Modern digital manufacturing processes, such as additive manufacturing, are cyber-physical in nature and utilize complex, process-specific simulations for both design and manufacturing. Although computational simulations can be used to optimize these complex processes, they can take hours or days — an unreasonable cost for engineering teams leveraging iterative design processes. Hence, more rapid computational methods are necessary in areas where computation time presents a limiting factor. When existing data from historical examples is plentiful and reliable, supervised machine learning can be used to create surrogate models that can be evaluated orders of magnitude more rapidly than comparable finite element approaches. However, for applications that necessitate computationally-intensive simulations, even generating the training data necessary to train a supervised machine learning model can pose a significant barrier. Unsupervised methods, such as physics-informed neural networks, offer a shortcut in cases where training data is scarce or prohibitive. These novel neural networks are trained without the use of potentially expensive labels. Instead, physical principles are encoded directly into the loss function. This method substantially reduces the time required to develop a training dataset, while still achieving the evaluation speed that is typical of supervised machine learning surrogate models. We propose a new method for stochastically training and testing a convolutional physics-informed neural network using the transient 3D heat equation- to model temperature throughout a solid object over time. We demonstrate this approach by applying it to a transient thermal analysis model of the powder bed fusion manufacturing process.

We have developed a new truss core panel by origami forming to get the higher aspect ratio than that by multi-stage press molding. Our object is to apply the new origami truss core to the train floor. Whether or not this goal can be achieved depends on whether this new origami truss core with a high aspect ratio has excellent sound insulation characteristics. Therefore, as a development of the analysis technology by FEM which accurately estimates the sound insulation characteristics, at first, the relation between the aspect ratio and the sound insulation performance is discussed in the flat plate with one core. So far, sound insulation simulations using FEM did not match with theory of the mass law. However, this can be achieved by setting the end of the transmitted side to be a nonreflective boundary. In this paper, to generalize this method, it is determined theoretically that the sound pressures from the FEM can be separated accurately into the sound pressures of the forward and backward waves from Helmholtz’s equation. Then, the sound insulation characteristics of a flat plate obtained using the proposed theoretical method and the conventional method are compared while assuming that the flat plate is a rigid body. In addition, the validity of the proposed method is confirmed by evaluating the effect of attaching a sound absorbing material to the plate. Furthermore, application of the proposed method to a flat plate with a truss core and an examination of whether a high aspect ratio is advantageous for sound insulation are also presented.

The goal of this research is to demonstrate the applicability of the Human Error and Functional Failure Reasoning (HEFFR) framework to complex engineered systems. Human errors are cited as a root cause of a majority of accidents and performance losses in complex engineered systems. However, a closer look would reveal that such mishaps are often caused by complex interactions between human fallibilities, component vulnerabilities, and poor design. Hence, there is a growing call for risk assessments to analyze human errors and component failures in combination. The HEFFR framework was developed to enable such combined risk assessments. Until now, this framework has only been applied to simple problems, and it is prone to be computationally heavy as complexity increases. In this research, we introduce a modular HEFFR assessment approach as means of managing the complexity and computational costs of the HEFFR simulations of complex engineered systems. Then, we validate the proposed approach by testing the consistency of the HEFFR results between modular and integral assessments and between different module partitioning assessments. Next, we perform a risk assessment of a train locomotive using the modular approach to demonstrate the applicability of the HEFFR framework to complex engineered systems. The results show that the proposed modular approach can produce consistent results while reducing complexity and computational costs. Also, the results from the train locomotive HEFFR analysis show that the modular assessments can be used to produce risk insights similar to integral assessments but with a modular context.

Estimating the form and functional performance of a design in the early stages can be crucial for a designer for effective ideation Humans have an innate ability to guess the size, shape, and type of a design from a single view. The brain fills in the unknowns in a fraction of a second. However, humans may struggle with estimating the performance of designs in the early stages of the design process without making prototypes or doing back-of-the-envelope calculations. In contrast, machines need information about the full 3D model of a design to understand its structure. Machines can estimate the performance using pre-defined rules, expensive numerical simulations, or machine learning models. In this paper, we show how information about the form and functional performance of a design can be estimated from a single image using machine learning methods. Specifically, we leverage the image-to-image translation method to predict multiple projections of an image-based design. We then train deep neural network models on the predicted projections to provide estimates of design performance. We demonstrate the effectiveness of our method by predicting the aerodynamic performance from images of aircraft models. To estimate ground truth aero-dynamic performance, we run CFD simulations for 4045 3D aircraft models from the ShapeNet dataset and use their lift-to-drag ratio as the performance metric. Our results show that single images do carry information for both form and functional performance. From a single image, we are able to produce six additional images of a design in different orientations, with an average Structural Similarity Index score of 0.872. We also find image-translation methods provide a promising direction in estimating the performance of design. Using multiple images of a design (gathered through image-translation) to predict design performance yields a recall value of 47%, which is 14% higher than a base guess, and 3% higher than using a single image. Our work identifies the potential and provides a framework for using a single image to predict the form and functional performance of a design during the early-stage design process. Our code and additional information about our work are available at http://decode.mit.edu/projects/formfunction/ .

Robotic Wire Arc Additive Manufacturing (WAAM) is the layer-by-layer deposition of molten metal to build a three-dimensional part. In this process, the fed metal wire is melted using an electric arc as a heat source. The process is sensitive to the arc conditions, such as arc length. While building WAAM parts, the metal beads overlap at corners causing material accumulation. Material accumulation is undesirable as it leads to uneven build height and process failures caused by arc length variation. This paper introduces a deposition speed regulation scheme to avoid the corner accumulation problem and build parts with uniform build height. The regulated speed has a complex relationship with the corner angle, bead geometry, and molten metal dynamics. So we need to train a model that can predict suitable speed regulations for corner angles encountered while building the part. We develop an unsupervised learning technique to characterize the uniformity of the bead profile of a WAAM built layer and check for anomalous bead profiles. We train a model using these results that can predict suitable speed regulation parameters for different corner angles. We test this model by building a WAAM part using our speed regulation scheme and validate if the built part has uniform build height and reduced corner defects.

Soil-borne plant-parasitic nematodes are microscopic, eellike roundworms. The root-knot nematodes ( Meloidogyne spp.) and root-lesion nematodes ( Pratylenchus vulnus ) are among the most damaging in California, which are difficult to control and can spread easily in soil on tools, boots, and infested plants. Root-knot nematodes can attack many different crops, including nut and fruit trees, usually cause unusual swellings, called galls, on affected plants’ roots. It is not easy to recognize the infestations of these nematodes. For instance, researchers need to dig up walnut trees with symptoms, wash or gently tap the soil from the roots, and examine the roots for galls. The nematode extraction procedures, identification, and enumeration under a microscope are tedious and time-consuming. Therefore, in this article, the authors proposed to use a low-cost contactless radio frequency tridimensional sensor “Walabot,” and Deep Neural Networks (DNNs), to perform the early detection of nematodes in a walnut site. Radiofrequency reflectance of walnut leaves from different nematode infestation levels was measured. The hypothesis was that waveforms generated from walnut leaves can estimate the damage caused by nematodes. DNNs with Tensor-Flow were used to train and test the proposed method. Results showed that the Walabot predicted nematode infestation levels with an accuracy of 82%, which showed great potentials for early detection of nematodes.

We present a simulation-based classification approach for large deployed structures with localized operational excitations. The method extends the two-level Port-Reduced Reduced-Basis Component (PR-RBC) technique to provide faster solution estimation to the hyperbolic partial differential equation of time-domain elastodynamics with a moving load. Time-domain correlation function-based features are built in order to train classifiers such as Artificial Neural Networks and Support-Vector Machines and perform damage detection. The method is tested on a bridge-shaped structure with a moving vehicle (playing the role of a digital twin) in order to detect cracks’ existence. Such problem has 45 parameters and shows the merits of the two-level PR-RBC approach and of the correlation function-based features in the context of operational excitations, other nuisance parameters and added noise. The quality of the classification task is enhanced by the sufficiently large synthetic training dataset and the accuracy of the numerical solutions, reaching test classification errors below 0.1% for disjoint training set of size 7 × 10 3 and test set of size 3 × 10 3 . Effects of the numerical solutions accuracy and of the sensors locations on the classification errors are also studied, showing the robustness of the proposed approach and the importance of constructing a rich and accurate representation of possible healthy and unhealthy states of interest.

Systematic enumeration and identification of unique 3D spatial topologies of complex engineering systems such as automotive cooling layouts, hybrid-electric power trains, and aero-engines are essential to search their exhaustive design spaces to identify spatial topologies that can satisfy challenging system requirements. However, efficient navigation through discrete 3D spatial topology options is a very challenging problem due to its combinatorial nature and can quickly exceed human cognitive abilities at even moderate complexity levels. Here we present a new, efficient, and generic design framework that utilizes mathematical spatial graph theory to represent, enumerate, and identify distinctive 3D topological classes for an abstract engineering system, given its system architecture (SA) — its components and interconnections. Spatial graph diagrams (SGDs) are generated for a given SA from zero to a specified maximum crossing number. Corresponding Yamada polynomials for all the planar SGDs are then generated. SGDs are categorized into topological classes, each of which shares a unique Yamada polynomial. Finally, for each topological class, one 3D geometric model is generated for an SGD with the fewest interconnect crossings. Several case studies are shown to illustrate the different features of our proposed framework. Design guidelines are also provided for practicing engineers to aid the utilization of this framework for application to different types of real-world problems.

In this work, we investigate the problem of level curve tracking in unknown scalar fields using a limited number of mobile robots. We design and implement a long short term memory (LSTM) enabled control strategy for a mobile sensor network to detect and track desired level curves. Based on the existing work of cooperative Kalman filter, we design an LSTM-enhanced Kalman filter that utilizes the sensor measurements and a sequence of past fields and gradients to estimate the current field value and gradient. We also design an LSTM model to estimate the Hessian of the field. The LSTM enabled strategy has some benefits such as it can be trained offline on a collection of level curves in known fields prior to deployment, where the trained model will enable the mobile sensor network to track level curves in unknown fields for various applications. Another benefit is that we can train using larger resources to get more accurate models, while utilizing a limited number of resources when the mobile sensor network is deployed in production. Simulation results show that this LSTM enabled control strategy successfully tracks the level curve using a mobile multi-robot sensor network.

One of the most important steps in the structural synthesis of planetary gear trains is to eliminate degenerate structures. First, the graph theory is used to represent planetary gear trains (PGTs). Second, a procedure is developed to identify fundamental geared entities (FGEs). Further, the single-planet FGEs are shown to have one-DOF and, therefore, cannot constitute a degenerate structure. It is this that allows a significant reduction in the calculation in relation to other methods of diagnosing degenerate structures. Third, using the concepts of FGEs and the notation of the associated adjacency matrix, an algorithm is developed for the detection of degenerate structures in PGTs. The algorithm is based on the fact that any degenerate structure is a PGT formed by two fundamental geared entities with common edges and/or vertices equal to or more than 3. Forth, the concept of connectivity between single-planet FGEs is introduced and a simple, straightforward approach for deducting the connectivity matrix from the adjacency matrix is developed. The new vertex-edge mobility criterion does not require combinatorial analysis. Besides, the method is applicable to one and multiple degrees of freedom PGTs, it is also applicable to multi-planet PGTs and complex PGTs, including contrary examples found in the literature.

A task motion trajectory usually needs to be determined for the training process and mechanism design for rehabilitation patients since they are not capable of providing a normal motion. In this paper, a machine-learning-based approach of gait trajectory prediction for lower limb rehab patients is proposed to provide the basis for the design of simple 1-degree-of-freedom (DOF) rehab mechanisms. First, a large amount of gait trajectories from various healthy volunteers are collected along with their body parameters, and a normalization method is presented to trim/expand these trajectory samples to a standard length and timing while retaining their shape and velocity information. Then, these normalized gait samples are clustered and regressed into a limited number of representative trajectories with K-means algorithm, and the cluster index is recorded as the label for each trajectory. Next, a genetic-algorithm-optimized support vector machine method is adopted to train a classifier for the trajectories, obtaining the correspondence between body parameters and cluster labels of gait trajectories. As a result, once a group of body parameters are input into the classifier, it can predict a most suitable gait trajectory for the specific patient. It shows that the accuracy of trajectory prediction reaches 96% both on training set and test set which verifies the effectiveness of the method. In the end, a 1-DOF gait rehab mechanism design example is provided to illustrate the application of the proposed method. Taking the predicted result from the classifier as the task motion trajectory for the synthesis of mechanisms, a 1-DOF six-bar mechanism is designed and the patient-mechanism matching can be realized.

The design of valve lift curves for valve train cam systems is formulated as an optimization problem, in which the maximum principle is applied. The solution is obtained by utilizing the maximum principle approach for an equivalent linear quadratic optimal control problem. The outcome of the design shows trade-offs between several design criteria. The method is applied to the specification of the lift curve for a finger-follower cam system, and design charts are developed for different choices of design parameters. The method presented in this study provides a unified way of valve lift curve design for those cam systems where the objective functions can be clearly defined.

The application of fault tree techniques to the analysis of gear trains is considered. The techniques involve the decomposition of the system into a logic diagram or fault tree in which certain basic or primary events lead to a specified top event which signifies the total failure of the system. The fault trees are used to obtain minimal cut sets from which the modes of system failure and, hence, the reliability measure for the top event can be calculated. The method of constructing fault trees and the subsequent estimation of reliability of the system is illustrated through a nine speed compound gear train and a five speed epicyclic gear train.

This paper describes several mechanical balancers, based on the planetary gear trains known as the hypotrochoid and epitrochoid trains, for reducing or eliminating second-order out-of-balance in mechanical systems. It is shown that by proper arrangement of the planetary gear trains, a balancer can be obtained for the elimination of second-order shaking forces or second-order shaking moments or a combination of both shaking forces and moments. The advantage of this type of balancer is that the carrier of the gear train needs only to run at the primary speed of the mechanical system to be balanced. Therefore, the balancer can be designed to be concentric with the primary rotating shaft of the machine using the primary shaft as the carrier. For example, for the balance of the second-order shaking force of an in-line four-cylinder four-stroke internal combustion piston engine, the balancer can be placed on the third main bearing or one on each of the second and fourth main bearings.

This paper uses the Euler parameters of the motion of the planet of a spherical epicyclic gear train to obtain a curve on the surface of a hypersphere in four dimensions. This curve, called the image curve, represents the rotational motion of the planet as it rolls without slipping on the fixed gear. Two image curves are obtained for two different choices of moving and fixed reference frames and it is shown that they are related by an orthogonal transformation in four dimensional space. The differential properties of the image curve are computed and it is found that the curvature is constant and the torsion is periodic in the motion parameter. A reference position is chosen and the canonical frame and instantaneous invariants of the motion are determined in terms of the dimensions of the gear train.

The cycloidal speed reducer, or cycloid drive, is an epicyclic gear train in which the profile of the planet gear is an epitrochoid and the annular sun gear has rollers as its teeth. The cycloid drive has very high efficiency and small size, in comparison with a conventional gear mechanism, making it an attractive candidate for limited space applications. On the other hand, in this type of transmissions there exist two major drawbacks, namely, backlash and torque ripple Backlash, the angle through which the output shaft can rotate when the input shaft is held fixed, has a degrading effect on the output accuracy. Torque ripple, the variation in mechanical advantage as the input shaft rotates, causes vibrations and could lead to dynamic instability of the machinery. If the cycloid drive were manufactured to the ideal dimensions, there would be no backlash nor torque ripple. However, in reality, there will always be some machining tolerances. In this paper an analytical model is developed which models the cycloid drive with machining tolerances. This model is used in Part II of this investigation to determine the effect of machining tolerances on backlash and torque ripple. As a result, simple and practical equations for design synthesis of this type of drives are formulated.

The application of fault tree techniques to the analysis of gear trains is considered. The techniques involve the decomposition of the system into a logic diagram or fault tree in which certain basic or primary events lead to a specified top event which signifies the total failure of the system. The fault trees are used to obtain minimal cut sets from which the modes of system failure and hence the reliability measure for the top event can be calculated. The method of constructing fault trees and the subsequent estimation of reliability of the system is illustrated through a nine speed compound gear train and a five speed epicyclic gear train.

Dynamic frequency response of mechanical and structural systems is of interest in design problems that are subjected to harmonically varying external loads caused by the reciprocating power train or other rotating machineries such as motors, fans, compressors, and forging hammers [1]. For example, airplane body and wing structures are subjected to a harmonic load transmitted from the propulsion system. Also, ship vibrations resulting from the propeller and engine excitation can cause noise problem, cracks, fatigue failure of tailshaft, and discomfort to crew. When a machine or any structure oscillates in some form of periodic or random motion, the motion generates alternating pressure waves that propagate from the moving surface at the velocity of sound. For instance, the interior sound pressure in an automobile compartment can occur when the input forces transmitted from road and power train excite the vehicle compartment boundary panels. These motions with frequencies between 20 Hz and 20 KHz stimulate the hearing mechanism of human [2].

The purpose of this paper is to propose a new graph representation to represent the kinematic structure of a planetary spur-gear train efficiently. Based on the graph representation, the kinematic analysis of planetary spur-gear trains is largely simplified. An interactive computer program is developed for the kinematic analysis of planetary spur-gear trains with any numbers of degrees of freedom. By only inputting the graph representation of a planetary spur-gear trains and the data of the mating gear pairs, all possible fundamental circuits are determined and the rotational displacement equations are derived and solved automatically.

In this paper, the trajectory of bevel planetary gear trains has been studied. The parametric equations of trajectory are derived. It is shown that the trajectory generated by a tracer point on the planet of a bevel planetary gear train is analogous to that of a spur planetary gear train. Two cases, gear ratio equal to one and two, are presented in detail including the geometric description, plane of symmetry, extent of trajectory, number of nodes (cusps) and their locations. The criteria for the existence of cusps are verified algebraically, and interpreted from geometrical point of view.