The so-called multilevel ant-colony algorithm, which is a relatively new metaheuristic search technique for solving optimisation problems, was applied and studied, and possible approaches to parallelisation of this algorithm were elaborated. The multilevel ant-colony algorithm performed very well and is superior to classical k-METIS and Chaco algorithms; it is even comparable with the combined evolutionary/multilevel scheme used in the JOSTLE evolutionary algorithm and returned solutions that are better than the currently available solutions in the Graph Partitioning Archive.
COBISS.SI-ID: 18247719
An approach of enhancing yield in the production of piezoresistive pressure sensors for automotive applications is proposed. The main idea is to introduce an advanced pre-check, which sorts out potentially bad sensors before the actual calibration. Reject margin is initially estimated on the basis of the production first series and continuously adjusted during the mass production in accordance with the quality management strategy. Yield enhancement of 5.7% in massive MAP sensor production in HYB, Šentjernej, was achieved by applying the described approach, resulting in the total 99.4% yield.
COBISS.SI-ID: 20958503
In this paper we present a case study of application of FE model in fault localisation in ceramic pressure sensor structures. The sensing elements are thick-film resistors acting as strain gauges, which translate the strain into an electrical signal. In the design phase, FE analysis was used to analyse the sensitivity of the thick-film resistors to the applied pressure. The same model was used for non-destructive fault diagnosis and troubleshooting of the prototype series. Selected examples illustrate the approach. FE model can also be used in the production test process.
COBISS.SI-ID: 20958759
Our achievements in optimization of electric-motor design are described. We optimized parameter values of rotor and stator geometries resulting in the minimum power losses. The solution based on a genetic algorithm was applied in practice in optimization a universal motor produced by Domel company for vacuum cleaners. Numerous solutions were elaborated considering different optimization criteria, such as material costs, power loss, mechanical design restrictions, etc. We achieved solutions in which the power losses in iron and copper were reduced by at least 20%.
COBISS.SI-ID: 18729767
We developed a neural network model for the prediction of apparent viscosity ofalumina–paraffin suspensions used in low-pressure injection moulding process. The model is based on a three-layer neural network with a backpropagation-learning algorithm. The training data were collected by the rotational viscometry followed by a nonlinear regression. The network is trained to predict the values of power-law model parameters suitable to describe non-Newtonian fluids. The approach helps to reduce the amount of experiments which simplifies the procedure in practice.
COBISS.SI-ID: 21387047