|
Abstract:
Artificial
neural network (ANN) was utilized as a new method to calculate the
thermodynamic performance of refrigeration compressors. However, there
are some shortcomings in the previous study. A modified ANN method is
developed by taking into accounts physical features of the classical
thermodynamic model of refrigeration compressors. The multilayer forward
ANN in the new method is of less inputs than the old one, thus its
structure is compact. The inputs and outputs are naturally normalized.
Therefore, no operation on normalization or denomalization is needled.
Several kinds of small-scale refrigeration compressors are modeled with
the presented ANN method, satisfactory precision and generalization are
obtained.
Key words:
Refrigeration compressor Thermodynamic calculation
Artificial neural network
CLC No: TH45 TB652
国家教委留学回国人员基金、上海市青年科技启明星计划和上海交通大学科技发展基金资助项目. Received 20000627,
received in revised form 20000912
|