Ant Colony Inspired Machine Learning Algorithm for Identifying and Emulating Virtual Sensors

TL;DR

This paper proposes an end-to-end algorithmic solution, to realise virtual sensors in complex systems using an Ant Colony inspired technique, FAC2T, and splits the dataset into blocks and clusters each of them individually.

Abstract

The scale of systems employed in industrial environments demands a large number of sensors to facilitate meticulous monitoring and functioning. These requirements could potentially lead to inefficient system designs. The data coming from various sensors are often correlated due to the underlying relations in the system parameters that the sensors monitor. In theory, it should be possible to emulate the output of certain sensors based on other sensors. Tapping into such possibilities holds tremendous advantages in terms of reducing system design complexity. In order to identify the subset of sensors whose readings can be emulated, the sensors must be grouped into clusters. Complex systems generally have a large quantity of sensors that collect and store data over prolonged periods of time. This leads to the accumulation of massive amounts of data. In this paper we propose an end-to-end algorithmic solution, to realise virtual sensors in such systems. This algorithm splits the dataset into blocks and clusters each of them individually. It then fuses these clustering solutions to obtain a global solution using an Ant Colony inspired technique, FAC2T. Having grouped the sensors into clusters, we select representative sensors from each cluster. These sensors are retained in the system while the other sensors readings are emulated by applying supervised learning algorithms.

Figures (7)

• Figure 1: Proposed end-to-end algorithm. $M_0$ is the minimum number of required representative sensors obtained from PCA.
• Figure 2: The figure presents a visualization of the processes involved in one iteration performed by FAC2T, for one ant. We fix $\beta \ = \ 2$ and we have used a proxy pheromone matrix to impart the understanding. After the sampling and swapping process, the pheromone matrix scores are updated based on the evaluated metric. The algorithm, depicted above, is repeated for all ants in the colony(one in this case) to complete one iteration. Multiple iterations are performed until convergence is achieved.
• Figure 3: Plot of the Objective Function, $g$, evaluated across 100 synthetic datasets for 20, 25 and 30 clusters respectively from left to right. In each plot, a comparison is drawn between Ideal Clustering, FAC2T Clustering and K-Means Clustering.
• Figure 4: Metric maximization performance of FAC2T
• Figure 5: Plots of Actual Sensor Data and Predicted Sensor Data on the Test Data for Sensor 01 and Sensor 33 are shown in the first and second row respectively. The plots from left to right represent LBFR, ANN and SVR respectively