Method for optimizing design of diesel engine piston skirt based on model simulation

Abstract

The present application relates to the internal combustion engine parts simulation technical field, specifically for the diesel engine piston skirt optimization design method based on model simulation, contain model construction, interference mapping, index solution and iterative optimization steps, the system through the mapping time-varying geometry interference field, simulate the nonlinear deformation of cylinder liner, its core is according to dynamic contact topology, the solution line heat sensitivity entropy and oil film robustness index, to quantify the degree of system away from failure boundary, according to the comparison of index and threshold, self-adapting adjustment weight and correction line, the present application realizes the change from static ideal surface to dynamic special-shaped channel design, effectively improves the lubrication reliability under real working condition.

Description

Technical Field

[0001] This invention relates to the field of internal combustion engine component design and simulation technology, specifically a diesel engine piston skirt optimization design method based on model simulation. Background Technology

[0002] With the continuous advancement of high-performance diesel engine technology, its operating conditions are becoming increasingly demanding, which places higher demands on the piston skirt's ability to withstand extremely low frictional power consumption and cylinder scoring under harsh conditions.

[0003] Currently, conventional piston skirt design methods are often based on the assumption of an ideal cylindrical surface or static boundary conditions for tribological optimization, focusing on pursuing low friction indicators under steady state. However, in actual operation, the cylinder liner is subjected to the combined effects of bolt preload, burst pressure, and non-uniform thermal load, resulting in complex nonlinear thermomechanical coupling deformation. Because traditional designs lack an effective quantitative assessment mechanism for the risk of such dynamic geometric topology mismatch, the design scheme is difficult to adapt to the real-time breathing and twisting of the cylinder liner, and cannot maintain oil film stability under harsh operating conditions while ensuring extremely low friction power consumption, which can easily lead to lubrication failure or even cylinder scoring accidents.

[0004] Therefore, how to achieve synergistic optimization of low frictional power consumption and high cylinder stress resistance of the piston skirt under complex working conditions where the cylinder liner undergoes nonlinear deformation has become an urgent problem to be solved in this field. Summary of the Invention

[0005] To address the aforementioned technical problems, this invention provides a model simulation-based optimization design method for diesel engine piston skirts. Specifically, the technical solution of this invention includes:

[0006] Step 1: Obtain the initial geometric topology data of the diesel engine piston skirt and cylinder liner, and discretize them to construct a three-dimensional contact mesh model; based on thermodynamic boundary conditions and multibody dynamics equations, establish a fluid lubrication coupled simulation environment, and define the design variable space of the skirt profile;

[0007] Step 2: Construct a time-varying geometric disturbance field and map the time-varying geometric disturbance field onto the inner surface of the cylinder liner of the three-dimensional contact mesh model to simulate the nonlinear deformation of the cylinder liner under thermomechanical coupling and generate a dynamic contact topology containing disturbance features.

[0008] Step 3: In the fluid lubrication coupling simulation environment, iterative calculations are performed on the piston skirt loaded with the dynamic contact topology. The transient oil film thickness field and local specific pressure distribution in the contact area are extracted. The profile thermal sensitivity entropy, which characterizes the degree of deformation dispersion of the profile under temperature disturbance, is calculated. Combined with the lubrication state boundary, an oil film robustness index reflecting the degree of the system's distance from the failure boundary is constructed.

[0009] Step 4: Preset failure risk threshold, compare the oil film robustness index with the failure risk threshold, and adaptively adjust the weight configuration of the objective function according to the comparison result to generate corrected profile data for the current disturbance state.

[0010] Step 5: Based on the corrected profile data, update the design variable space of the skirt profile, and repeat steps 2 to 4 until the oil film robustness index meets the convergence condition, and output the final piston skirt profile optimization design scheme.

[0011] Preferably, step one includes:

[0012] S11. Obtain the original profile coordinate point set of the piston skirt and the cylindrical surface equation of the cylinder liner under ideal conditions through 3D scanning or parametric modeling.

[0013] S12. Perform spline interpolation fitting on the original profile coordinate point set to generate a skirt surface model, and divide the skirt surface model and cylinder liner into hexahedral meshes to form several contact elements, each contact element being marked as the kth node.

[0014] S13. Collect the engine's burst pressure curve, piston heat load distribution data, and lubricating oil rheological parameters under all operating conditions, and input them as boundary conditions into the fluid lubrication coupling simulation environment to initialize the system state vector.

[0015] Preferably, step two includes:

[0016] S21. Collect the static deformation of the cylinder liner under preload and the thermal expansion deformation under different thermal loads.

[0017] S22. The static deformation amount and the thermal expansion deformation amount are superimposed in the time domain to construct a time-varying geometric disturbance field that varies with the crank angle.

[0018] S23. Extract the normal displacement component of the k-th node on the inner surface of the cylinder liner, use the time-varying geometric disturbance field to correct the normal displacement component, update the node coordinates of the three-dimensional contact mesh model, and form a dynamic contact topology of a non-standard cylindrical surface.

[0019] Preferably, step three includes:

[0020] S31. Based on the Reynolds equation and contact mechanics model, calculate the instantaneous oil film thickness at the k-th node at the current time step. Contact pressure with roughness ;

[0021] S32. Statistically analyze the oil film rupture frequency of all contact units throughout the entire cycle. Use the oil film rupture frequency and the profile thermal sensitivity entropy as evaluation parameters, and calculate the oil film robustness index using a preset negative correlation mapping function. The oil film robustness index is negatively correlated with both the oil film rupture frequency and the thermal sensitivity entropy of the profile.

[0022] The thermal sensitivity entropy of the skirt profile characterizes the degree of dispersion of radial deformation distortion of the skirt profile when a small disturbance occurs in the temperature field.

[0023] Preferably, in step four, the oil film robustness index is compared with the failure risk threshold, and the weight configuration of the objective function is adaptively adjusted based on the comparison result, specifically including:

[0024] S41. Define the optimization objective function, which includes a first sub-objective and a second sub-objective, where the first sub-objective is to minimize frictional power consumption and the second sub-objective is to maximize topology tolerance.

[0025] S42, when the oil film robustness index When the failure risk threshold is lower than the threshold, the system is determined to be in a high-risk mixed lubrication state, and a first weighting strategy is generated: reduce the weight of the first sub-objective, increase the weight of the second sub-objective, and drive the search algorithm to find a conservative profile with higher geometric tolerance.

[0026] S43, when the oil film robustness index When the failure risk threshold is higher than or equal to the threshold, the system is determined to be in a safe lubrication state, and a second weighting strategy is generated: the weight of the first sub-objective is increased, the weight of the second sub-objective is decreased, and the search algorithm is driven to find an aggressive profile with lower frictional resistance.

[0027] Preferably, step four also includes:

[0028] S44. A deep reinforcement learning algorithm is used as the search strategy, and the oil film robustness index is used as the environmental reward signal.

[0029] S45. In response to the first weighting strategy, the deep reinforcement learning algorithm punishes geometric features that lead to low robustness by generating negative reward signals, thereby suppressing the generation of profiles with excessive local curvature.

[0030] S46. In response to the second weighting strategy, the deep reinforcement learning algorithm rewards geometric features that result in low frictional power consumption by generating positive reward signals, thereby exploring profile generation that reduces contact area.

[0031] S47. Output the weighted and optimized design variable increments and add them to the current skirt profile data to generate the corrected profile data.

[0032] Preferably, in step three, the specific steps for calculating the thermal sensitivity entropy of the profile, which characterizes the sensitivity of the profile to changes in the temperature field, are as follows:

[0033] S331. Introduce a set of random temperature disturbance vectors into the fluid lubrication coupling simulation environment;

[0034] S332. Calculate the variance distribution of the radial deformation of the skirt profile under the action of the random temperature disturbance vector.

[0035] S333. Perform an integral operation on the variance distribution over the skirt surface domain to obtain a scalar value characterizing the deterministic nature of the profile deformation, which is used as the thermal sensitivity entropy of the profile. The higher the entropy value, the more likely the profile is to generate unpredictable contact drift under thermal shock.

[0036] Preferably, step five includes:

[0037] S51. Monitor the convergence trend of the oil film robustness index in real time;

[0038] S52. If, in a series of preset number of iterations, the oil film robustness index remains above the failure risk threshold and the rate of change of frictional power consumption is lower than the preset convergence limit, then the optimization is deemed complete.

[0039] S53. Extract the modified profile data generated in the last iteration, convert it into CNC machining code or a three-dimensional solid model file, and output it as the final piston skirt profile optimization design scheme.

[0040] Compared with the prior art, the present invention has the following beneficial effects:

[0041] 1. This method breaks through the limitations of traditional design based solely on the assumption of an ideal cylindrical surface, significantly improving the adaptability of the design scheme under real harsh working conditions. By constructing a time-varying geometric disturbance field and mapping it to a three-dimensional contact mesh model, this invention can accurately simulate the nonlinear breathing and torsional deformation of the cylinder liner under the combined action of bolt preload, burst pressure, and non-uniform thermal load. This dynamic contact topology construction method transforms the design environment of the piston skirt profile from a static ideal surface to a dynamic irregular channel, thereby ensuring that the optimized profile can still maintain a good matching relationship when the cylinder liner undergoes severe thermomechanical coupling deformation, effectively reducing the risk of lubrication failure caused by geometric topology mismatch.

[0042] 2. This method solves the problem of traditional design lacking quantitative assessment of geometric topological mismatch risk by introducing profile thermal sensitivity entropy. Existing technologies often struggle to predict the impact of minute temperature disturbances on profile performance, while this invention uses the physical quantity of profile thermal sensitivity entropy to mathematically quantify the degree of deformation dispersion of the skirt profile under temperature field disturbances. By screening low-entropy insensitive profiles, this invention can design structures with higher geometric toughness, ensuring that the piston can maintain the deterministic contact state under local temperature variations caused by cooling system fluctuations or combustion instability, thus avoiding unpredictable contact drift.

[0043] 3. This method achieves dynamic synergistic optimization between extremely low frictional power consumption and high cylinder scoring robustness, resolving the core contradiction in high-performance diesel engine design. Based on the comparison mechanism between the oil film robustness index and the failure risk threshold, this invention establishes an adaptive weight adjustment strategy: when the system is in a high-risk mixed lubrication state, it automatically focuses on improving the topology fault tolerance to ensure survival; when the system is in a safe lubrication state, it automatically focuses on reducing frictional resistance to improve efficiency. This intelligent decision-making logic of prioritizing survival and then development ensures that the final generated profile can maintain low friction in steady state and maintain the oil film without rupture under harsh operating conditions, significantly improving the overall reliability of the engine.

[0044] 4. This method utilizes deep reinforcement learning algorithms to break through the boundaries of human experience-based design, enabling the discovery of superior non-intuitive profile features. Unlike traditional gradient search algorithms, this invention employs deep reinforcement learning as the search strategy and uses the oil film robustness index as the environmental reward signal. By positively rewarding low-friction features and negatively penalizing high-risk geometric features, such as excessive local curvature, the agent can explore complex profiles with both smooth transitions and micro-lubricating wedge effects within a vast design variable space. This method not only avoids getting trapped in local optima during the optimization process but also generates high-performance geometric shapes that are difficult for human designers to conceive intuitively, further unlocking the performance potential of the piston skirt. Attached Figure Description

[0045] The present invention will be further explained below with reference to the accompanying drawings and embodiments:

[0046] Figure 1 This is a flowchart of the method of the present invention. Detailed Implementation

[0047] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to specific embodiments.

[0048] Example 1:

[0049] Please see Figure 1A model simulation-based optimization design method for diesel engine piston skirts, comprising the following steps:

[0050] Step 1: Obtain the initial geometric topology data of the diesel engine piston skirt and cylinder liner, and discretize them to construct a three-dimensional contact mesh model; based on thermodynamic boundary conditions and multibody dynamics equations, establish a fluid lubrication coupled simulation environment, and define the design variable space of the skirt profile;

[0051] Step 2: Construct a time-varying geometric disturbance field and map it onto the inner surface of the cylinder liner of the three-dimensional contact mesh model to simulate the nonlinear deformation of the cylinder liner under thermomechanical coupling and generate a dynamic contact topology containing disturbance features.

[0052] Step 3: In the fluid lubrication coupled simulation environment, iterative calculations are performed on the piston skirt loaded with dynamic contact topology. The transient oil film thickness field and local specific pressure distribution in the contact area are extracted. The profile thermal sensitivity entropy, which characterizes the degree of deformation dispersion of the profile under temperature disturbance, is calculated. Combined with the lubrication state boundary, an oil film robustness index reflecting the degree of the system's distance from the failure boundary is constructed.

[0053] Step 4: Preset failure risk threshold, compare the oil film robustness index with the failure risk threshold, and adaptively adjust the weight configuration of the objective function based on the comparison results to generate corrected profile data for the current disturbance state.

[0054] Step 5: Based on the corrected profile data, update the design variable space of the skirt profile, and repeat steps 2 to 4 until the oil film robustness index meets the convergence condition, and output the final piston skirt profile optimization design scheme.

[0055] This embodiment provides a model simulation-based optimization design method for diesel engine piston skirts. This method aims to address the core technical contradiction in existing piston skirt design methods: the difficulty in simultaneously achieving extremely low frictional power consumption and cylinder scoring resistance under harsh operating conditions. Particularly when the cylinder liner undergoes nonlinear thermomechanical coupling deformation, traditional designs often suffer from lubrication failure due to a lack of quantitative assessment of geometric topological mismatch risks. The system constructs a high-fidelity digital twin environment to acquire initial geometric topological data of the diesel engine piston skirt and cylinder liner, and performs discretization processing to construct a three-dimensional contact mesh model. This step involves not only simple geometric modeling but also the digital reconstruction of physical boundaries. Based on thermodynamic boundary conditions and multibody dynamics equations, a fluid lubrication coupling simulation environment is established. Within this environment, a design variable space for the skirt profile is defined, containing a set of parameters describing the skirt's convexity, ellipticity, and non-standard irregular features. Specifically, the design variable space is defined as a vector... ,in, The maximum height of the central convex point. The axial position factor of the convex point. This is the ellipticity coefficient. This represents the radial adjustment amount for the B-spline control points;

[0056] A time-varying geometric disturbance field is introduced to simulate the harsh environment under real-world working conditions. This disturbance field is not merely a static deformation but includes the dynamic response of the cylinder liner under the combined effects of bolt preload, burst pressure, and non-uniform thermal load. The time-varying geometric disturbance field is mapped onto the inner surface of the cylinder liner in a three-dimensional contact mesh model to simulate the nonlinear deformation of the cylinder liner under thermomechanical coupling. This process generates a dynamic contact topology containing disturbance characteristics. Rather than operating within a standard cylindrical cylinder liner, the piston is more accurately described as breathing and twisting in an irregularly shaped channel over time. The motion provides a realistic physical scenario for subsequent robustness assessment; multiphysics coupling calculations and robustness quantification are performed, and iterative calculations are conducted on the piston skirt with dynamic contact topology in a fluid lubrication coupling simulation environment; the transient oil film thickness field and local specific pressure distribution of the contact area are extracted; based on this, the profile thermal sensitivity entropy is calculated; the profile thermal sensitivity entropy refers to a physical quantity that characterizes the degree of geometric deformation or distortion of the skirt profile under small temperature disturbances; if the entropy value is too high, it means that the profile is extremely sensitive to temperature and is prone to uncontrollable contact drift;

[0057] Combining the lubrication state boundary, i.e., the minimum oil film thickness limit, an oil film robustness index is constructed. This index reflects the safe distance between the current system state and the cylinder scoring failure boundary, i.e., the direct metal-to-metal welding boundary. Then, adaptive weight adjustment and profile correction are performed, and a failure risk threshold is preset. The oil film robustness index is compared with the failure risk threshold. This step is the core of the decision-making of this method: if the index is lower than the threshold, it indicates that the system is in a high-risk state. The algorithm automatically adjusts the weight configuration of the objective function, focusing on survival, i.e., sacrificing some friction power consumption indicators to prioritize improving the topology fault tolerance. If the index is higher than the threshold, it indicates that the system is in a safe zone. The algorithm focuses on efficiency and increases the weight of reducing friction power consumption. Based on the weight adjustment results, corrected profile data for the current disturbance state is generated. Entering the iterative closed loop and output, based on the corrected profile data, the design variable space of the skirt profile is updated, and steps two to four are repeated until the oil film robustness index meets the convergence condition, i.e., optimal performance is achieved under the premise of ensuring safety, and the final piston skirt profile optimization design scheme is output.

[0058] This embodiment, by introducing a time-varying geometric disturbance field and profile thermal sensitivity entropy, achieves for the first time the pre-simulation of various possibilities for imperfect cylinder liner operating conditions during the design phase. This method breaks the limitation of traditional methods that only optimize friction under ideal cylindrical surfaces, and can design a piston skirt profile with geometric toughness. This profile not only has low friction under steady state, but more importantly, it can still maintain the oil film without rupture when the cylinder liner undergoes severe thermal deformation, thereby significantly reducing the risk of cylinder scoring in high-performance diesel engines.

[0059] Example 2:

[0060] Step one includes:

[0061] S11. Obtain the original profile coordinate point set of the piston skirt and the cylindrical surface equation of the cylinder liner under ideal conditions through 3D scanning or parametric modeling.

[0062] S12. Perform spline interpolation fitting on the original profile coordinate point set to generate the skirt surface model, and perform hexahedral meshing on the skirt surface model and cylinder liner to form several contact elements, each contact element being marked as the kth node.

[0063] S13. Collect the engine's burst pressure curve, piston heat load distribution data, and lubricating oil rheological parameters under all operating conditions, and input them as boundary conditions into the fluid lubrication coupling simulation environment to initialize the system state vector.

[0064] This embodiment further specifies step one of Embodiment 1, focusing on how to construct a high-precision simulation initialization environment; perform geometric data acquisition and equation definition by acquiring the original profile coordinate point set of the existing piston skirt through a high-precision laser 3D scanner, or by directly exporting point cloud data from CAD parametric modeling software; simultaneously, define the cylindrical surface equation of the cylinder liner under ideal conditions; perform spline fitting and mesh generation. To ensure the continuity and convergence of the simulation calculation, the original profile coordinate point set is fitted with bicubic B-spline interpolation to generate a smooth skirt surface model; perform hexahedral mesh generation on the skirt surface model and the cylinder liner; compared to tetrahedral meshes, hexahedral meshes have higher accuracy and better Jacobian matrix quality in fluid lubrication calculations; form several contact elements, each contact element is marked as the k-th node, and the total number of nodes is N;

[0065] After completing the multiphysics boundary condition input, key data under all engine operating conditions were collected, including: the explosion pressure curve, which was obtained from the measured indicator diagram data of the combustion analyzer; piston heat load distribution data, which was obtained from the temperature field results of finite element thermal analysis; and lubricating oil rheological parameters, including viscosity-temperature relations, i.e., Vogel equation parameters and compressive viscosity coefficient. Specifically, the Vogel equation is used to calculate the lubricating oil viscosity at normal pressure, and its expression is:

[0066] ;

[0067] in, Absolute temperature The characteristic constant of the lubricating oil is obtained by performing nonlinear least-squares fitting on the measured viscosity data of the target lubricating oil at at least three different temperature points; the compressive viscosity coefficient. The formula used to characterize the enhancing effect of pressure on viscosity is as follows: ,in, The local oil film pressure is used as the boundary condition input into the fluid lubrication coupling simulation environment to initialize the system state vector, including initial displacement, velocity, and oil film pressure.

[0068] This embodiment constructs a digital twin base capable of accurately capturing micron-level oil film changes by using high-precision spline fitting and hexahedral mesh generation, combined with realistic combustion and thermal load boundary conditions. This provides the necessary computational accuracy guarantee for subsequent capture of minute geometric disturbances.

[0069] Example 3:

[0070] Step two includes:

[0071] S21. Collect the static deformation of the cylinder liner under preload and the thermal expansion deformation under different thermal loads.

[0072] S22. The static deformation and thermal expansion deformation are superimposed in the time domain to construct a time-varying geometric disturbance field that varies with the crank angle.

[0073] S23. Extract the normal displacement component of the k-th node on the inner surface of the cylinder liner, use the time-varying geometric disturbance field to correct the normal displacement component, update the node coordinates of the three-dimensional contact mesh model, and form a dynamic contact topology of the non-standard cylindrical surface.

[0074] This embodiment further specifies step two of Embodiment 1, detailing the construction process of the time-varying geometric disturbance field. The focus is on digitizing the physical deformation into a specific correction matrix; performing multi-source deformation acquisition to collect two basic deformation data of the cylinder liner: static deformation. The radial displacement field of the cylinder liner caused by the preload of the cylinder head bolts was calculated using finite element analysis. The axial height From a circumferential perspective, the data format typically presents as a fourth wave. Characteristics; thermal expansion deformation Under calibrated operating conditions, such as 100% load, the radial expansion deformation of the cylinder liner due to a non-uniform temperature field is described here. Represents engine load parameters; Note: In this embodiment, inward deformation is defined as positive, therefore, the amount of outward expansion deformation is taken as negative when substituted into the interference field calculation; A time-varying interference field is constructed to simulate the dynamic breathing effect of the cylinder liner with the fluctuation of the combustion pressure under real working conditions, and to construct the interference field with the crankshaft angle. Changing time-varying geometric disturbance field The specific superposition formula is as follows:

[0075] ;

[0076] in, This is a normalized instantaneous in-cylinder combustion pressure curve, with values ​​ranging from 0 to 1. The dynamic breathing coefficient, determined by the cylinder liner stiffness, typically ranges from 0.05 to 0.15. This formula quantifies the nonlinear time-varying characteristics under thermo-mechanical coupling. The parameters of this time-varying geometric disturbance field formula are fitted based on measured cylinder deformation data from multiple engine bench tests. The superposition mathematical form used is necessary for accurately decoupling static preload deformation and dynamic thermo-mechanical coupling deformation in simulation calculations, thus ensuring the authenticity of subsequent contact topology generation. The dynamic contact topology update is completed by extracting the original normal coordinates of the k-th node on the inner surface of the cylinder liner. The normal displacement component is corrected using a time-varying geometric disturbance field. The correction formula is as follows:

[0077] ;

[0078] Note: Indicates the first The axial height of each node, Indicates the first The circumferential angle of each node, where the negative sign indicates that inward deformation leads to a reduction in the cylinder radius; through this step, the node coordinates of the three-dimensional contact mesh model are updated, forming a non-standard cylindrical surface topology that dynamically creeps over time, providing accurate physical boundaries for subsequent lubrication calculations.

[0079] Example 4:

[0080] Step three includes:

[0081] S31. Based on the Reynolds equation and contact mechanics model, calculate the instantaneous oil film thickness at the k-th node at the current time step. Contact pressure with roughness ;

[0082] S32. Statistically analyze the oil film rupture frequency of all contact units throughout the entire cycle. Use the oil film rupture frequency and profile thermal sensitivity entropy as evaluation parameters, and calculate the oil film robustness index using a preset negative correlation mapping function. Among them, the oil film robustness index is negatively correlated with the oil film rupture frequency and the thermal sensitivity entropy of the profile;

[0083] Among them, the thermal sensitivity entropy of the skirt profile characterizes the degree of dispersion of the radial deformation distortion of the skirt profile when a small perturbation occurs in the temperature field.

[0084] This embodiment further specifies step three of Embodiment 1, focusing on the mathematical definition and calculation logic of the oil film robustness index, thus solving the problem of black-box evaluation indicators; it performs instantaneous lubrication parameter calculation based on the two-dimensional average Reynolds equation:

[0085] ;

[0086] in, These are the circumferential and axial coordinates along the unfolded surface of the piston skirt, respectively. The fluid dynamic pressure distribution to be solved; The nominal oil film thickness takes into account surface roughness; The dynamic viscosity of the lubricating oil is determined by the Vogel equation and the compressive viscosity coefficient in Example 2. The instantaneous relative sliding speed at the piston-cylinder liner contact interface takes into account the dynamic deformation speed of the cylinder liner; For pressure-flow factor, The above flow factor is a correction coefficient introduced based on the Patir-Cheng average flow model, used to characterize the anisotropic effect of surface roughness texture directionality on lubricating oil flow. Let time be the variable; using the Greenwood-Tripp micro-convexity contact model, calculate the instantaneous oil film thickness at the k-th node. Contact pressure with roughness Rough contact pressure The specific calculation formula is as follows:

[0087] ;

[0088] in, For the surface density of the microconvex body, Let be the average radius of curvature of the micro-convexity. The overall roughness of the contact surfaces. The combined elastic modulus of the two contact surfaces. The statistical function related to the height distribution of the micro-protrusions is assumed to follow a standard Gaussian distribution. Specifically, the height distribution function of the micro-protrusions involved in the rough contact pressure calculation above is... The definition is as follows:

[0089] ;

[0090] This integral is used to describe the situation where the normalized gap is... Statistical contact characteristics at time, among which, For integration variables; For film thickness ratio, For the normalized height variable of the micro-convex body, in actual calculations when When the function value is approximately 0, the oil film robustness index is calculated. To quantify the complex hydrodynamic results into a single index, the input parameters are defined as: 1. Oil film rupture frequency. Statistical full cycle Within, all nodes satisfy ,in, 1. Overall roughness of the contact surface; 2. Thermal sensitivity entropy of the profile. : The scalar value obtained by integration in subsequent step S333; Construct a pre-defined negative correlation mapping function, the specific mathematical expression of which is:

[0091] ;

[0092] in, The total computing scale over the entire cycle is calculated as follows:

[0093] ;

[0094] in, For the number of times the walk, This represents the total number of discrete nodes in the space. The base entropy value has the same dimensions as... To maintain consistency, the value is taken as the system entropy value of the initial design scheme in the unoptimized state;

[0095] Alternatively, you can use parentheses to extract the negative sign:

[0096] ;

[0097] in, It is the product of the total number of computation steps and the total number of nodes in the entire cycle, used to normalize the frequency; The baseline entropy value is taken as the entropy value of the initial design scheme or an empirical constant; it should be clarified here that, given... It is based on the integral of the variance of geometric deformation, see the example, and its physical essence is a measure of geometric uncertainty. Should be taken and A dimensionless geometric scalar, such as the integral of variance in the initial state, rather than the unit of thermodynamic entropy. This is to ensure the physical correctness of dimensionless calculations; For example, take the weighting coefficient. Function characteristics: The range of values ​​for this exponent is... When the frequency of rupture or sensitivity entropy As the number of pistons increases, the exponent exhibits a Gaussian decay trend, strictly satisfying the negative correlation. The mathematical form of exponential decay used here is determined based on statistical analysis of the piston cylinder scoring failure probability distribution during the R&D phase. This form is effective for reducing the frequency of oil film rupture. With the thermal sensitivity entropy of the profile These two physical quantities with different dimensions are mapped to a unified dimensionless probability space. This formula is necessary and forms the basis for standardized decision-making in subsequent optimization algorithms; it maps physical field data into dimensionless probability space values, intuitively representing the system's distance from the cylinder scoring failure boundary. The level of safety.

[0098] Example 5:

[0099] In step four, the oil film robustness index is compared with the failure risk threshold, and the weight configuration of the objective function is adaptively adjusted based on the comparison results. Specifically, this includes:

[0100] S41. Define the optimization objective function, which includes a first sub-objective and a second sub-objective, where the first sub-objective is to minimize frictional power consumption and the second sub-objective is to maximize topology tolerance.

[0101] S42, When the oil film robustness index When the weight is below the failure risk threshold, the system is determined to be in a high-risk mixed lubrication state, and a first weight strategy is generated: reduce the weight of the first sub-objective and increase the weight of the second sub-objective to drive the search algorithm to find a conservative profile with higher geometric tolerance.

[0102] S43, when the oil film robustness index When the failure risk threshold is higher than or equal to the threshold, the system is determined to be in a safe lubrication state, and a second weighting strategy is generated: the weight of the first sub-objective is increased, the weight of the second sub-objective is decreased, and the search algorithm is driven to find an aggressive profile with lower frictional resistance.

[0103] This embodiment further specifies step four of embodiment 1, transforming the qualitative weight adjustment strategy into a programmable adaptive iterative algorithm; constructing a comprehensive cost function to characterize the optimization objective, and defining the total cost function. The weighted combination of the first sub-objective and the second sub-objective:

[0104] ;

[0105] The constraints are ;in, The normalized full-cycle triboelectric power consumption represents the first sub-objective, which needs to be minimized. The normalized profile thermal sensitivity entropy, a lower value indicates a higher topological tolerance, and therefore serves as the inverse metric for the second sub-objective, needing to be minimized in the cost function. To eliminate differences in physical dimensions, the normalization process uses the following formula:

[0106] ;

[0107] in, This is the currently calculated frictional power consumption value. The thermal sensitivity entropy of the current curve is calculated; the extreme value parameter in the denominator is... Taken from the initial sample set or previous history The maximum and minimum values ​​in each iteration are used to prevent gradient explosion or numerical overload.

[0108] Set failure risk threshold ; Execute the first-weighted strategy under high-risk conditions: in response to If the system is determined to be in a high-risk zone, then prioritizing increasing the fault tolerance rate and executing the weight update formula is necessary. Note: An iteration count index is introduced here. This is to distinguish it from the physical time variable of the Reynolds equation in Examples 3 and 4. :

[0109] ;

[0110] ;

[0111] in, The attenuation factor is set to 0.8. This operation forces the algorithm to quickly reduce its focus on triboelectric power consumption and increases the penalty weight for thermal sensitivity entropy. This shifts the optimization focus to improving fault tolerance; the second weighting strategy in a safe state is executed: in response to The system is determined to be in a safe zone, at which point extreme performance can be pursued, and the weight update formula is executed:

[0112] ;

[0113] ;

[0114] in, To determine the increment step, a value of 0.05 is used. The upper limit of the friction weight is set, such as 0.9; this strategy simulates the engineering logic of prioritizing survival over development, and dynamically adjusts the weight. Approaching the physical friction limit while ensuring that cylinder scoring does not occur.

[0115] Example 6:

[0116] Step four also includes:

[0117] S44. A deep reinforcement learning algorithm is used as the search strategy, and the oil film robustness index is used as the environmental reward signal.

[0118] S45. In response to the first weighting strategy, the deep reinforcement learning algorithm punishes geometric features that lead to low robustness by generating negative reward signals, thereby suppressing the generation of profiles with excessive local curvature.

[0119] S46. In response to the second weighting strategy, the deep reinforcement learning algorithm rewards geometric features that lead to low frictional power consumption by generating positive reward signals, thereby exploring the generation of profiles that reduce contact area.

[0120] S47. Output the weighted and optimized design variable increments, and add them to the current skirt profile data to generate corrected profile data.

[0121] This embodiment is a further specification of Embodiment 5, introducing deep reinforcement learning as the core search strategy; a DRL environment is constructed, and deep reinforcement learning algorithms, such as PPO or DDPG, are used as the search strategy; State: originates from the simulation environment, and its physical meaning is the current profile parameters and the current oil film distribution characteristics; specifically, State is defined as a feature vector. ;in, The minimum oil film thickness across the entire computational domain. For the maximum rough contact pressure, The standard deviation of the oil film thickness distribution. The design variable vector for the current time step;

[0122] Action: Originates from the policy network, and its physical meaning is the incremental adjustment of the profile design variables; Reward: Originates from the calculation module, and its physical meaning is the oil film robustness index. Frictional power consumption serves as the basis for the environmental reward signal; to meet code-level reproduction requirements, the specific reward function construction must be clearly defined, and the overall reward function must be defined. The dynamic weights calculated in direct coupling embodiment 5 and The formula is as follows:

[0123] ;

[0124] The first term is a performance gain term, rewarding the reduction of frictional power consumption; the second term is a risk penalty term, affecting the oil film robustness index. Below the threshold A non-zero penalty is generated at this time; The curvature smoothing weight coefficients are dimensionally corrected, in meters. The construction logic of this reward function originates from the control strategy of prioritizing survival over development, as verified in R&D experiments, which includes... The piecewise function setting is necessary to ensure that the algorithm generates a forced nonlinear penalty gradient when the oil film robustness exponent falls below the safety threshold, thereby effectively avoiding the optimization from getting trapped in a high-risk region of local optima; this formula ensures that when Example 5 determines that the system is in a high-risk state and increases At that time, the agent can immediately receive a strong penalty signal, thus being forced to change its search direction; The local curvature of the profile is obtained by applying the discrete Laplacian operator or the second-order central difference to the coordinates of the current node and its neighboring nodes. The specific calculation formula adopts the second-order central difference form (for axial profile):

[0125] ;

[0126] Alternatively, the discrete Laplace operator form can be used, where... This represents the corrected radial displacement value of the current node. This represents the axial grid node spacing;

[0127] Used to quantify the micro-geometric irregularities of the profile surface. Adding or penalizing excessively steep transition zones results in negative deductions, thus inhibiting the generation of this type of line and forcing the agent to explore smoother geometric structures. For the current iteration step Frictional power consumption, when season Equal to the frictional power consumption value of the initial design scheme; implement negative reward and penalty to suppress high-risk characteristics, responding to the first weighted policy, i.e., the high-risk state. The deep reinforcement learning algorithm generates a strong negative reward signal in the first branch of the above formula; this signal is effective against geometric features that lead to low robustness, such as excessive local curvature. The generation of this type of line is suppressed by increasing the penalty for overly steep transition regions, thus forcing the agent to explore smoother geometries.

[0128] Implement positive reward guidance to explore performance limits, in response to the second-weighted policy, namely a safe state. The deep reinforcement learning algorithm generates a positive reward signal based on the second branch of the above formula; this signal rewards the geometric features that lead to low frictional power consumption, thereby encouraging the agent to explore profile generation that reduces the contact area and optimizes the lubricating oil wedge effect; after completing the profile correction output, the agent outputs the weighted optimization of the design variable increment according to the current policy network, and superimposes it on the current skirt profile data to generate the corrected profile data;

[0129] This embodiment utilizes the exploratory capabilities of deep reinforcement learning to discover non-intuitive line features that are difficult for human designers to conceive based on experience. Through tens of thousands of virtual iterations, the AI ​​agent learns how to dynamically adjust the line shape according to the current risk level, thereby achieving design results far exceeding those of traditional gradient optimization algorithms.

[0130] Example 7:

[0131] In step three, the thermal sensitivity entropy of the profile, which characterizes the sensitivity of the profile to changes in the temperature field, is calculated. The specific steps are as follows:

[0132] S331. Introduce a set of random temperature disturbance vectors in the fluid lubrication coupling simulation environment;

[0133] S332. Calculate the variance distribution of the radial deformation of the skirt profile under the action of a random temperature disturbance vector.

[0134] S333. Integrate the variance distribution on the surface domain of the skirt to obtain a scalar value characterizing the deterministic nature of the profile deformation, which is used as the profile thermal sensitivity entropy. The higher the entropy value, the more likely the profile is to generate unpredictable contact drift under thermal shock.

[0135] This embodiment is a concretization of the method for calculating the thermal sensitivity entropy of the profile involved in step three of embodiment 1 and embodiment 4; random temperature disturbances are introduced, and a set of random temperature disturbance vectors are introduced into the fluid lubrication coupling simulation environment. Its dimension is defined as The vector follows a multivariate Gaussian distribution. Symbols are used here. To distinguish it from the total number of nodes defined in Example 2 , The covariance matrix of the temperature perturbation is used to simulate local temperature variations caused by cooling system fluctuations or combustion instabilities; the radial deformation variance is calculated to determine the response of the skirt profile to radial deformation under the action of a random temperature perturbation vector; to address the high computational cost of repeatedly performing thermal-structural coupled finite element analysis (FEA) in the optimization loop, this step pre-constructs a thermal sensitivity matrix. Calculate the radial deformation vector using linearized mapping. :

[0136] ;

[0137] in, for Jacobian matrix of order 1 For the number of nodes, The temperature vector dimension, its elements The thermal sensitivity matrix was obtained by performing finite difference perturbation analysis on the initial model. elements in Constructed based on the following finite difference formula:

[0138] ;

[0139] in, Indicates the first Radial deformation of each node, Indicates the first One temperature variable, The preferred range for the applied temperature perturbation step size is [value range missing]. ;

[0140] In the preferred embodiment, it is assumed that the thermal-structural stiffness characteristics remain linear within a small range of design variable increments; therefore, the Jacobian matrix... The matrix remains constant during optimization; or, in another implementation, each interval... The next iteration, for example Recall the finite element solver to update To correct the nonlinear error introduced by large geometric deformation;

[0141] Based on this, for any position on the surface of the skirt Calculate its radial deformation. Statistical variance under multiple random disturbances or through Direct analytical calculation; statistical variance The calculation formula is as follows:

[0142] ;

[0143] in, The total number of samples for the Monte Carlo random perturbation. For the first Radial deformation value under the second disturbance for Mean radial deformation from each sample;

[0144] Perform entropy integration and quantization, and determine the variance distribution in the skirt surface domain. Integral operations are performed to obtain a scalar value characterizing the deterministic nature of the profile deformation, which is used as the profile thermal sensitivity entropy. The calculation formula is:

[0145] ;

[0146] Among them, the higher the entropy value, the more chaotic and unpredictable the deformation distribution of the profile under thermal shock, that is, the easier it is to generate unpredictable contact drift.

[0147] This embodiment introduces profile thermal sensitivity entropy to evaluate the stability of mechanical structures from an information theory perspective. By penalizing profiles with high entropy values, this method can screen out those blunt profiles that can maintain a relatively stable geometric shape even when the temperature field fluctuates drastically, thereby fundamentally improving the piston's resistance to thermal shock.

[0148] Example 8:

[0149] Step five includes:

[0150] S51. Real-time monitoring of the convergence trend of the oil film robustness index;

[0151] S52. If the oil film robustness index remains above the failure risk threshold and the rate of change of friction power consumption is lower than the preset convergence limit in the continuous preset number of iterations, then the optimization is considered complete.

[0152] S53. Extract the corrected profile data generated in the last iteration, convert it into CNC machining code or a 3D solid model file, and output it as the final piston skirt profile optimization design scheme.

[0153] This embodiment further specifies step five in embodiment 1, describing the optimized termination determination and output implementation; it performs convergence trend monitoring and real-time monitoring of the oil film robustness index. and triboelectric power consumption The convergence trend; during this period, a sliding window is established to record the results of the most recent M iterations; a double convergence judgment is performed, and the optimization is deemed complete only if the following two conditions are met simultaneously; safety constraint: in a continuous preset number of iterations, such as 50 iterations, the oil film robustness index All remain within the failure risk threshold Above; stability constraint: the rate of change of frictional power consumption is lower than the preset convergence limit; implement engineering output, extract the corrected profile data generated in the last iteration; through the computer-aided manufacturing interface, convert it into CNC machining code, such as G code or a general three-dimensional solid model file, such as STEP format; output it as the final piston skirt profile optimization design scheme, which can be directly used to guide high-precision CNC lathes or grinding machines to perform piston forming processing;

[0154] This embodiment ensures that the optimization result is not only theoretically optimal, but also manufacturable and safe in engineering. The strict convergence judgment logic guarantees that the output design scheme must be a robust solution that has been fully verified, avoiding unreliable designs caused by premature convergence or oscillation of the algorithm.

[0155] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. A method for model-based simulation-aided design optimization of a diesel engine piston skirt, characterized in that, The specific steps include: Step 1: Obtain the initial geometric topology data of the diesel engine piston skirt and cylinder liner, and discretize them to construct a three-dimensional contact mesh model; based on thermodynamic boundary conditions and multibody dynamics equations, establish a fluid lubrication coupled simulation environment, and define the design variable space of the skirt profile; Step 2: Construct a time-varying geometric disturbance field and map the time-varying geometric disturbance field onto the inner surface of the cylinder liner of the three-dimensional contact mesh model to simulate the nonlinear deformation of the cylinder liner under thermomechanical coupling and generate a dynamic contact topology containing disturbance features. Step 3: In the fluid lubrication coupling simulation environment, iterative calculations are performed on the piston skirt loaded with the dynamic contact topology. The transient oil film thickness field and local specific pressure distribution in the contact area are extracted. The profile thermal sensitivity entropy, which characterizes the degree of deformation dispersion of the profile under temperature disturbance, is calculated. Combined with the lubrication state boundary, an oil film robustness index reflecting the degree of the system's distance from the failure boundary is constructed. Step 4: Preset failure risk threshold, compare the oil film robustness index with the failure risk threshold, and adaptively adjust the weight configuration of the objective function according to the comparison result to generate corrected profile data for the current disturbance state. Step 5: Based on the corrected profile data, update the design variable space of the skirt profile, and repeat steps 2 to 4 until the oil film robustness index meets the convergence condition, and output the final piston skirt profile optimization design scheme. Step three includes: S31、based on the Reynolds equation and the contact mechanics model, calculate the instantaneous oil film thickness of the kth node at the current time step with rough contact pressure ; S32. Statistically analyze the oil film rupture frequency of all contact units throughout the entire cycle. Use the oil film rupture frequency and the profile thermal sensitivity entropy as evaluation parameters, and calculate the oil film robustness index using a preset negative correlation mapping function. The oil film robustness index is negatively correlated with both the oil film rupture frequency and the thermal sensitivity entropy of the profile. The thermal sensitivity entropy of the skirt profile characterizes the degree of dispersion of radial deformation distortion of the skirt profile when a small disturbance occurs in the temperature field. In step three, the specific steps for calculating the thermal sensitivity entropy of the profile, which characterizes the profile's sensitivity to changes in the temperature field, are as follows: S331. Introduce a set of random temperature disturbance vectors into the fluid lubrication coupling simulation environment; S332. Calculate the variance distribution of the radial deformation of the skirt profile under the action of the random temperature disturbance vector. S333. Perform an integral operation on the variance distribution over the skirt surface domain to obtain a scalar value characterizing the deterministic nature of the profile deformation, which is used as the thermal sensitivity entropy of the profile. The higher the entropy value, the more likely the profile is to generate unpredictable contact drift under thermal shock.

2. The method for model-based simulation-assisted optimization design of a diesel engine piston skirt according to claim 1, characterized in that Step one includes: S11. Obtain the original profile coordinate point set of the piston skirt and the cylindrical surface equation of the cylinder liner under ideal conditions through 3D scanning or parametric modeling. S12. Perform spline interpolation fitting on the original profile coordinate point set to generate a skirt surface model, and divide the skirt surface model and cylinder liner into hexahedral meshes to form several contact elements, each contact element being marked as the kth node. S13. Collect the engine's burst pressure curve, piston heat load distribution data, and lubricating oil rheological parameters under all operating conditions, and input them as boundary conditions into the fluid lubrication coupling simulation environment to initialize the system state vector.

3. The method for model-based simulation-assisted design optimization of a diesel engine piston skirt according to claim 2, characterized in that Step two includes: S21. Collect the static deformation of the cylinder liner under preload and the thermal expansion deformation under different thermal loads. S22. The static deformation amount and the thermal expansion deformation amount are superimposed in the time domain to construct a time-varying geometric disturbance field that varies with the crank angle. S23. Extract the normal displacement component of the k-th node on the inner surface of the cylinder liner, use the time-varying geometric disturbance field to correct the normal displacement component, update the node coordinates of the three-dimensional contact mesh model, and form a dynamic contact topology of a non-standard cylindrical surface.

4. The method for model-based simulation-assisted design optimization of a diesel engine piston skirt according to claim 3, characterized in that In step four, the oil film robustness index is compared with the failure risk threshold, and the weight configuration of the objective function is adaptively adjusted based on the comparison result. Specifically, this includes: S41. Define the optimization objective function, which includes a first sub-objective and a second sub-objective, where the first sub-objective is to minimize frictional power consumption and the second sub-objective is to maximize topology tolerance. S42, when the oil film robustness index When the failure risk threshold is lower than the threshold, the system is determined to be in a high-risk mixed lubrication state, and a first weighting strategy is generated: reduce the weight of the first sub-objective, increase the weight of the second sub-objective, and drive the search algorithm to find a conservative profile with higher geometric tolerance. S43, when the oil film robustness index When the failure risk threshold is higher than or equal to the threshold, the system is determined to be in a safe lubrication state, and a second weighting strategy is generated: the weight of the first sub-objective is increased, the weight of the second sub-objective is decreased, and the search algorithm is driven to find an aggressive profile with lower frictional resistance.

5. The diesel engine piston skirt optimization design method based on model simulation according to claim 4, characterized in that, Step four also includes: S44. A deep reinforcement learning algorithm is used as the search strategy, and the oil film robustness index is used as the environmental reward signal. S45. In response to the first weighting strategy, the deep reinforcement learning algorithm generates a negative reward signal to penalize geometric features that lead to low robustness, thereby suppressing the generation of profiles with excessive local curvature. S46. In response to the second weighting strategy, the deep reinforcement learning algorithm rewards geometric features that result in low frictional power consumption by generating positive reward signals, thereby exploring profile generation that reduces contact area. S47. Output the weighted and optimized design variable increments and add them to the current skirt profile data to generate the corrected profile data.

6. The diesel engine piston skirt optimization design method based on model simulation according to claim 1, characterized in that, Step five includes: S51. Monitor the convergence trend of the oil film robustness index in real time; S52. If, in a series of preset number of iterations, the oil film robustness index remains above the failure risk threshold and the rate of change of frictional power consumption is lower than the preset convergence limit, then the optimization is deemed complete. S53. Extract the modified profile data generated in the last iteration, convert it into CNC machining code or a 3D solid model file, and output it as the final piston skirt profile optimization design scheme.

Images