Analysis of different fouling predictive models in a heat exchanger from experimental data
Résumé de l’article :
We present a methodology that incorporates the implementation and validation of experimental data analysis for the creation of a predictive tool of heat exchanger fouling effects. The goal is to determine a realistic fouling kinetics in order to develop an adapted maintenance practice for minimizing energetic and intervention costs. The test bench implemented is equipped with plate and gaskets heat exchanger provided with industrial plates sizes. Particles are injected into the cold fluid to simulate a fouling. A metrology device is used on heat exchanger in order to control thermal and hydraulic performance through usual parameters. Several realistic conditions are tested during a learning phase to establish predictive models based on these experimental data. Results are analyzed using: (i) the asymptotic fouling models of Kern and Seaton [1] (ii) several predictive models from a statistical approach by different methods (multiple linear regression, artificial neural network). For each of them, conclusions are made about the accuracy of the modes and their application limits.
Author |
Christophe WEBER, Brice TREMEAC, Christophe MARVILLET and Cathy CASTELAIN |
Publication date |
2016 |
Keywords |
Fouling,heat exchanger, data analysis, predictive method, experimental data, multiple linear regression, artificial neural network. |
♦ The full version is only available for GRETh "PREMIUM" members!
♦ If you are not a GRETh member, you can consult the offer proposed by GRETh clicking here as well as the conditions of membership by clicking here.