Table 1.

Survey of similar studies on operational states/modes detection in the literature.

Article	Overview	Method	Limitation
Liu et al. [26]	Diagnoses the 8 most common faults seen in photovoltaic arrays.	Gaussian kernel fuzzy C-means clustering algorithm	The number of different states is a priori known.
Tejedor et al. [27]	Detects and classifies threats and identifies activities.	Several techniques such as Gaussian mixture models	Classify new data into well-defined operational states.
Pan et al. [28]	Based on electrocardiogram signals, identifies a pilot’s fatigue status.	Learning vector quantization (LVQ) and support vector machines (SVM)
Wu et al. [29]	Identifies the operating state of converter transformer.	Deep belief network optimization algorithm
Simon and Litt [30]	Extracts steady-state engine operating points.	Mean and standard deviation calculations are combined with domain-specific logic constraints.	Only focusing on one mode of operation.
Davison and Craig [31]	Measures how close an engine is to steady state during operation.	Estimates the rate of change of the assessed parameter and variation about that change
Celis et al. [32]	Identifies the operating regimes of industrial gas turbines, and monitors and diagnoses steady state conditions during operation.	Moving window approach
Mikielewicz et al. [33]	Analyzes the partial load of gas turbines.	Based on theoretical analysis on micro turbines
Zhang et al. [34]	Determines the number of operational modes represented by clusters.	Multivariate statistical process control data (MSPC)	Determine only how many modes of operation there are.