ErrorSim: A tool for error propagation analysis of Simulink models

Abstract

This paper introduces a new lightweight tool for simulative error propagation analysis of Simulink models. The tool allows a user to inject different types of faults that are common for embedded control systems and analyze error propagation to critical system parts and outputs. The intended workflow comprises the following three steps: (i) setup faulty and critical blocks of a Simulink model, (ii) setup and run simulations, and (iii) observe and examine the obtained results. The tool is implemented in MATLAB using the callback block functions from the Simulink API. The graphical user interface allows the injection of several types of faults including computing hardware faults such as single and multiple bit-flips, sensor faults such as offsets, stuck-at faults, and a noise, and network faults such as time delays and packet drops. The fault occurrence and duration can be specified either with the classical reliability metrics like mean time to failure and mean time to repair, or failure rates with classical (normal, exponential, Poisson, Weibull etc.) or custom user-defined probability distributions. The error propagation to the selected critical blocks is reported with several statistical metrics including the mean number of errors, failure rate, and mean error value, as well as performance indexes such as integral squared error, integral absolute error, and integral time-weighted absolute error. The reported numerical results support standard reliability and safety assessment methods such as fault tree analysis and failure mode and effects analysis. The paper demonstrates the tool with a case study Simulink model of fault-tolerant control for a passenger jet.