Realistic Simulation of Sensor/Actuator Faults for a Dependability Evaluation of Demand-Controlled Ventilation and Heating Systems

Abstract

In the development of fault-tolerant systems, simulation is a common technique used to obtain insights into performance and dependability because it saves time and avoids the risks of testing the behavior of real-world systems in the presence of faults. Fault injection in a simulation offers a high controllability and observability, and thus is ideal for an early dependability analysis and fault-tolerance evaluation. Heating, ventilation, and air conditioning (HVAC) systems in critical infrastructures, such as airports and hospitals, are safety-relevant systems, which not only determine energy consumption, system efficiency, and occupancy comfort but also play an essential role in emergency scenarios (e.g., fires, biological hazards). Hence, fault injection serves as a practical and essential solution to assess dependability in different fault scenarios of HVAC systems. Hence, in this paper, we present a simulation-based fault injection framework with a combination of two techniques, simulator command and simulation code modification, which are applied to fault injector blocks as saboteurs and an automated fault injector algorithm to automatically activate fault cases with certain fault attributes. The proposed fault injection framework supports a comprehensive range of faults and various fault attributes, including fault persistence, fault type, fault location, fault duration, and fault interarrival time. This framework considers noise in a demand-controlled ventilation (DCV) and heating system as a type of HVAC system since it has been demonstrated that any fault injection scenario is accompanied by some impacts on energy consumption, occupancy comfort, and a fire risk. It also supports the reproducibility for a set of specific fault scenarios or random fault injection scenarios. The system model was implemented and simulated in Matlab/Simulink, and fault injector blocks were developed by Stateflow diagrams. An experimental evaluation serves as the assessment of the presented fault injection framework with a defined example of fault scenarios. The results of the evaluation show the correctness, system behavior, accuracy, and other parameters of the system, such as the heater energy consumption and heater duty cycle of the fault injection framework in the presence of different fault cases. In conclusion, the present paper provides a novel simulation-based fault injection framework, which combines simulator command techniques and simulation code modifications for a realistic and automatic fault injection with comprehensive coverage of various fault types and a consideration of noise and uncertainty, allowing for reproducibility of the results. The outputs achieved from the fault injection framework can be applied to fault-tolerant studies in other application domains.

Fault Injection with Multiple Fault Patterns for Experimental Evaluation of Demand-Controlled Ventilation and Heating Systems

“…Empirical models relate to more complex and detailed behavior and structural descriptions, such as simulation and physical models implemented as hardware and software features. Kiamanesh et al [ 8 ] discussed FI methods and presented a realistic, simulation-based FI framework for evaluating DCV and heating systems. Their introduced FI combines two techniques, simulator command, and simulation code modification, implemented with Stateflow for single faults.…”

“…In previous works, individual faults were injected into HVAC systems [ 8 , 12 , 13 , 46 , 47 ] and the injection of multiple faults was considered in other domains, such as semiconductor technology [ 10 , 27 , 30 , 31 , 34 , 35 , 37 , 39 ]. Hence, the injection of multiple faults with corresponding attributes is a research gap for DCV and heating systems.…”

“…White noise fault : A random number is added to the value, which is determined by Gaussian or uniform probability distributions [ 8 ].…”

“…Faults in HVAC systems can cause temperature fluctuations, occupancy discomfort, excess ventilation, and overheating. Fault management is a significant component of a building management system (BMS) for mitigating faults and their high-level symptoms [ 7 , 8 ]. For example, Teraoka et al [ 7 ] proposed a fault management framework, BDSherlock, based on two lists.…”

“…FI provides insights into the system’s behavior by deliberately introducing faults in different scenarios and conditions. FI frameworks for HVAC systems have been introduced in previous work [ 8 ], and the impact of individual faults on system behavior has been evaluated. However, many HVAC systems (e.g., hospitals, airports, multi-story office buildings) are large-scale distributed systems with thousands of components, including sensors, actuators, computational nodes, and communication links which are vulnerable and prone to faults.…”