Anomaly Detection in Discrete Manufacturing Systems by Pattern Relation Table Approaches

. 2020 Oct 12;20(20):5766. doi: 10.3390/s20205766

Subroutine 4Pattern Table Building Subroutine (

L P, E, D

)

Input: Loop Pattern $L P$ , Event Set E, Training Data Set D

Output: Pattern Table T

Pattern set $P \leftarrow L P \cup E$ , Pattern table $T_{| P | \times | P |} \leftarrow \emptyset$

for sequence $d_{i}$ in D do

$S L$ = Sequence Splitting Routine( $d_{i}$ , $L P$ , D)

for ( $s l_{i}, s l_{i + 1}$ ) in $S L$ do

$T (s l_{i}, s l_{i + 1}) = *$ if $s l_{i}$ equals to $s l_{i + 1}$

$T (s l_{i}, s l_{i + 1}) = \to$ if $T (s l_{i}, s l_{i + 1})$ is empty or →, and $T (s l_{i + 1}, s l_{i})$ is empty

$T (s l_{i}, s l_{i + 1}), T (s l_{i + 1}, s l_{i}) = ‖$ if $T (s l_{i}, s l_{i + 1})$ is empty or →, and $T (s l_{i + 1}, s l_{i}) = \to$

end for

return T