Self-Training of a Fault-Free Model for Residential Air Conditioner Fault Detection and Diagnostics

Published

September 15, 2015

Author(s)

William V. Payne, Piotr A. Domanski, Jaehyeok Heo, Zhimin Du

Abstract

This study is meant to be applied to residential, unitary, air-source, single-speed, vapor compression air-conditioning systems with nominal cooling capacities less than 19 kW (65,000 Btu•h-1 or 5.4 tons). In this research, a method for self-training FDD and the new concepts of zone density and operational domain coverage rate are introduced. Various fault-free correlation techniques, such as ANN (artificial neural network), MPR (multivariable polynomial regression), and combined ANN-MPR, are demonstrated and compared for self-training FDD. For the validation of this method, a full FDD algorithm was developed to include eight scenarios (1- fault-free, 2-compressor valve leakage, 3-improper outdoor air flow, 4-improper indoor air flow, 5-liquid-line restriction, 6-refrigerant undercharge, 7-refrigerant overcharge, and 8- presence of non-condensable gases) using data from 72 fault-free and 84 faulty tests.

Citation

Technical Note (NIST TN) - NIST TN 1881

Report Number

NIST TN 1881

NIST Pub Series

Technical Note (NIST TN)

Pub Type

NIST Pubs

Download Paper

https://doi.org/10.6028/NIST.TN.1881

Keywords

Fault detection, fault diagnosis, heat pump, self-training

Citation

Payne, W. , Domanski, P. , Heo, J. and Du, Z. (2015), Self-Training of a Fault-Free Model for Residential Air Conditioner Fault Detection and Diagnostics, Technical Note (NIST TN), National Institute of Standards and Technology, Gaithersburg, MD, [online], https://doi.org/10.6028/NIST.TN.1881 (Accessed April 25, 2024)