Hybrid Machine Learning-Based Fault-Tolerant Sensor Data Fusion and Anomaly Detection for Fire Risk Mitigation in IIoT Environment

. 2025 Mar 28;25(7):2146. doi: 10.3390/s25072146

Notations	Descriptions
${S_{1} D A, S_{2} D A, S_{3} D A, S_{4} D A,$ $S_{5} D A}$	Data acquision from RGB, IR, Gas, Smoke and Flame sensor
${S_{1} N_{i}, S_{2} N_{i}, S_{3} N_{i},$ $S_{4} N_{i}, S_{5} N_{i}}$ _.	Normalized the sensor raw values
${S_{1} M, S_{2} M, S_{3} M, S_{4} M, S_{5} M}$	Mean of the Normalized data
${S 1 S D, S 2 S D, S 2 S D, S 3 S D,$ $S 4 S D, S 5 S D}$	Standard Deviation of the normalized data
${D y n a m i c U p p e r T h r e s h o l d i,$ $D y n a m i c L o w e r T h r e s h o l d i}$	For Anomaly detection.
$S i_a n o m a l y$	record the anomaly detected for faulty sensor
${S_{1} A D, S_{2} A D, S_{3} A D,$ $S_{4} A D, S_{5} A D}$	Anomaly detection of each sensor
DE(m_i)	Deng Entropy
${M L_{1}, M L_{2}, M L_{3}, M L_{4}, M L_{5}}$	Class probabilities prediction from each of 5 ML Models
$C W F$	Combined weight factor
PWC	Percentage of weight (%) for calculating combined weight factor
${m (f i r e)}_{K}$ ${m (n o f i r e)}_{K}$	Probabilistic model adjustment based on sensor reliability
$T o t a l C o n f i d e n c e$	confidence score from all models (both normal and faulty sensors)
$m (f i r e)$ $(n o f i r e)$ m( $Θ$ +)	Belief mass for fire, Belief mass for No fire Belief mass for entire hypothesis
M_new(fire)	Dynamically Belief
H(m_i,m_j)	Hellinger Distance
${S_{1} D P, S_{2} D P, S_{3} D P,$ $S_{4} D P, S_{5} D P}$	Pre-processed sensor data
m_combined(A)	Dempster’s Rule of Combination
S₁RGB, S₂IR, S₃Gas, S₄Smoke, S₅Flame	Multi sensor data type
Belief(A), Plausibility(A)	Belief and Plausibility for each class