Carlos A. Duchanoy

A novel recurrent neural network soft sensor via a differential evolution training algorithm for the tire contact patch

In this paper we propose a novel Recurrent Neural Network Soft Sensor designed to estimate and predict the contact area that tires of a car are making with the ground. This is one of the most critical issues regarding car modelling for improving its performance. The proposed sensor is particularly useful for an active suspension because it allows its suspension to be prepared instead of reacting to a disturbance. The recurrent neural network enables the Soft Sensor to have a correct prediction of the contact area of the tire. This sensor uses data from 11 sensors mounted on the car while the tire contact patch is obtained by means of frustrated total internal reflection phenomenon. The training process of the Recurrent Neuronal Network presents several difficulties caused by the existence of spurious valleys. For this reason, we address this problem as an optimization problem, solved by using a modified differential evolution algorithm. Our Soft Sensor performance is successfully validated by physical experiments under real operation.

Hardware implementation of the elitist compact Genetic Algorithm using Cellular Automata pseudo-random number generator ☆

Abstract In this paper the design and implementation of two versions of the compact Genetic Algorithm (cGA), with and without mutation and elitism, and a Cellular Automata-based pseudo-random number generator on a Field Programmable Gate Arrays (FPGAs) are accomplished. The design is made using a Hardware Description Language, called VHDL. Accordingly, the obtained results show that it is viable to have this searching algorithm in hardware to be used in real time applications.

Optimal redesign of a shock absorber using physical variables and the full dynamics of the suspension system

ABSTRACT In this paper a nonlinear car model, comprising the dynamic behavior of the suspension system, which includes the body displacement, body acceleration, wheel displacement, tire deformation, suspension travel, suspension geometry, pitch, and roll, has been designed. The main contribution introduced in this model is that it considers the nonlinearities caused by the geometry of the suspension system and includes a detailed tire model. This model is used by a dynamic multiobjective optimization methodology in order to improve the passenger comfort and the vehicle safety. As result of the optimization a set of nondominated solutions is presented.

Design of a Damper for a Suspension System via Evolutive Optimization

In this paper a dynamic optimization methodology for designing a passive automotive damper is proposed. The methodology proposes to state the design problem as a dynamic optimization one by considering the nonlinear dynamic interactions between the damper and the other elements of the suspension system, emphasizing geometry, dimensional and movement constraints. In order to obtain realistic simulations of the suspension, a link between a Computer-Aided Engineering Model (CAEM) and a multi-objective dynamic optimization algorithm is developed. As design objectives we consider the vehicle safety and the passenger comfort which are represented by the contact area of the tire and the vibrations of the cockpit respectively. The damper is optimized by stating a set of physical variables that determine the stiffness and damping coefficients as independent variables for the dynamic optimization problem, they include the spring helix diameter, the spring wire diameter, the oil physical characteristics and the bleed orifice diameters among others. The optimization algorithm that we use to solve the problem at hand is a multi-objective evolutive optimization algorithm. For this purpose we developed a parameterized model of the damper which is used to link the CAE tools and the optimization software, thus enabling fitness evaluations during the dynamic optimization process. By selecting the physical characteristics of the damper as design variables instead of the typical stiffness and damping coefficients, it is possible to consider important design constrains as the damper size, movement limitations and anchor points. As result of the proposed methodology a set of blueprints of non dominated Pareto configurations of the damper are provided to the decision maker.Copyright

A novel tire contact patch soft sensor via Neural Networks

In this paper, a novel application of a soft sensor via Neural Networks is proposed. One of the most critical issues in the car is the area where the tires are in contact with the ground, these are the only points of contact between the vehicle and the road. The importance lies in the fact that keeping the tire contact patch within the operational parameters ensures the safety, comfort and maneuverability of the vehicle. There are some techniques to measure the tire contact patch, but the operating conditions of the car make impossible the implementation; so in this work the tire contact patch is obtained by the frustrated total internal reflection phenomenon and with the measures of 12 sensors in the car, a Neural Network is trained and tested. Thus, a soft sensor that accurately estimates this variable is the main contribution of this work.

Nonlinear Full-Car Model for Optimal Dynamic Design of an Automotive Damper

In this paper a nonlinear full-car model, comprising the dynamic behavior of the suspension system, which includes the body displacement, body acceleration, wheel displacement, tire deformation, suspension travel, suspension geometry, pitch and roll has been designed. The main improvement introduced to this model is that it considers the nonlinearities caused by the geometry of the suspension system and includes a detailed tire model. This is used by a dynamic optimization methodology in order to improve the passenger comfort and the vehicle safety, which are represented by the chassis displacement and the contact area of the tires, respectively. The optimization algorithm used to solve the problem at hand is a multi-objective artificial bee colony algorithm (MOABC). As result of the optimization a set of nondominated solutions is presented.