Abstract
Based on testability verification, this paper studies the design of a fault injection system applied during the design stage of an auto-driving instrument. Previous designs often prioritized functional requirements while ignoring testability verification, leading to significant equipment maintenance challenges. This paper proposes a fault injection system to address these issues. Experimental results demonstrate that the system can inject faults rapidly and effectively, providing quantifiable testability metrics.
Introduction
As electronic systems in military and aerospace sectors grow in complexity due to Large Scale Integration (LSI), maintenance becomes a critical bottleneck. Traditionally, "testability"—the ease with which a system's status is determined and faults are isolated—was treated as an afterthought. This "design first, test later" paradigm results in prolonged diagnostic times, high maintenance costs, and late-stage discovery of flaws that are expensive to fix. This study advocates for integrating testability verification early in the design cycle using a robust fault injection system.
Theoretical Framework: Fault Propagation
To verify testability, one must understand how a physical fault evolves into a system-level failure. The researchers utilize a propagation model where a Fault (the root cause) triggers an Error (an incorrect internal state). If the system's fault-tolerant mechanisms fail to detect or mitigate this error, it leads to a Failure.
By injecting faults at the start of this chain, designers can evaluate the effectiveness of Built-In Test (BIT) logic and fault-tolerant designs in a controlled, accelerated environment.
Methodology: Semi-Physical Simulation
The core of the proposed solution is a semi-physical simulation test environment. This setup bridges the gap between pure software models and final hardware prototypes by combining hardware-in-the-loop components with software-simulated faults.
The system architecture includes an equipment semi-physical test simulator that builds the test model and manages fault injection, connected to a Universal Automatic Test System (UTS) that executes diagnostic algorithms. The hardware utilizes FPGA-based control boards and RS422 communication to ensure high-speed data acquisition and precise fault timing.
Standardized Metrics and Results
To move away from subjective assessments, the study adopts metrics based on IEEE Std 1522-2004:
- Fault Detection Rate (FDR): The percentage of faults the system successfully notices.
- Fault Isolation Rate (FIR): The percentage of faults pinpointed to a specific replaceable unit.
- False Alarm Rate (FAR): The frequency of incorrect failure reports.
The system was evaluated using an Army autopilot system across 65 failure modes. The experimental results closely aligned with theoretical analysis, validating the system's accuracy:
| Metric | Analysis Results | Actual Results |
|---|---|---|
| FDR | 97.4% | 98.0% |
| FIR | 89.2% | 92.3% |
| FAR | <5% | 4.83% |
Conclusion
The developed fault injection system provides a reliable, quantitative method for verifying testability during the design phase. By identifying diagnostic weaknesses early, the system enhances equipment reliability and reduces long-term maintenance burdens. Future work will focus on increasing automation in the analysis phase and scaling the system for more complex electronic equipment.
Summary updated