Humans are daily presented with tasks that they complete with little effort or even consideration of the planning that goes into the movement. Movements such as manual manipulation tasks are completed with ease, even though the complexities and years of learned behavior are largely hidden from the person. Digital human modeling (DHM) and specifically optimization-based posture and motion prediction methodologies have employed numerical methods in order to simulate/predict/analyze human movements. However, these movements are heavily constrained such that the planning of the motion/posture is explicitly provided in the formulation of the problem. This paper presents the addition of cognitive principles into the optimization-based posture and motion prediction formulations. The simulation/prediction of manual manipulation tasks is considered such that a single formulation can accomplish multiple tasks. It adopts a theory from cognitive psychology referred to as the end-state comfort effect in order to derive general constraints for the prediction of the initial and final posture states that frame the movement related to the manual manipulation task. It considers multiple tasks from the literature that have been heavily studied through experimentation in order to evaluate the efficacy of the formulation. The results show strong correlation with observations reported in the literature.

In this paper, we introduce a physics-driven regularization method for training of deep neural networks (DNNs) for use in engineering design and analysis problems. In particular, we focus on the prediction of a physical system, for which in addition to training data, partial or complete information on a set of governing laws is also available. These laws often appear in the form of differential equations, derived from first principles, empirically validated laws, or domain expertise, and are usually neglected in a data-driven prediction of engineering systems. We propose a training approach that utilizes the known governing laws and regularizes data-driven DNN models by penalizing divergence from those laws. The first two numerical examples are synthetic examples, where we show that in constructing a DNN model that best fits the measurements from a physical system, the use of our proposed regularization results in DNNs that are more interpretable with smaller generalization errors, compared with other common regularization methods. The last two examples concern metamodeling for a random Burgers’ system and for aerodynamic analysis of passenger vehicles, where we demonstrate that the proposed regularization provides superior generalization accuracy compared with other common alternatives.

Implementing the concept of a digital twin in full production provides enough data on each individual assembly for real-time control of production processes. Taking advantage of this opening, this paper proposes individualized locator adjustments as a new method to improve the geometrical quality of assemblies. In this method, all locators in the assembly fixture can be adjusted for each individual assembly based on the scan data of the mating parts of that assembly. The optimal adjustment of every locator for each individual assembly is obtained using an optimization algorithm and nonrigid variation simulation tools (computer-aided tolerancing tools). This method is applied to three industrial cases and geometrical variations and the mean deviation from nominal positions are compared to nonindividualized adjustments and also when there are no adjustments. The results show that applying this method, an improvement of up to 81% in geometrical variation and 78% in the mean deviation of assemblies can be obtained compared to assemblies without adjustments. These improvements are 60% and 57% higher than nonindividualized adjustments of locators for the variation and the mean deviation, respectively. Moreover, a modification on the optimization algorithm has been proposed that reduces the amount of required adjustments.

The ability to track human operators' hand usage when working in production plants and factories is critically important for developing realistic digital factory simulators as well as manufacturing process control. We propose a proof-of-concept instrumented glove with only a few strain gage sensors and a microcontroller that continuously tracks and records the hand configuration during actual use. At the heart of our approach is a trainable system that can predict the fourteen joint angles in the hand using only a small set of strain sensors. First, ten strain gages are placed at various joints in the hand to optimize the sensor layout using the English letters in the American Sign Language (ASL) as a benchmark for assessment. Next, the best sensor configurations for three through ten strain gages are computed using a support vector machine (SVM) classifier. Following the layout optimization, our approach learns a mapping between the sensor readouts to the actual joint angles optically captured using a Leap Motion system. Five regression methods including linear, quadratic, and neural regression are then used to train the mapping between the strain gage data and the corresponding joint angles. The final proposed model involves four strain gages mapped to the fourteen joint angles using a two-layer feed-forward neural network (NN).

Reducing the volume of support structures is a critical means for saving materials and budgets of additive manufacturing, and tree structure is an effective topology for this purpose. Although a few articles in literature and commercial software have been devoted to developing tree-supports, those tree-supports are generated based on geometry optimization or user-defined parameters, which cannot guarantee a minimum volume with robust fabrication guarantee. To address this issue, we propose a set of formulas for stably growing the tree-supports with physical constraints based on 3D printing experiments using fused decomposition modelling (FDM) machines, and a volume minimization mechanism using a hybrid of particle swarm optimization (PSO) method and a greedy algorithm. We show that this combination is effective in reducing the volume of tree-supports and the simulations reveal that the volume curves monotonically descent to a constant within a short time, and our experimental results show that the models with the tree-supports can be manufactured stably.

During present offshore gas-condensate production, multiphase flow-meters, due to its exceedingly high cost, are being substituted by a soft sensing (SS) technique for estimating total and single-well flowrates through sensor measurements and physical models. In this work, the inverse problem is solved by data reconciliation (DR), minimizing weighted sum of errors with constraints integrating multiple two-phase flow models. The DR problem is solved by parallel genetic algorithm (PGA) without complex calculations required by conventional optimization. The newly developed SS method is tested by data from a realistic gas-condensate production system. The method is proved of good accuracy and robustness with invalid individual pressure sensor or unavailable total flowrate measurements. Meanwhile, the proposed method shows good parallel performance and the time cost of each DR process can meet the demand of engineering application.

Tolerance analysis aims toward the verification impact of the individual tolerances on the assembly and functional requirements of a mechanism. The manufactured products have several types of contact and are inherent in imperfections, which often causes the failure of the assembly and its functioning. Tolerances are, therefore, allocated to each part of the mechanism in purpose to obtain an optimal quality of the final product. Three main issues are generally defined to realize the tolerance analysis of a mechanical assembly: the geometrical deviations modeling, the geometrical behavior modeling, and the tolerance analysis techniques. In this paper, a method is proposed to realize the tolerance analysis of an over-constrained mechanical assembly with form defects by considering the contacts nature (fixed, sliding, and floating contacts) in its geometrical behavior modeling. Different optimization methods are used to study the different contact types. The overall statistical tolerance analysis of the over-constrained mechanical assembly is carried out by determining the assembly and the functionality probabilities based on optimization techniques combined with a Monte Carlo simulation (MCS). An application to an over-constrained mechanical assembly is given at the end.

Topology optimization for additive manufacturing has been limited to the design of single-piece components that fit within the printer's build volume. This paper presents a gradient-based multicomponent topology optimization method for structures assembled from components built by powder bed additive manufacturing (MTO-A), which enables the design of multipiece assemblies larger than the printer's build volume. Constraints on component geometry for powder bed additive manufacturing are incorporated in a density-based topology optimization framework, with an additional design field governing the component partitioning. For each component, constraints on the maximum allowable build volume (i.e., length, width, and height) and the elimination of enclosed cavities are imposed during the simultaneous optimization of the overall topology and component partitioning. Numerical results of the minimum compliance designs revealed that manufacturing constraints, previously applied to single-piece topology optimization, can unlock richer design exploration space when applied to multicomponent designs.

The goal of this research is to optimize an object's macroscopic topology and localized gradient material properties (GMPs) subject to multiple loading conditions simultaneously. The gradient material of each macroscopic cell is modeled as an orthotropic material where the elastic moduli in two local orthogonal directions we call x and y can change. Furthermore, the direction of the local coordinate system can be rotated to align with the loading conditions on each cell. This orthotropic material is similar to a fiber-reinforced material where the number of fibers in the local x and y -directions can change for each cell, and the directions can as well be rotated. Repeating cellular unit cells, which form a mesostructure, can also achieve these customized orthotropic material properties. Homogenization theory allows calculating the macroscopic averaged bulk properties of these cellular materials. By combining topology optimization with gradient material optimization and fiber orientation optimization, the proposed algorithm significantly decreases the objective, which is to minimize the strain energy of the object subject to multiple loading conditions. Additive manufacturing (AM) techniques enable the fabrication of these designs by selectively placing reinforcing fibers or by printing different mesostructures in each region of the design. This work shows a comparison of simple topology optimization, topology optimization with isotropic gradient materials, and topology optimization with orthotropic gradient materials. Finally, a trade-off experiment shows how different optimization parameters, which affect the range of gradient materials used in the design, have an impact on the final objective value of the design. The algorithm presented in this paper offers new insight into how to best take advantage of new AM capabilities to print objects with gradient customizable material properties.

Optimal layouts for structural design have been generated using topology optimization approach with a wide variety of objectives and constraints. Minimization of compliance is the most common objective but the resultant structures often have stress concentrations. Two new objective functions, constructed using an upper bound of von Mises stress, are presented here for computing design concepts that avoid stress concentration. The first objective function can be used to minimize mass while ensuring that the design is conservative and avoids stress concentrations. The second objective can be used to tradeoff between maximizing stiffness versus minimizing the maximum stress to avoid stress concentration. The use of the upper bound of von Mises stress is shown to avoid singularity problems associated with stress-based topology optimization. A penalty approach is used for eliminating stress concentration and stress limit violations which ensures conservative designs while avoiding the need for special algorithms for handling stress localization. In this work, shape and topology are represented using a density function with the density interpolated piecewise over the elements to obtain a continuous density field. A few widely used examples are utilized to study these objective functions.

This paper addresses some important theoretical issues for constrained least-squares fitting of planes and parallel planes to a set of points. In particular, it addresses the convexity of the objective function and the combinatorial characterizations of the optimality conditions. These problems arise in establishing planar datums and systems of planar datums in digital manufacturing. It is shown that even when the set of points (i.e., the input points) are in general position, (1) a primary planar datum can contact 1, 2, or 3 input points, (2) a secondary planar datum can contact 1 or 2 input points, and (3) two parallel planes can each contact 1, 2, or 3 input points, but there are some constraints to these combinatorial counts. In addition, it is shown that the objective functions are convex over the domains of interest. The optimality conditions and convexity of objective functions proved in this paper will enable one to verify whether a given solution is a feasible solution, and to design efficient algorithms to find the global optimum solution.

Fourier descriptor (FD)-based path synthesis algorithms for generation of planar four-bar mechanisms require assigning time parameter values to the given points along the path. An improper selection of time parameters leads to poor fitting of the given path and suboptimal four-bar mechanisms while also ignoring a host of mechanisms that could be potentially generated otherwise. A common approach taken is to use uniform time parameter values, which does not take into account the unique harmonic properties of the coupler path. In this paper, we are presenting a nonuniform parametrization scheme in conjunction with an objective function that provides a better fit, leverages the harmonics of the four-bar coupler, and allows imposing additional user-specified constraints.

To obtain real-time interactions in the virtual cockpit system (VCS), a real-time trajectory generation method based on dynamical nonlinear optimization and regression prediction for the haptic feedback manipulator (HFM) is presented in this paper. First, a haptic feedback system based on servoserial manipulator is constructed. Then, the trajectory planning problem for the HFM is formulated as a nonlinear optimization problem to balance the motion time and power consumption and ensure the safety of physical human–robot interactions (pHRI). Multiple optimization problems are solved to generate the optimal database off-line. Finally, the classified multivariate (CM) regression method is presented to learn the database and generate optimal trajectories with arbitrary initial and objective positions on-line. Results show that trajectories with rapidity, safety, and lower power consumption can be generated in real-time by this method, which lay a basis of haptic interactions in the VCS.

The mass production paradigm strives for uniformity, and for assembly operations to be identical for each individual product. To accommodate geometric variation between individual parts, tolerances are introduced into the design. However, this method can yield suboptimal quality. In welded assemblies, geometric variation in ingoing parts can significantly impair quality. When parts misalign in interfaces, excessive clamping force must be applied, resulting in additional residual stresses in the welded assemblies. This problem may not always be cost-effective to address simply by tightening tolerances. Therefore, under new paradigm of mass customization, the manufacturing approach can be adapted on an individual level. This paper focuses on two specific mass customization techniques: permutation genetic algorithms (GA) and virtual locator trimming. Based on these techniques, a six-step method is proposed, aimed at minimizing the effects of geometric variation. The six steps are nominal reference point optimization, permutation GA configuration optimization, virtual locator trimming, clamping, welding simulation, and fatigue life evaluation. A case study is presented, which focuses on the selective assembly process of a turbine rear structure of a commercial turbofan engine, where 11 nominally identical parts are welded into a ring. Using this simulation approach, the effects of using permutation GAs and virtual locator trimming to reduce variation are evaluated. The results show that both methods significantly reduce seam variation. However, virtual locator trimming is far more effective in the test case presented, since it virtually eliminates seam variation. These results underscore the potential of virtual trimming and GAs in manufacturing, as a means both to reduce cost and increase functional quality.

There are a large number of real-world engineering design problems that are multi-objective and multiconstrained, having uncertainty in their inputs. Robust optimization is developed to obtain solutions that are optimal and less sensitive to uncertainty. Since most of complex engineering design problems rely on time-consuming simulations, the robust optimization approaches may become computationally intractable. To address this issue, an advanced multi-objective robust optimization approach based on Kriging model and support vector machine (MORO-KS) is proposed in this work. First, the main problem in MORO-KS is iteratively restricted by constraint cuts formed in the subproblem. Second, each objective function is approximated by a Kriging model to predict the response value. Third, a support vector machine (SVM) classifier is constructed to replace all constraint functions classifying design alternatives into two categories: feasible and infeasible. The proposed MORO-KS approach is tested on two numerical examples and the design optimization of a micro-aerial vehicle (MAV) fuselage. Compared with the results obtained from other MORO approaches, the effectiveness and efficiency of the proposed MORO-KS approach are illustrated.

As the core component of an automobile, the internal combustion engine (ICE) nowadays is still a typical complex engineering system. Tolerance design for ICEs is of great importance since small changes in the dimensions and clearances of ICE components may result in large variations on the performance and cost of manufactured products. In addition, uncertainty in tolerance design has great impact on the engine performance. Although tolerance optimization for the key components of ICEs has been discussed, few of them take uncertainty into consideration. In this regard, robust optimization (RO) for the tolerances of ICEs remains a critical issue. In this work, a novel RO approach is proposed to deal with the tolerance optimization problem for ICEs under parameter and model uncertainties, even considering metamodeling uncertainty from Gaussian processes (GP). A typical parameter uncertainty in ICEs exists in the rotation speed which can vary randomly due to the inherent randomness. AVL EXCITE software is used to build the simulation models of ICE components, which brings in model uncertainty. GP models are used as the analysis model in order to combine the corresponding simulation and experimental data together, which introduces metamodeling uncertainty. The proposed RO approach provides a general and systematic procedure for determining robust optimal tolerances and has competitive advantages over traditional experience-based tolerance design. In addition to the ICE example, a numerical example is utilized to demonstrate the applicability and effectiveness of the proposed approach.

Fixture locators are used to precisely locate and stably support a workpiece so that the desired position and orientation (pose) of the workpiece relative to the cutting tool can be maintained during machining or inspection process. It is believed that manufacturing errors of locators and locating datum surfaces are key factors for the pose error between the workpiece and the cutting tool. Optimizing the layout of locators is helpful to reduce the pose error so as to improve machining accuracy of the workpiece. In order to minimize the pose error, we introduced, for the first time, a singular value decomposition (SVD) technique for the location matrix to derive error amplification factors (EAF) in six degrees-of-freedom of the workpiece. The EAF principle defines the maximal singular value, the condition number, the product of all singular values and the manipulability as the maximal error amplification factor, the relative error amplification factor, the error ellipsoid volume and the location stability, respectively. The four defined indices taken as objective functions are optimized, by a nondominated sort genetic algorithm (NSGA-II), so that an optimal layout of locators is obtained due to the minimization of the pose error. Also, the feasibility of the proposed method was comprehensively validated by simulation and machining experiments.

The precise control of mass and energy deposition associated with additive manufacturing (AM) processes enables the topological specification and realization of how space can be filled by material in multiple scales. Consequently, AM can be pursued in a manner that is optimized such that fabricated objects can best realize performance specifications. In the present work, we propose a computational multiscale method that utilizes the unique meso-scale structuring capabilities of implicit slicers for AM, in conjunction with existing topology optimization (TO) tools for the macro-scale, in order to generate structurally optimized components. The use of this method is demonstrated on two example objects including a load bearing bracket and a hand tool. This paper also includes discussion concerning the applications of this methodology, its current limitations, a recasting of the AM digital thread, and the future work required to enable its widespread use.

In many system-engineering problems, such as surveillance, environmental monitoring, and cooperative task performance, it is critical to allocate limited resources within a restricted area optimally. Static coverage problem (SCP) is an important class of the resource allocation problem. SCP focuses on covering an area of interest so that the activities in that area can be detected with high probabilities. In many practical settings, primarily due to financial constraints, a system designer has to allocate resources in multiple stages. In each stage, the system designer can assign a fixed number of resources, i.e., agents. In the multistage formulation, agent locations for the next stage are dependent on previous-stage agent locations. Such multistage static coverage problems are nontrivial to solve. In this paper, we propose an efficient sequential sampling algorithm to solve the multistage static coverage problem (MSCP) in the presence of resource intensity allocation maps (RIAMs) distribution functions that abstract the event that we want to detect/monitor in a given area. The agent's location in the successive stage is determined by formulating it as an optimization problem. Three different objective functions have been developed and proposed in this paper: (1) L2 difference, (2) sequential minimum energy design (SMED), and (3) the weighted L2 and SMED. Pattern search (PS), an efficient heuristic algorithm has been used as optimization algorithm to arrive at the solutions for the formulated optimization problems. The developed approach has been tested on two- and higher dimensional functions. The results analyzing real-life applications of windmill placement inside a wind farm in multiple stages are also presented.

In this paper, we present a pattern development method for soft product design. We utilize a surface fattening method based on a mass-spring model to create 2D patterns unfolding from a three-dimensional (3D) model. Multilevel meshes are proposed to expedite the flattening process, and a boundary optimization method is employed to guarantee 2D patterns can be sewn well. We apply the proposed method to the design of real soft products. Experimental results show that it can deal with complex surfaces efficiently and robustly, and manufactured products are satisfactory.