can be expressed by the following formula:

[0061] ;

[0062] in, Let be the ply angle of the k-th layer (in radians), with a value range of [0, π]. This is the elastic modulus in the fiber direction. It is the transverse elastic modulus.

[0063] It is worth noting that although the above description outlines the general process for establishing a set of layered constitutive equations when the nodal elements are composite materials, the specific implementation details may vary depending on the material properties, layup method, and usage environment of the automotive wiring harness. For example, glass fiber reinforced materials commonly used in automotive wiring harnesses typically have a high fiber-direction elastic modulus (E1) and a low transverse elastic modulus (E2). This means that in the stiffness and durability analysis of the wiring harness in the bending region, special attention needs to be paid to the influence of the layup angle (θ) on the overall stiffness. Furthermore, for complex wiring harness structures (such as multi-layer shielded wiring harnesses), the core layer may contain a metal braided layer, and the influence of its in-plane shear modulus (G12) and Poisson's ratio (ν12) on electromagnetic interference resistance needs to be modeled separately, which differs significantly from traditional composite materials.

[0064] Furthermore, when establishing local and global coordinate systems, if there are large bending angles in the wiring harness (such as wiring harness paths bypassing the motor compartment), the transformation matrix between the local and global coordinate systems may involve complex three-dimensional rotation calculations. Especially when the ply angle is atypical (such as 45°), an accurate mathematical model must be introduced to describe the anisotropy of material properties. For example, in the design of high-voltage wiring harnesses for some vehicles, local material properties need to be dynamically adjusted in conjunction with the temperature field and vibration environment to ensure that the deformation of the core layer and fiber layer does not exceed the material's yield strength.

[0065] S3: Based on the stress distribution data, an improved ant colony algorithm is used to optimize the wiring path of the wire harness, wherein the improved ant colony algorithm introduces an entropy weight method to dynamically adjust the stress weight of different node units.

[0066] First, the path search space is initialized by constructing a three-dimensional mesh search space based on the start and end points of the wire harness. When obstacles exist, the obstacle area is marked as a no-passage area. If the wire harness needs to pass through fixed points, these points are set as mandatory nodes. An ant colony is established based on the initial conditions, and the colony size and maximum number of iterations are set.

[0067] Next, the entropy weight method is used to calculate the stress weight of the node element. By calculating the information entropy of the stress data, the weight coefficient of each node stress is dynamically determined. Then, the path iterative search begins, and each ant moves in the search space based on the transition probability. When an ant passes through a high-stress area, the penalty factor of the path is increased according to the stress weight. If the path passes through a low-stress area, the penalty factor is decreased. Based on the comprehensive evaluation of the total path length and stress weight, the fitness value of the path is calculated.

[0068] The pheromone concentration is then updated, with increased concentrations for paths with better fitness values; pheromone increments are appropriately reduced when a path passes through a stress concentration region; if a better path is found, the pheromone intensity of that path is increased; and a pheromone decay factor is used to prevent the algorithm from getting trapped in local optima.

[0069] Next, path smoothing is performed, and curve fitting is applied to the obtained optimal path. When the path has sharp turns, a circular arc transition is used for smoothing. If the smoothed path interferes with obstacles, the radius of the transition arc is adjusted locally. Ensure that the smoothed path meets the minimum bending radius requirement of the harness.

[0070] Finally, the solution is verified by re-analyzing the stress of the optimized routing path. When a region with excessive stress is found, the stress weight of that region is increased and the process is iterated again. If all constraints are met, the final routing solution is output and the optimized routing path is provided to the downstream design stage.

[0071] The entropy weight method for calculating the stress weights of node elements involves the following steps: First, stress data from all nodes is collected and standardized, mapping the data to a range of zero to one. Then, the stress percentage of each node is calculated, and information entropy is calculated based on these percentages. The more uniform the stress distribution at a node, the greater its information entropy. Next, the information entropy is subtracted from 1 to obtain a difference coefficient, which is then normalized to obtain the initial weights. Finally, the weights are dynamically adjusted based on the relationship between the node stress value and a preset threshold. When the stress at a node approaches or exceeds the strength limit, its weight coefficient is increased to strengthen avoidance of that area, while for safer areas with lower stress, its weight coefficient is decreased. These weight coefficients are continuously updated during iterative optimization, thereby achieving adaptive path planning based on stress distribution. More specifically, the calculation of the stress weights of node elements is shown in the following formula:

[0072] ;

[0073] in, Let be the final weight value of the i-th node. Let n be the information entropy of the i-th node, and n be the total number of nodes. Let be the equivalent stress value of the i-th node. The preset stress threshold, The maximum stress value among all nodes. The average stress value. The weighting adjustment coefficient has a range of values ​​[0.1, 0.5]. The stress sensitivity coefficient ranges from [0.5, 1.5].

[0074] On the other hand, a pheromone decay factor is used to avoid the algorithm getting trapped in local optima, specifically as follows:

[0075] When the algorithm performs pheromone decay operations, it first sets a global decay factor ρ to control the degree of pheromone retention after each iteration. If the pheromone concentration of a path exceeds the threshold θ, a larger decay factor ρ1 is used to force decay of that path. When the optimal solution remains unchanged for N consecutive iterations, it is determined that the algorithm is trapped in a local optimum, and the system increases the global decay factor to ρ2 to accelerate pheromone evaporation. If the selection probability of a certain path is too high and exceeds the threshold p, an additional decay mechanism is activated for that path, using a larger decay factor ρ3 to quickly eliminate its dominant position. When the pheromone difference among multiple paths is detected to be too large, the decay enhancement operation is uniformly performed on the K paths with the highest pheromone concentration. If the path selection is found to be too concentrated during the iteration process, the decay factor is temporarily increased to ρ4 to increase the randomness of path search. When the search falls into a repetitive loop, an emergency perturbation mechanism is triggered, increasing the global decay factor to the maximum value ρmax to forcibly break the established fixed pattern. If the optimization target oscillates, the decay factor is dynamically adjusted to fluctuate within the range of ρmin to ρmax to balance local development and global exploration. When the algorithm rediscovers a better solution, the decay factor is gradually restored to the initial level, and effective pheromone is re-accumulated.

[0076] When determining the pheromone decay factor, the base decay factor ρ is initially set to 0.5 as a standard reference value. If the path pheromone exceeds the threshold, the forced decay factor ρ1 is set to 0.7 to accelerate the evaporation of abnormal pheromones. When a local optimum is detected, the global decay factor ρ2 is increased to 0.8 to disrupt the inherent path. If the probability of selecting a certain path is too high, the dedicated decay factor ρ3 for that path is increased to 0.85 to quickly reduce its influence. When path selection tends to be singular, the temporary decay factor ρ4 is set to 0.9 to forcibly introduce randomness. To ensure algorithm convergence, the minimum decay factor ρmin is limited to above 0.3. When a severe local optimum occurs, the maximum decay factor ρmax can reach 0.95 to achieve drastic perturbation. If the optimization effect is not good, the decay factor is increased by 0.05 every certain number of iterations until it reaches ρmax. When a better solution is found, the decay factor is decreased by 0.05 each time, but not lower than ρmin. If the problem size increases, the baseline values ​​of all decay factors are adjusted accordingly to increase the search space.

[0077] For example, when planning paths in a 15x15 grid map, the system initially sets the global pheromone decay factor ρ = 0.3. If the initial pheromone concentration of each feasible path is 10, after the first iteration, the pheromone concentration of all paths will decay to 7. When path A is detected to have a selection probability greater than 80% for 5 consecutive iterations, it is determined to be trapped in a local optimum, and the system increases the decay factor of path A to 0.6, causing its pheromone to decay rapidly. If the pheromone of path A was originally 50, it will drop to 20 after forced decay. When the pheromone concentration of the suboptimal path B is detected... When the pheromone concentration is 15, which is too large compared to path A, the global decay factor is increased to 0.5 to accelerate the elimination of pheromone differences between paths. If, after the 20th iteration, the pheromone levels of paths A and B drop to 25 and 12 respectively, and a new feasible path C is found, and path C shows a better path cost in subsequent iterations, its decay factor is reduced to 0.2 to slow down pheromone evaporation. If the pheromone level of path C accumulates to 40 after 10 iterations, while paths A and B decay to 10 and 5 respectively, then the algorithm has successfully escaped the local optimum and found a better solution.

[0078] Preferably, when using a pheromone decay factor for path selection, the performance is significantly improved compared to traditional methods. Without a pheromone decay mechanism, the algorithm's path selection probability often concentrates on the initially discovered local optimum after 15 iterations, leading to excessively fast convergence. When the decay factor is introduced, even if the pheromone concentration of the optimal path reaches its peak, it will gradually decay over time, reserving search space for other possible better paths. If the path length of path A discovered in the initial stage is 100, after 30 iterations, due to the effect of the decay factor, the algorithm successfully discovers path B with a path length of 85. When the iteration reaches 50 rounds, the system can continue to search for better paths based on path B, eventually discovering path C with a path length of 78. If the average path length is used as the evaluation index, the optimization result after using the decay factor is about 18% lower than that of the traditional method. When evaluating the algorithm's convergence speed, using the decay factor improves the convergence speed by about 25% while maintaining global search capability.

[0079] S4: Based on the optimization results of the wiring path, adjust the position of the fixing bracket of the wire harness, and add a protective structure at the node unit where the stress distribution data exceeds the preset stress threshold.

[0080] The process of adjusting the position of the wire harness fixing brackets and adding protective structures based on the optimization results of the wiring path is as follows: First, fixing brackets are set at path turning points and long straight sections according to the minimum support density requirements, and the bracket arrangement is appropriately densified in areas with interference risks; then, the stress state of the node units is evaluated, and areas where the stress exceeds 80% of the preset threshold are marked as high-risk areas; for these high-risk areas, reinforcing ribs are used near the fixing brackets to strengthen them, protective covers are set up in areas of external impact, and vibration damping devices are added in areas of concentrated vibration; finally, the stress concentration is reduced by adjusting the bracket positions, and the effectiveness of the scheme is verified by finite element analysis to ensure that the stress of each node is lower than the preset threshold.

[0081] In summary, the model-driven wire harness development and optimization method based on embodiments of the present invention has been clarified. It constructs a digital twin model by collecting actual operating parameters and subdivides the model into a large number of node units for precise stress analysis, enabling accurate identification of potential stress concentration areas. This solves the problem that existing wiring solutions, which lack accurate analysis of stress distribution under actual operating conditions, easily lead to stress concentration and wire harness damage.

[0082] Here, those skilled in the art will understand that the specific operations of each step in the above-described model-driven harness development and optimization method have been referenced above. Figure 1 The model-driven approach to harness development and optimization has been described in detail in the previous section, and therefore, its repeated description will be omitted here.

[0083] In summary, the model-driven wire harness development and optimization method based on embodiments of the present invention has been clarified. It constructs a digital twin model by collecting actual operating parameters and subdivides the model into a large number of node units for precise stress analysis, enabling accurate identification of potential stress concentration areas. This solves the problem that existing wiring solutions, which lack accurate analysis of stress distribution under actual operating conditions, easily lead to stress concentration and wire harness damage.

Claims

1. A model-driven method for developing and optimizing wire harnesses, characterized in that, include: Collect the actual operating parameters of the wiring harness, and construct a digital twin model of the wiring harness based on the actual operating parameters; The digital twin model of the wire harness is divided into N node units, and stress analysis is performed on the N node units to obtain stress distribution data for each node unit. Based on the stress distribution data, the wiring path of the wire harness is optimized; Based on the optimization results of the wiring path, the position of the fixing bracket of the wire harness is adjusted, and a protective structure is added at the node unit where the stress distribution data exceeds the preset stress threshold. The wire harness digital twin model integrates the physical property database, geometric property database, and material property database through a data mapping algorithm; The vibration transfer function of each segment of the wire harness is calculated by the physical feature sub-vector in the digital feature vector. If the amplitude of the vibration transfer function of a certain segment of the wire harness exceeds the safety limit, the stiffness parameter of that segment of the wire harness is stored in the physical property database. The digital feature vector is obtained by fusing the actual operating parameters through a neural network, and then by orthogonal decomposition to obtain physical feature sub-vectors characterizing the physical properties of the wire harness, geometric feature sub-vectors characterizing the shape of the wire harness, and material feature sub-vectors characterizing the material of the wire harness.

2. The model-driven wire harness development and optimization method according to claim 1, characterized in that, The geometric feature sub-vectors are used to construct the three-dimensional spatial curve equation of the wire harness. When the curvature of the three-dimensional spatial curve at any point exceeds the safe curvature, the stress concentration factor at that point is calculated and stored in the geometric characteristic database of the wire harness digital twin model.

3. The model-driven wire harness development and optimization method according to claim 2, characterized in that, Constructing the equation of the three-dimensional space curve includes: The geometric feature sub-vectors are dimensionality reduced to extract the key control point coordinate sequence of the wire bundle in three-dimensional space; When the spacing between control points is less than a preset distance d, control point filtering is performed; Based on the control point coordinate sequence, the three-dimensional spatial curve equation is constructed.

4. The model-driven wire harness development and optimization method according to claim 3, characterized in that, The process of obtaining stress distribution data for each node element includes: Based on the location information of each node unit, the corresponding boundary conditions and constraints are extracted. After setting the boundary conditions, different constitutive equations are constructed based on the material structure of each node element. The deformation of the nodal elements is calculated based on the displacement boundary conditions, and the strain components in each direction are calculated based on the deformation. Substituting the strain components into the constitutive equation yields the stress components of each nodal element.

5. The model-driven wire harness development and optimization method according to claim 4, characterized in that, When the node element is a composite material, a set of layered constitutive equations is established; The set of constitutive equations for the layered structure determines the number of layers in the composite material and the material properties of each layer, records the layup angles, and treats the core layer as an independent layer, establishing a local coordinate system and a global coordinate system. If the two coordinate systems do not coincide, the constitutive equations of each layer are transformed from the local coordinate system to the global coordinate system using the coordinate transformation matrix. Finally, based on the principle of equivalent stiffness superposition, the constitutive equations of each layer are combined into a global stiffness matrix, thereby obtaining a set of layered constitutive relation equations describing the mechanical behavior of composite material nodal elements.

6. The model-driven wire harness development and optimization method according to claim 5, characterized in that, Ant colony optimization is used to find the optimal wiring path for the wire harness; The ant colony algorithm uses the entropy weight method to dynamically adjust the stress weight of different node units, and adopts the pheromone decay factor to avoid the ant colony algorithm getting trapped in local optima.

7. The model-driven wire harness development and optimization method according to claim 6, characterized in that, Its features are, The entropy weighting method dynamically adjusts the stress weights of different node elements, including: Calculate the stress percentage for each node element and calculate the information entropy based on these percentages; The difference coefficient is obtained by subtracting the information entropy from one, and the initial weight is obtained by normalizing the difference coefficient. The weights are dynamically adjusted based on the relationship between the stress value of the node element and a preset threshold.

8. The model-driven wire harness development and optimization method according to claim 7, characterized in that, When the optimal solution remains unchanged for multiple consecutive iterations, it is determined that it has entered a local optimum. By setting different pheromone decay factors, the pheromone differences between paths can be eliminated more quickly, thereby escaping the local optimum and finding a better solution.

Images