Development of an automated compressor performance mapping using artificial neural network and multiple compressor technologies

Abstract In the last decades, several technological improvements to positive displacement compressors have been developed and introduced into market. During the process of implementing new compressor technologies, high-accuracy numerical models are essential to predict the performance at both component and system levels. ANSI/AHRI Standard 540 10-coefficient cubic polynomial model is still the industry-standard compressor mapping approach despite the well documented limitations. In order to generate accurate compressor maps by using the 10-coefficient cubic polynomial models, compressor manufacturers are required to obtain, in some cases, more than 20 compressor-calorimeter data points depending on the compressor type and operating envelope. This paper attempts to address the need for more generalized and versatile compressor mapping methodologies as well as to reduce the time-consuming and expensive compressor calorimeter testing. To this end, an automated compressor performance mapping approach based on artificial neural network (ANN) is proposed to identify the compressor operating envelope and map the performance of any positive displacement compressors for HVAC&R applications with minimum number of data points. In addition, the paper also aims at demonstrating the feasibility and reliability of the proposed automated compressor performance mapping approach for training the ANN models in comparison to conventional 10-coefficient cubic polynomial maps. Three different compressor types, i.e. reciprocating, scroll, and rotary rolling piston, have been considered as test cases.

[1]  O. Cortés,et al.  Optimization of operating conditions for compressor performance by means of neural network inverse , 2009 .

[2]  James E. Braun,et al.  Experimental and numerical analyses of a 5 kWe oil-free open-drive scroll expander for small-scale organic Rankine cycle (ORC) applications , 2018, Applied Energy.

[3]  Jimmy Ba,et al.  Adam: A Method for Stochastic Optimization , 2014, ICLR.

[4]  James E. Braun,et al.  Theoretical analysis of dynamic characteristics in linear compressors , 2020 .

[5]  Shengwei Wang,et al.  A comparison of the effect of empirical and physical modeling approaches to extrapolation capability of compressor models by uncertainty analysis : a case study with common semi-empirical compressor mass flow rate models , 2018 .

[6]  Sergio Ledesma,et al.  A comparison between the modeling of a reciprocating compressor using artificial neural network and physical model , 2015 .

[7]  James E. Braun,et al.  Performance limit for economized cycles with continuous refrigerant injection. , 2011 .

[8]  M Gholamrezaei,et al.  Compressor map generation using a feed-forward neural network and rig data , 2010 .

[9]  James E. Braun,et al.  PDSim: Demonstrating the capabilities of an open-source simulation framework for positive displacement compressors and expanders , 2020 .

[10]  Xianting Li,et al.  Utilization of ANN and ANFIS models to predict variable speed scroll compressor with vapor injection , 2017 .

[11]  Sergio Ledesma,et al.  Analysis and modeling of a variable speed reciprocating compressor using ANN , 2015 .

[12]  Vincent Lemort,et al.  PDSim: A general quasi-steady modeling approach for positive displacement compressors and expanders , 2020 .

[13]  Wenhua Li.  Simplified steady-state modeling for hermetic compressors with focus on extrapolation , 2012 .

[14]  Vincent Lemort,et al.  Liquid-flooded compression and expansion in scroll machines - Part I: Model development , 2012 .

[15]  Eckhard A. Groll,et al.  Comprehensive model of a hermetic reciprocating compressor , 2017 .

[16]  Eckhard A. Groll,et al.  Two-phase injected and vapor-injected compression: Experimental results and mapping correlation for a R-407C scroll compressor , 2018 .