Generating Pseudo-Data to Enhance the Performance of Classification-Based Engineering Design: A Preliminary Investigation

Machine learning for classification has been used widely in engineering design, for example, feasible domain recognition and hidden pattern discovery. Training an accurate machine learning model requires a large dataset; however, high computational or experimental costs are major issues in obtaining a large dataset for real-world problems. One possible solution is to generate a large pseudo dataset with surrogate models, which is established with a smaller set of real training data. However, it is not well understood whether the pseudo dataset can benefit the classification model by providing more information or deteriorates the machine learning performance due to the prediction errors and uncertainties introduced by the surrogate model. This paper presents a preliminary investigation towards this research question. A classification-and-regressiontree model is employed to recognize the design subspaces to support design decision-making. It is implemented on the geometric design of a vehicle energy-absorbing structure based on finite element simulations. Based on a small set of real-world data obtained by simulations, a surrogate model based on Gaussian process regression is employed to generate pseudo datasets for training. The results showed that the tree-based method could help recognize feasible design domains efficiently. Furthermore, the additional information provided by the surrogate model enhances the accuracy of classification. One important conclusion is that the accuracy of the surrogate model determines the quality of the pseudo dataset and hence, the improvements in the machine learning model.

Ontologies As a Tool for Design and Material Engineers

Functionally Graded Materials (FGMs), initially conceptualized in the ’80, have recently attracted a great research interest thanks to the advent of additive manufacturing (AM) technologies. AM permits to gradationally varying the spatial composition or porosity inside an object resulting in a corresponding spatial change in material properties. The data about this new class of materials are radically different from the traditional engineering materials and require information about the object geometry. Moreover, traditional methods for product design are not sufficient to represent heterogeneous objects. The full exploitation of these technologies requires the synergy of material science, product modeling and manufacturing domain. Ontologies can play a crucial role for the integration, making the information accessible and understandable to both experts from different domains and machines. In this paper, a prototypical ontology for the characterization of FGM objects is proposed. Firstly, an already existing FGM ontology is analyzed, highlighting shortcomings and possible improvements. Then, the new ontology is proposed, focusing on the classes and relationships for accommodating material knowledge and geometrical information. The core idea, retrieved from the literature on heterogeneous object representation and transposed in an ontological fashion, is based on the mapping between the geometrical 3D space and the n-dimensional material space. After presenting the new ontology, a benchmark case study is described to test the effectiveness of this approach along with some competency questions an engineer might be interested in. The proposed ontology represents a first, crucial building block for a more complex system aiming to support the communication and knowledge sharing among different actors in engineering.

Algorithm-Based Support for the Redesign of Product Variants

Incremental Manufacturing is a novel manufacturing approach where product variants are manufactured based on a finalization of pre-produced parts through additive and subtractive manufacturing processes. This approach allows a multi-scale production with the possibility to scale product variants as well as the production volume. In order to ensure high economic efficiency of the manufacturing concept, there is a need for pre-produced parts that come as close as possible to the final variant geometries to ensure that only variant-specific features need to be added by additive or subtractive manufacturing steps. Furthermore, to ensure high economies of scale, a high degree of commonality should be ensured for the pre-produced parts manufactured in mass production. In this context, a graph-based method is developed that enables an automated analysis of product families, based on physical and functional attributes, for standardization potentials. The method thus provides support for the strategic definition of pre-produced parts and is embedded in an overall approach for the redesign of products for Incremental Manufacturing. For the demonstration of the approach, which is based on 3D Shape and Graph Matching methods, a first case study is carried out using a guiding bush product family as an example.

A Novel Training Bike and Camera System to Evaluate Pose of Cyclists

Aerodynamic drag force can account for up to 90% of the opposing force experienced by a cyclist. Therefore, aerodynamic testing and efficiency is a priority in cycling. An inexpensive method to optimize performance is required. In this study, we evaluate a novel indoor setup as a tool for aerodynamic pose training. The setup consists of a bike, indoor home trainer, camera, and wearable inertial motion sensors. A camera calculates frontal area of the cyclist and the trainer varies resistance to the cyclist by using this as an input. To guide a cyclist to assume an optimal pose, joint angles of the body are an objective metric. To track joint angles, two methods were evaluated: optical (RGB camera for the two-dimensional angles in sagittal plane of 6 joints), and inertial sensors (wearable sensors for three-dimensional angles of 13 joints). One (1) male amateur cyclist was instructed to recreate certain static and dynamic poses on the bike. The inertial sensors provide excellent results (absolute error = 0.28°) for knee joint. Based on linear regression analysis, frontal area can be best predicted (correlation > 0.4) by chest anterior/posterior tilt, pelvis left/right rotation, neck flexion/extension, chest left/right rotation, and chest left/right lateral tilt (p < 0.01).

Manufacture, Modeling, and Testing of Wire-Reinforced Metallic Pressure Vessel Prototypes

This article describes the manufacture, testing, and finite element modeling of prototype pressure vessels made of steel and reinforced with high-strength steel wire in the cylindrical part. Vessel prototypes were manufactured with pipe fittings and either no wire reinforcement, one layer of wire reinforcement, or two layers of wire reinforcement, with the purpose of developing an improved understanding of the effect of the wire reinforcement, and the number of reinforcement layers on prototype pressure strength. Pressure tests were conducted for instrumented vessels to determine strength up to 70 bar with a test system equipped with pressure and velocity regulators to guarantee the stability of the supplied flow and improve measurement accuracy and repeatability. Finite element modeling is conducted with the commercial code ANSYS and equivalent orthotropic properties obtained with the unit cell method, assuming a high value for the volume fraction of steel wire, and a matrix with low elastic properties compared with those of the steel wire. The results show that there is an interaction between the cylindrical part and the reinforcing wire, and that this relation is affected by external factors resulting from manufacturing process and material properties. Strain reduction in prototypes with thicker reinforcement is an indicator of the improvement on pressure resistance.

Use of a Trophic Structure Test Bed to Validate a New Systems-of-Systems Resilience Metric

Systems of Systems (SoS) combine complex systems such as financial, transportation, energy, and healthcare systems to provide greater functionality. A failure in a constituent system, however, can render the entire SoS ineffective by causing cascading faults. One method to prevent constituent faults from compromising SoS performance is to increase the SoS’s “resilience,” a measure of the SoS’s ability to cope with these faults and efficiently recover. Attempts to engineer improved resilience require a metric to measure resilience across different SoS architectures (network arrangements). In a previous work, the System of System Resilience Metric (SoSRM) was presented as a possible solution, but this new metric requires additional testing. This work examines the key question: “How can natural ecosystem characteristics be used to validate the SoSRM metric?” We hypothesize that the analysis of a test bed of generic ecosystems will produce SoSRM values that will positively correlate with a triangular trophic structure (wide base), validating SoSRM as a useful design metric. First principles for test bed creation are presented including biodiversity, trophic structure, and the role of detritus. SoSRM is measured for 31 case studies in a trophic structure test bed. Ecosystem network structure is quantified with graph theory. SoSRM correlates as expected with ecosystem network structure (r2 = .5016, n = 31), thus providing a validation of SoSRM as a design tool. As a final check, tests are conducted to ensure SoSRM is independent of trivial network characteristics (i.e. the number of nodes or links). By validating SoSRM, we provide a foundation for future work that focuses on increasing SoS resilience with biologically inspired design heuristics.

Text Writing Using Air Gestures for People With Limited Hand Motor Skills

Mobile devices and laptops are the main ICT tools to exchange information among people in the world. All the applications are designed by following a specific interaction style based either touchscreen or mouse and keyboard, which can be performed only with detailed movements of hands and fingers. Traditional interaction becomes difficult for elderly who have diseases limiting the hand motor skills, such as arthritis and brain stroke. The use of simple air gestures can be adopted as alternative interaction style to interact with smartphones, tablets and laptops. The aim of this research work is the development of an application that allows text writing using air gestures for people with limited hand motor skills. The application embeds several computer vision algorithms and convolutional neural networks software modules to detect and drawn alphanumeric characters and recognizing them using both mobile devices and laptops. The preliminary results obtained show that the approach is robust, and it can easily detect the alphanumeric characters written with the movement of the wrist.

Evaluation of Metal Lattice Structures With X-Ray Micro-Computed Tomography: Dimensional Accuracy and Manufacturability

Additive manufacturing enables product designers to incorporate complexity onto their designs on multiple size scales. Computer-aided design methods, such as topology optimization and lattice design, have emerged as software tools for applications where part consolidation and weight reduction are desired. Still, a more delicate control of hierarchical complexity and submillimeter-sized features would unlock a widely unexplored frontier of new design possibilities. However, the complexity of a design can respectively affect the manufacturing process. In powder bed fusion, the diameter, power and speed of the laser spot and the resulting size of the melt pool define the attainable feature resolution and accuracy in comparison with the original design intent. X-ray computed tomography can be a useful tool in validation and provide a detailed, volumetric representation of a part with internal features. This paper examines the design accuracy of 316L metal lattice structures and density of solid cubes with industrial X-ray micro-computed tomography. Accessible tools with open source software are presented for CT data analysis. The nominal values are compared against the as-built and CT scanned samples for surface area, volume, and dimensional accuracy. A CT voxel size of 30–40 μm allows to identify printability issues and general trends in the part density in comparison to the geometry changes. However, a finer voxel size in the submicron range would be required to properly detect and localize internal porosity and evaluate surface topography.

Powdered Metal Heat Treatment Apparatus Designed for a Laboratory

Metal additive manufacturing is a rapidly growing and sophisticated industry however the manufacturing processes and equipment for the heat treatment of the needed powdered metals is underdeveloped. Heat treatment is a key step in the powdered metal production process and is often needed to produce desired material properties. The objective of this paper is to examine the design of a heat treatment machine that addresses the needs of a laboratory performing research on powdered metals. The device was designed to address the three criteria of a heat treatment device; treatment, environment, and containment. The treatment criterion is accomplished by continuous powder flow through a furnace. The environment criterion is accomplished through a gas handling system capable of creating both an argon and vacuum environment. Finally, the containment criterion is accomplished through a network of tubes that provides structure to contain the powder. The design of this machine will allow research and development labs to heat treat powdered to a higher quality at a significantly faster rate.

An Investigation on the Design of Formate and Generate Face Milled Hypoid Gears

Hypoid gears are transmission elements that transfer power and moment between shafts whose axes do not intersect. They are similar in structure to spiral bevel gears. However, there are many advantages compared to spiral bevel gears in terms of load carrying capacity and rigidity. Hypoid gear pairs are mostly used as powertrain on the rear axles of cars and trucks. Hypoid gears are manufactured by two essential methods called face-milling and face-hobbing, and there are mainly two relative kinematic movements (Formate® and Generate). In this study, the gears produced with the Face-milling method are discussed. Face milled hypoid gears can be manufactured with both Formate® and Generate, while pinions can only be manufactured with the Generate method. The most crucial factor that determines the performance of hypoid gears is the geometry of hypoid gears. The gear and pinion geometry is directly dependent on the tool geometry, machine parameters, and relative motion between the cradle and the workpiece. The gear geometry determines the contact shape and pressure during power transmission. In this study, the mathematical equation of the cutting tool is set. After that, using differential geometry, coordinate transformation, and the gearing theory, the mathematical equation of hypoid gear is obtained.