faultDetect: Process detection

In chengfeifan/monitor: Write Process monitoring method

Description

After the model builds, use it to detect if a single multivariate observation is beyond normal operating conditions.

Usage

faultDetect(threshold_object, observation, ...)

Arguments

threshold_object	An object of classes "threshold" and "pca" returned by the internal threshold() function.
observation	A single row of an xts data matrix to compare against the thresholds
...	Lazy dots for additional internal arguments

Details

This function takes in a threshold object returned by the threshold() function and a single observation as a matrix or xts row. Internally, the function multiplies the observation by the projection matrix returned within the threshold object, calculates the SPE and T2 process monitoring statistics for that observation, and compares these statistics against their corresponding threshold values to determine if the observation lies outside the expected boundaries. The function then returns a row vector of the SPE test statistic, a logical indicator marking if this statistic is beyond the threshold, the Hotelling's T2 statistic, and an indicator if this statistic is beyond the threshold. Observations with monitoring statistics beyond the calculated threshold are marked with a 1, while observations within normal operating conditions are marked with a 0. These threshold values are passed from the threshold() function through this function via a returned threshold object. This object will be used in higher function calls.

Value

A named 1 x 4 matrix with the following entries for the single row observation passed to this function:

• SPE – the SPE statistic value

• SPE_Flag – the SPE fault indicator, where 1 represents a flag and 0 marks that the observation is within the normal operating conditions

• T2 – the T2 statistic value

• T2_Flag – the T2 fault indicator, defined the same as SPE_Flag

Examples

data(normal_data)
data(fault_data)
normal_embed <- embed(scale_pca(normal = normal_data, fault = normal_data),2)
fault_embed <- embed(scale_pca(normal=normal_data,fault=fault_data),2)
model_pca <- PCA_model(data=normal_embed, kernel_num=5)
threshold_pca <- threshold(pca_object=model_pca)
fault_detection <- faultDetect(threshold_object = threshold_pca, observation =  fault_embed)

chengfeifan/monitor documentation built on May 14, 2019, 2:29 p.m.