No abstract

No abstract

No abstract

Cyber-physical systems (CPS) are finding increasing application in many domains. CPS are composed of sensors, actuators, a central decision-making unit, and a network connecting all of these components. The design of CPS involves the selection of these hardware and software components, and this design process could be limited by a cost constraint. This study assumes that the central decision-making unit is a binary classifier, and casts the design problem as a feature selection problem for the binary classifier where each feature has an associated cost. Receiver operating characteristic (ROC) curves are a useful tool for comparing and selecting binary classifiers; however, ROC curves only consider the misclassification cost of the classifier and ignore other costs such as the cost of the features. The authors previously proposed a method called ROC Convex Hull with Cost (ROCCHC) that is used to select ROC optimal classifiers when cost is a factor. ROCCHC extends the widely used ROC Convex Hull (ROCCH) method by combining it with the Pareto analysis for cost optimization. This paper proposes using the ROCCHC analysis as the evaluation function for feature selection search methods without requiring an exhaustive search over the feature space. This analysis is performed on 6 real-world data sets, including a diagnostic cyber-physical system for hydraulic actuators. The ROCCHC analysis is demonstrated using sequential forward and backward search. The results are compared with the ROCCH selection method and a popular Pareto selection method that uses classification accuracy and feature cost.

Condition based maintenance with prognostics (CBM+) is an area of research that interests many in the industrial, energy, and defense sectors. Interest in this concept is focused on lowering overall cost of operations, while also increasing equipment availability and mission readiness. Many applications, however, include power constraints and extensive lifecycle requirements that pose a challenge for existing embedded sensing systems. In some cases these systems can be expected to operate for years to decades without access to wired electricity or reliable energy harvesting sources. In this study a battery powered sensor node is presented that collects operational (pressure, acceleration, position) and environmental (temperature) information to identify and track faults seeded into an instrumented hydraulic test stand. The experimental setup is described in this paper, along with the range of baseline, damage cases, and severities imposed upon the system. Machine learning algorithms are developed specifically to leverage features that can be processed at the sensor node, then applied using low-power, computationally-limited microcontrollers. Several classifiers are considered in this analysis, including random forest and classification trees. The results discussed include prediction accuracies, training and testing requirements, as well as physical power consumption measured using actual hardware. Findings indicate that small sized random forest algorithms (up to 5 trees) can be implemented at the node and provide lower error rates; however they operate with the higher computing times and power requirements when compared to other machine learning techniques. Conversely, classification trees provide a good trade-off in accuracy and computing time, prolonging the operational life of the sensor node given a finite capacity battery as the power source.