Explore the vast range of titles available.

Engineering design research integrating artificial intelligence (AI) into computer-aided design (CAD) and computer-aided engineering (CAE) is actively being conducted. This study proposes a deep learning-based CAD/CAE framework in the conceptual design phase that automatically generates 3D CAD designs and evaluates their engineering performance. The proposed framework comprises seven stages: (1) 2D generative design, (2) dimensionality reduction, (3) design of experiment in latent space, (4) CAD automation, (5) CAE automation, (6) transfer learning, and (7) visualization and analysis. The proposed framework is demonstrated through a road wheel design case study and indicates that AI can be practically incorporated into an end-use product design project. Engineers and industrial designers can jointly review a large number of generated 3D CAD models by using this framework along with the engineering performance results estimated by AI and find conceptual design candidates for the subsequent detailed design stage.

A two-step optimization strategy for the inductance control of through-silicon via (TSV)-based 3-dimensional (3-D) inductor is developed based on the stacked autoencoder model and multi-level particle swarm optimization algorithm. The design parameters are divided into two categories, including structural and geometrical parameters. The non-linear relationship among the inductance value, structural and geometrical parameters is described by stacked autoencoder model. Based on the multi-level particle swarm optimization algorithm, the two-step optimization strategy is developed to optimize the structural and geometrical parameters to control the inductance value of TSV-based 3-D inductor. The effectiveness of the developed two-step optimization strategy is verified by two cases. According to the Q3D extractor, the simulated inductance value (1240.55 pH) well agrees with the desired value (1200 pH). Therefore, the inductance value of TSV-based 3-D inductor can be controlled by the developed two-step optimization strategy.

Crash-box is a significant part of automotive passive safety system and serves as a main energy absorption device in frontal impact scenario. In this paper, a novel crash-box integrating an outer thin-walled tube and an inner auxetic cellular core has been designed and comprehensively investigated under axial load. The impact simulation of the auxetic crash-box has been carried out, and the results show that introduction of the auxetic core can improve its energy absorption capacity without increasing too much peak impact force. Based on the sensitivity analysis, the effects of its geometric parameters on the crashworthiness performance have been studied. Finally, collaborative optimization of the auxetic crash-box has been performed to simultaneously improve its crashworthiness under the low-speed (15 km/h) and high-speed (40 km/h) impact cases. In the optimization procedure, the least square support vector regression (LS-SVR) method and an improved particle swarm optimization (IPSO) algorithm with time-varying coefficients have been utilized. The results demonstrate that the optimized crash-box can comprehensively improve the energy absorption and impact force characteristics effectively. The auxetic crash-box and the collaborative optimization approach provide extensive references for the application of auxetic structure in vehicle crashworthiness design.

Due to the limited joint position space and the consideration of reducing the moment of inertia and vibration for industrial robot, the design optimization for rotate vector (RV) reducer is becoming a new and urgent problem in industry. Currently, the existing researches focus on deterministic design optimization, which may cause unreliable designs without considering uncertainties. Therefore, the study focuses on the implementation of reliability-based design optimization (RBDO) to the RV reducer. The aim is to make the RV reducer smaller in size while ensuring a higher reliability. Firstly, a modified advanced mean value (MAMV) method is proposed to improve efficiency and robustness of the advanced mean value method, which encounters inefficiency and numerical instability for the concave or highly nonlinear performance measure functions. Secondly, a mathematical model of RBDO for the RV reducer is established. Thirdly, the proposed MAMV method is integrated into double-loop method to optimize the established mathematical model of RBDO with different target reliability. The results show that the proposed MAMV method is efficient compared with other methods. In addition, the volume of the RV reducer is correspondingly reduced by 9.44%, 7.89%, and 5.66% compared with that before optimization when the target reliabilities are 99.38%, 99.87%, and 99.98%.

Improving the functionality of soft continuum manipulators to expand their application space has always been an important development direction for soft robotics. It remains very challenging to calculate the deformations of soft materials and predict the basic structure of soft fingers under complex objective functions and constraints. This work develops a cable-driven soft robotic gripper with multi-input and multi-output using topology optimization considering geometric nonlinearity, which not only performs adaptive grasping but also enables finer manipulations such as rotating or panning the target. A scheme that can describe adaptive grasping behavior is proposed, which converts the contact between the clamping surface and the object into a boundary condition to circumvent complex contact nonlinearities. An additive hyperelasticity technique is used to overcome numerical instabilities, and the finite element analysis is performed in ANSYS. Numerical simulations and experimental results are performed to demonstrate the effectiveness of the optimization algorithm and to illustrate the application potential of the proposed gripper.

To achieve lightweight of automobile body side structure, a body optimization framework is established. The optimization framework includes numerical simulation, wall thickness, and material set parameterization, design of experiment, surrogate model, NSGA-II, and multi-criteria decision making (MCDM). The finite element model of automobile body side collision is established, and the accuracy of the model is verified by side collision experiment. The problem that it is difficult to embed discrete material variables into the optimization model is solved by using material set parameterization technology. The RBF surrogate model and NSGA-II are applied to the multi-objective lightweight optimization of the body side structure. A hybrid weight & GRA decision method is proposed for Pareto front data mining. Compared with other MCDM methods, the decision results using hybrid weight & GRA are more robust and reasonable. Through the integrated optimization of the wall thickness, material, and performance of the body side structure and the Pareto front data mining, the mass of the body side structure is reduced by 9.21 kg, the lightweight rate is up to 15.93%, and the crashworthiness performance indicators meet the design baseline requirements.

Optimization for the energy-absorbing structures can achieve their better crashworthiness and lightweight performance. However, traditional optimization methods cannot handle categorical responses such as deformation modes. This results some undesirable deformations which often appear in the optimization solution, making it difficult to guarantee the accuracy of optimization. To this end, a machine learning based optimization method for energy-absorbing structures was proposed in this study to remove the undesired deformations. In this method, a DOE method was used to get representative sample points in the design space; the machine learning techniques were adopted to build the prediction models for deformation modes and numerical responses; the Nondominated Sorting Genetic algorithm II (NSGA-II) was utilized for the multi-objective optimization. A case study on optimization of a shrink tube used in train energy absorption was used to verify the effectiveness of the optimization method. The optimization result for the shrink tube illustrated that the machine learning based optimization method can effectively remove the undesirable deformations for energy-absorbing structures. This study may pave a new way to improve the accuracy of energy-absorbing structure optimization.

In this paper, a framework for stochastic optimization of horizontal-axis wind turbine composite blades is presented. It is well known that the structural responses of the wind turbines (e.g., natural frequency, blade tip displacement) are affected by uncertainties in, for instance, wind conditions and material properties. These uncertainties can have an undesirable impact on the performance and reliability of wind turbine blades, and therefore must be accounted for. However, performing the stochastic optimization of wind turbine blades is challenging because of the computational cost and the need to incorporate several disciplines. To make the stochastic problem tractable, a surrogate-based optimization framework using Kriging and support vector machines with adaptive refinement was developed. The framework is based on blade element momentum theory for aerodynamics coupled with a fully parameterized finite element structural model. The framework is used to find the optimal chord and twist distribution of a composite blade and, notably, the optimal control features such as tip-speed ratio and pitch angle with respect to operating wind speeds. The objective function considered is the ratio of mass to the expected value of the Annual Energy Production subjected to several probabilistic constraints on the blade tip deflection, natural frequencies, and failure indices. Uncertainties in material properties, as well as wind conditions are considered. The results of this industrial application demonstrate that the framework can lead, in a reasonable number of function calls, to an optimal composite blade with higher efficiency and robustness to uncertainty.

Stamping is the main manufacturing process for sheet metal parts. However, during the stamping process, based on excessive blank holder force, unreasonable mold design, and other factors, it is easy to generate defects such as cracks in the drawing area and flange wrinkles. In this paper, a novel hybrid model based on a restricted Boltzmann machine and back-propagation neural network is proposed and its validity is verified through different testing functions. Additionally, an improved multi-objective particle swarm optimization (MOPSO) method based on a crowding operator is proposed and compared to several powerful existing algorithms. The proposed method was applied to the process optimization of a double-C part. The sensitivity of the forming quality to different process parameters was analyzed and a novel index was used to describe quality changes. A mapping relationship between the process parameters and forming quality was established based on the proposed hybrid model. Furthermore, optimal process parameters were obtained using MOPSO. The results demonstrated that the proposed method significantly reduces flange wrinkles without excessive thinning and improves the uniformity of formed parts.

The hexagonal periodic structure of the honeycomb is a magic product of nature and shows great mechanical potential. In this work, a type of metamaterial vibration isolator with a honeycomb structure is proposed. The strain, deformation, and natural frequency of the vibration isolator are calculated by the two-dimensional plane finite element model and the simulation accuracies are validated by the experiments. As the design of the metamaterial vibration isolator involves time-consuming finite-element simulation, a multi-fidelity sequential optimization approach based on feasible region analysis (MF-FA) is proposed. In the proposed method, the refined and coarse mesh models are developed as the high- and low-fidelity models, and a two-phase multi-fidelity updating strategy is carried out. In the first phase, sample points are added to the constraint boundary to find the feasible solution quickly, in the second phase, the quality of the feasible optimization solution is gradually improved in the feasible region until it converges to the global optimal solution. Finally, the optimized metamaterial vibration isolator is manufactured and its superiority is validated. Results illustrate that the proposed approach can obtain a desirable optimum, whose natural frequency error between the experimental and the expected value is improved by 12.67% compared with the initial design.