R. Fern'andez-Mart'inez

Combination of the Finite Element Method and Data Mining Techniques to Design and Optimize Bearings

Double-Row Tapered Roller Bearings are mechanical systems widely used in vehicles for the transmission of high load and moderate rotation speeds.

Design and Optimization of Welded Products Using Genetic Algorithms, Model Trees and the Finite Element Method

One of the fundamental requirements in the phases of design and manufacture of any welded product is the reduction of residual stresses and strains. These stresses and strains can cause substantial changes in the geometry of the finished products which often require subsequent machining in order to fit to the dimensions specified by the customer, and are usually caused by the contribution of an external heat flux in a small area. All welded joints contain welding seams with more or less regular geometry. This geometry gives the welded product the strength and quality required to support the mechanical demands of the design, and is affected by the parameters controlling the welding process (speed, voltage and current). Some researchers have developed mathematical models for predicting geometry based on the height, width and cord penetration, but is a difficult task as many of the parameters affecting the quality and geometry of the cord are unknown. As the welded product becomes more and more complex, residual stresses and strains are more difficult to obtain and predict as they depend greatly on the sequence followed to manufacture the product. Over several decades, the Finite Element Method (FEM) has been used as a tool for the design and optimization of mechanical components despite requiring validation with experimental data and high computational cost, and for this reason, the models based on FEM are currently not efficient. One of the potential methodologies used for adjusting the Finite Element models (FE models) is Genetic Algorithms (GA). Likewise, Data Mining techniques have the potential to provide more accurate and more efficient models than those obtained by FEM alone. One of the more common Data Mining techniques is Model Trees (MT). This paper shows the combination of FEM, GA and MT for the design and optimization of complex welded products.

Improvement in Manufacturing Welded Products through Multiple Response Surface Methodology and Data Mining Techniques

Gas Metal Arc Welding (GMAW) is an industrial process commonly used in manufacturing welded products. This manufacturing process is normally done by an industrial robot, which is controlled through the parameters of speed, current and voltage. These control parameters strongly influence the residual stress and the strength of the welded joint, as well as the total cost of manufacturing the welded components. Residual stress and tensile strength are commonly obtained via standardized hole-drilling and tensile tests which are very expensive to routinely carry out during the mass production of welded products. Over the past few decades, researchers have concentrated on improving the quality of manufacturing welded products using experimental analysis or trial-and-error results, but the cost of this methodology has proved unacceptable. Likewise, regression models based on Data Mining (DM) techniques have been used to improve various manufacturing processes, but usually require a relatively large amount of data in order to obtain acceptable results. By contrast, multiple response surface (MRS) methodology is a method for modelling and optimizing, which aims to identify the combination of input parameters that give the best output responses with a reduced number of data sets. In this paper, power consumption, cord area, tensile strength and tensile stress were modelled with quadratic regression (QR) models using Response Surface Methodology (RSM) and were compared with regression models based on DM (linear regression (LR), isotonic regression (IR), Gaussian processes (GP), artificial neural networks (ANN), support vector machines (SVM) and regression trees (RT)). The optimization of the parameters was conducted using RSM with quadratic regression and desirability functions, and was achieved when the residual stresses and power consumption were as low as possible, while strength and process speed were as high as possible.

Modeling of the Mechanical Properties of Carbon-Black Reinforced Rubber Blends by Machine Learning Techniques

Mastering the relationship between the final mechanical properties of carbon black reinforced rubber blends and their composition is a key advantage for an efficient design of the composition of the blend. In this work, some models to predict three relevant physical attributes of rubber blends — modulus at 100% deformation, Shore A hardness, and tensile strength — are built by machine learning methods and subsequently evaluated. Linear regression, artificial neural networks, support vector machine, and regression trees are used to generate the models. The number of used samples and the values for the input variables is determined by a Taguchi design of experiments, and prior to the modeling the uncertainty of the experimental data was analyzed.

Use of Classification Trees and Rule-Based Methods to Predict Shapes of Nano-Aggregates of Reinforcement Fillers

While manufacturing composite materials, reinforcement fillers inevitable collide with each other and subsequently they congregate to aggregates with different shapes. The shape of these nanoparticles aggregates are of great significance for the mechanical material properties and in consequence, knowing the percentage of aggregates of each shape within of a specific group of shapes can give an idea of the final properties of the material. This work classifies aggregates, a new dataset of 5713 carbon black aggregates gathered based on transmission electron microscopy image processing, using several classification trees and rule-based methods. Based on these methods several models are built, trained and tested to perform the classification. And then, the most reliable and accurate model to classify aggregates is selected, obtaining a testing accuracy of the 74.57% according to their shape.

Comparison Analysis of Regression Models Based on Experimental and FEM Simulation Datasets Used to Characterize Electrolytic Tinplate Materials

Currently, processes to characterize materials are mainly based on two methodologies: a good design of experiments and models based on finite element simulations. In this paper, in order to obtain advantages and disadvantages of both techniques, a prediction of mechanical properties of electrolytic tinplate is made from the data obtained in both methodologies. The predictions, and therefore, the comparative analysis are performed using various machine learning techniques: linear regression, artificial neural networks, support vector machines and regression trees. Data from both methodologies are used to develop models that subsequently are tested with their own method data and with data obtained from mechanical tests. The obtained results show that models based on design of experiments are more accurate, but the models based on finite element simulations better define the problem space.

Modeling Structural Elements Subjected to Buckling Using Data Mining and the Finite Element Method

Buckling of thin walled welded structures is one of the most common failure modes experienced by these structures in-service. The study of such buckling, to date, has been concentrated on experimental tests, empirical models and the use of numerical methods such as the Finite Element Method (FEM). Some researchers have combined the FEM with Artificial Neural Networks (ANN) to study both open and closed section structures but these studies have not considered imperfections such as holes, weld seams and residual stresses. In this paper, we have used a combination of FEM and ANN to obtain predictive models for the critical buckling load and lateral displacement of the center of the profile under compressive loading. The study was focused on ordinary Rectangular Hollow Sections (RHS) and on the influence of geometric imperfections while taking residual stresses into consideration.