978-1-4799-0059-6/13/$31.00 ©2014 IEEE. Root Cause Analysis on Changes in Chiller Performance Using Linear Regression Jun Okitsu R&D Department, Hitachi Asia Malaysia Sdn. Bhd., Kuala Lumpur, Malaysia jokitsu@has.hitachi.com.my Ken Naono R&D Department, Hitachi Asia Malaysia Sdn. Bhd., Kuala Lumpur, Malaysia knaono@has.hitachi.com.my Mohd Fatimie Irzaq Khamis Property Management & Maintenance, Universiti Teknologi PETRONAS, Tronoh, Perak, Malaysia fatimieirzaq@petronas.com.my Shaharin Anwar Sulaiman Department of Mechanical Engineering, Universiti Teknologi PETRONAS, Tronoh, Perak, Malaysia shaharin@petronas.com.my Mohd Amin Abd Majid Department of Mechanical Engineering, Universiti Teknologi PETRONAS, Tronoh, Perak, Malaysia mamin_amajid@petronas.com.my Abstract—Gas District Cooling (GDC) plants, designed to be environmentally efficient, require frequent maintenances, in order to avoid corrosions or leakages from the chemical reactions in Steam Absorption Chillers (SACs) of the plant. However, most of the plant experts face difficulty that the positive and the negative effects from the SAC maintenances are not clear. This is because there are various metrics to indicate GDC SAC performance, but they don’t have enough information to describe chiller internal conditions. The paper describes a method to detect the root cause of the GDC SAC performance changes. Specifically, (1) the chiller performance is modeled by linear regression on the performance related sensor data, and (2) the root cause is determined by time series analysis of the sensor contribution ratios to the performance in accordance of the concept of theory of constraints (TOC). Evaluations in Universiti Teknologi PETRONAS (UTP) GDC plant showed that the method determined the root cause correctly in 3 cases out of 4 problem cases. Because the method determines the root cause only from the plant operation historical data without any inspections, it is generalized to detect component failures and other plant anomalies. Keywords—chiller; performance degradation; maintenance; linear regression; root cause analysis; I. INTRODUCTION Gas District Cooling (GDC) provides electricity and chilled water to facilities with relatively low running cost and as a potential to reduce CO 2 emission [1]. Steam absorption chiller (SAC) is the key component of the GDC because SAC generates chilled water with low running cost by making use of waste heat and chemical reactions. The SAC has difficulty that corrosions and leakages in the system greatly affect its performance because of the chemical reactions. Thus, it is required for the SAC maintenance to detect and resolve the problems immediately. To realize it, plant experts were engaged to observe the SAC performance metrics, such as chilled water supply rate, COP and steam rate with sensors in the plant. However, it would be difficult for the metrics to detect the root cause of the problem immediately because the metrics do not have enough information to determine chiller internal conditions. Methods to detect root cause of plant abnormally were studied [2, 3]. As a consequence, detecting root causes of non- linear factors, such as control valves with excessive static friction, oscillating and sensors faults, can be solved satisfactorily [3]. However, detecting root cause of linear factors, such as performance deterioration due to aging and corrosion, still remains open research question [3]. Since the deterioration proceeds gradually, presence of noise and further factors makes it difficult to detect the root cause. The research objectives are (1) to develop a method to detect the root cause of a chiller performance deterioration caused by corrosions and leakages in the chiller, and (2) to evaluate the method through a root cause analysis on the SAC performance degradation in Universiti Teknologi PETRONAS (UTP) GDC plant.