Statistical Deadband: A Novel Approach for Event-Based Data Reporting

Abstract

Deadband algorithms are implemented inside industrial gateways to reduce the volume of data sent across different networks. By tuning the deadband sampling resolution by a preset interval ∆, it is possible to estimate the balance between the traffic rates of networks connected by industrial SCADA gateways. This work describes the design and implementation of two original deadband algorithms based on statistical concepts derived by John Bollinger in his financial technical analysis. The statistical algorithms proposed do not require the setup of a preset interval-this is required by non-statistical algorithms. All algorithms were evaluated and compared by computing the effectiveness and fidelity over a public collection of random pseudo-periodic signals. The overall performance measured in the simulations showed better results, in terms of effectiveness and fidelity, for the statistical algorithms, while the measured computing resources were not as efficient as for the non-statistical deadband algorithms.The objective of this work is to implement statistical algorithms based on Bollinger's theory as alternatives to the conventional SoD sampling scheme, which is an event-based data-reporting strategy.The experimental results showed a performance enhancement of the statistical deadband algorithms over the non-statistical send-on-delta schema. In order to reproduce the algorithms comparison in R [10], the pseudo-code of all of the algorithms in this paper have a corresponding function in the deadband package [11]. The algorithms described in the pseudo-code represent the concept to obtain the algorithm output, sample by sample, as in the online working mode. The corresponding functions in the package process the collection of all outputs for all input samples in a finite time period. The functions in the package have been developed to compare the overall performance of a hypothetical device that embeds a hardware or software implementation of the corresponding pseudo-code algorithms.This paper is organized as follows. Section 2 presents an introduction to the OPC deadband concepts that are related to the data collection process and the algorithms. Section 3 presents Bollinger's technical analysis. Section 4 presents the two original deadband algorithms. Section 5 provides the simulation results developed by using the deadband package. Finally, Section 6 draws some conclusions.