Fault Detection Using the Clustering-kNN Rule for Gas Sensor Arrays

. 2016 Dec 6;16(12):2069. doi: 10.3390/s16122069

Algorithm 2 Fault Detection based on the Clustering-kNN rule

Input: M clusters, clusters centers

C P_{1}

C P_{2}

,...,

C P_{M}

, new test sample

x_{n e w}

, the threshold

D_{α}^{2}

of the

D^{2}

statistic
Output: Fault detection result
1: Calculate the distance between the new test sample

x_{n e w}

and each cluster center, denoted by

D (x_{n e w}, C P_{i})

i = 1, 2, . . ., M

;
2: Computer the nearest cluster center

C P

x_{n e w}

according to

C P = arg min \{D (x_{n e w}, C P_{i})\}

i = 1, 2, . . ., M

;
3: Adopt the cluster of

C P

as the calculation subset of the new test sample

x_{n e w}

;
4: Apply the kNN algorithm to the calculation subset of

x_{n e w}

, and achieve the value of the

D^{2}

statistic;
5: If the value of

D^{2}

statistic is larger than the threshold

D_{α}^{2}

, output the faulty work condition; otherwise, output the normal work condition.