Neural Network Models For The Estimation Of Product Costs: An Application In The Automotive Industry

The estimation of the production cost per unit of a product during its design phase can be extremely difficult, especially if information about previous similar products is missing. On the other hand, most of the costs that will be sustained during the production activity are implicitly determined mainly in the design phase, depending on the choice of characteristics and performance of the new product. Hence, the earlier the information about costs becomes available, the better the trade-off between costs and product performances could be managed. These considerations have led to the development of different design rules and techniques, such as Design to Cost, which aims at helping designers and engineers understand the impact of their alternative decisions on the final cost of the developing product. Other approaches, which are based on information about already designed and industrialised products, aim at correlating the product cost with the product’s specific characteristics. The real challenging task is to determine such a correlation function that is generally quite difficult. The above observation led the authors believe that an Artificial Neural Network (ANN) could be the best tool to determine the correlation between a product’s cost and its characteristics. Several authors hold that an ANN could be seen as a universal regressor, able to approximate any kind of function within a desirable range, without the necessity to impose any kind of hypothesis a priori on the characteristics of the correlation function. Indeed, test results seem to confirm the validity of the neural network approach in this application field.