What Is Fault Diagnosis And Fault-tolerant Control?

Fault diagnosis and fault-tolerant control is a field of systems and control engineering concerned with identifying faults in physical systems and maintaining safe, acceptable operation in their presence. Fault diagnosis determines that a fault has occurred, identifies its location and type, and estimates its severity. Fault-tolerant control uses that diagnostic information, or pre-designed robustness, to keep the controlled system stable and within performance bounds even when one or more components have degraded. Together, these two functions form the architecture underpinning resilient engineering systems in aerospace, industrial automation, nuclear power, and other safety-critical domains.

The field integrates estimation theory, control design, reliability engineering, and formal safety analysis methods. IEEE Access has published a dedicated special section on recent advances in fault diagnosis and fault-tolerant control of aerospace engineering systems, reflecting the scale of current research activity.

Fault Diagnosis Methods

Fault diagnosis proceeds through three steps: detection, isolation, and identification. Detection flags that an anomaly exists; isolation determines which component is responsible; identification estimates the fault magnitude or mode. Model-based methods generate residuals by comparing measured outputs to observer or Kalman-filter predictions and evaluate them using statistical decision rules. Parity space and parameter estimation approaches offer complementary coverage for actuator and sensor faults, respectively. Data-driven methods use pattern recognition, principal component analysis, or trained neural networks applied to process measurements, making them well-suited to complex plants where physics-based models are expensive to construct. A bibliographical review of fault diagnosis and fault-tolerant control design for aerospace systems surveys these methods across aircraft, satellite, and spacecraft applications, cataloging the signal types, observer architectures, and performance metrics reported in the literature. The diagnostic output, typically a fault flag with an estimated magnitude and confidence level, is passed to the fault-tolerant control layer.

Fault Tree Analysis

Fault tree analysis (FTA) is a top-down deductive technique used during design to identify combinations of component failures that can lead to an undesired system-level event, the top event. Starting from the top event, an analyst constructs a graphical tree using AND and OR logic gates representing failure combinations and independent failure modes. Each leaf of the tree corresponds to a basic failure event with an associated probability or frequency. Quantitative FTA computes the probability of the top event from the component probabilities using minimal cut set analysis, while qualitative FTA identifies the most critical single points of failure. The NASA Fault Tree Handbook is the foundational reference for applying FTA to aerospace programs and has shaped how the method is taught and applied across safety-critical industries. FTA informs both the diagnostic architecture, by revealing which faults are most consequential and therefore most important to detect quickly, and the fault-tolerant control design, by identifying which components require hardware redundancy or software accommodation logic.

Fault-tolerant Control Design

Fault-tolerant control systems are classified as passive or active. Passive designs use fixed controllers with sufficient gain and phase margins to maintain stability across all fault scenarios in the specified set, without any detection or reconfiguration logic. Active designs use the fault diagnosis output to trigger controller reconfiguration: switching to a backup control law, redistributing actuator commands through control allocation, or adapting controller parameters online. Active systems can achieve tighter nominal performance because they do not need to be conservatively detuned for all possible faults; they compensate only when a specific fault is diagnosed. The trade-off is that active systems depend on the reliability and speed of the diagnostic subsystem, making the FD-FTC integration a critical design and verification challenge for applications such as aircraft fly-by-wire systems and robotic manipulators.

Applications

Fault diagnosis and fault-tolerant control has applications in a wide range of engineering fields, including:

• Aerospace control, where flight management computers monitor actuator and sensor health and reconfigure control laws after failures
• Nuclear power plant instrumentation, where redundant sensor channels and voting logic detect and accommodate measurement faults
• Industrial process control, where valve and pump faults are isolated and compensated to maintain continuous production
• Automotive safety systems, where electronic stability control and steer-by-wire architectures must tolerate sensor and actuator failures
• Unmanned aerial and ground vehicles, where onboard fault management is required for safe autonomous operation