An algorithm based fault tolerance technique for safety-critical applications

Abstract

h CONCLUSIONS The design of safety critical systems is a difficult and time consuming task. The traditional design of safety critical systems involves the use of proprietary hardware and software. The full custom design approach is becoming unacceptable because performing a complete redesign for each different class of control applications is unacceptable. Additionally, traditional safety critical systems are designed based on locating all faults in the system which can affect safety via diagnostics. The increasing complexity of the control algorithms which must be processed in a safety critical fashion makes the fault space of the system increase in size to the point where the design of diagnostics becomes quite difficult and time consuming. One promising methodology which can assist in reducing the design effort associated with safety critical systems while using Commercial Off The Shelf (COTS) components is Algorithm Based Fault Tolerance (ABET). Unfortunately, existing ABFT techniques are currently not suitable for safety critical applications. A new Safety Critical ABFT (SC-ABFT) technique is presented. The checking scheme for the SC-ABFT method is derived based on verifying the correctness of a given control application as it is being evaluated. Also, the probability of detecting a safety-critical error can made as close to 1.0 as desired by varying certain parameters.

Update on Common-Cause Failure Experience and Mitigation Practices

“…The V_Frame embodies the safety-critical algorithm-based fault tolerance (SC-ABFT) approach developed at UVa [115]. SC-ABFT provides a method for verifying whether applications are executed correctly within a certain probability.…”

“…Additionally, corresponding check-streams can be established to enable verification of correct execution in the field. This checking capability is implemented either in "a redundant processor executing software or a lowcomplexity custom hardware device" [115]. This type of system relies on having primarily discrete, as opposed to continuous, variables to allow the control system to be decomposed into a finite set of states.…”

“…The V_Frame implementation of the SC-ABFT was demonstrated in prototype form simulating the Los Angles Metro Green Line. The first demonstration at UVa involved a COTS-based test system using a Motorola 68040 processor card with supporting I/O implemented in a VME-based chassis [115]. Also, the check process was performed via a check algorithm that executed on a Motorola 68040 processor card.…”

Data criticality estimation in software applications

“…Different SIHFT techniques have been proposed to address different types of hardware error sources [6] [7]. The basic approach consists in improving the Built-In Error Detection Mechanisms of the system (exceptions, memory protection, etc.)…”