RflySim PHM - Fault Injection and Health Assessment Interface Library

---

The RflySim PHM (Prognostics and Health Management) library provides a comprehensive suite of tools for fault injection and health assessment, supporting modeling, injection, and evaluation of various hardware faults on UAVs, enabling validation of flight control system fault tolerance and fault-handling performance.

---

Module List

Power System Faults

Module	Function Description
Motor Fault	Motor fault injection, supporting fault modes such as reduced motor efficiency, stalling, and complete shutdown
Prop Fault	Propeller fault injection, supporting fault modes such as blade damage, detachment, and efficiency degradation
Battery Fault	Battery fault injection, supporting fault modes such as voltage drop, capacity decay, and increased internal resistance

Environmental Disturbance Faults

Module	Function Description
Wind Fault	Wind disturbance fault injection, supporting environmental disturbances such as gusts, turbulence, and wind shear
Load Fault	Load fault injection, supporting fault modes such as sudden load changes, center-of-gravity shift, and payload detachment

Sensor Faults

Module	Function Description
GPS Fault	GPS fault injection, supporting fault modes such as positioning drift, signal loss, and reduced accuracy
Sensor Fault	General-purpose sensor fault injection, supporting faults in IMU, magnetometer, barometer, and other sensors

Utility Modules

Module	Function Description
PHM 6DOF	Six-degree-of-freedom dynamics block connected to the PHM fault-model bus for state propagation under injected faults
Fault Params Extract	Fault parameter extraction module, used to extract and parse fault parameters from fault configuration files

---

Fault Type Descriptions

Power System Faults

Fault Type	Typical Manifestations	Application Scenarios
Motor efficiency degradation	Reduced rotational speed, slower response	Motor aging, thermal protection activation
Motor stalling	Complete loss of thrust	Mechanical failure, foreign object blockage
Propeller damage	Reduced thrust, increased vibration	Blade impact, fatigue cracks
Battery voltage drop	Insufficient power, premature landing	Low-temperature operation, deep discharge

Environmental Disturbance Faults

Fault Type	Typical Manifestations	Application Scenarios
Gust disturbance	Attitude oscillation, position deviation	Strong wind conditions, urban wind fields
Turbulence disturbance	Random vibration, control difficulty	Complex terrain, wake regions
Sudden load change	Center-of-gravity shift, reduced control margin	Payload release, suspended payload swing

Sensor Faults

Fault Type	Typical Manifestations	Application Scenarios
GPS drift	Increased positioning error, navigation failure	Signal obstruction, multipath effects
IMU drift	Attitude estimation bias, spin divergence	Temperature drift, vibration interference
Magnetometer interference	Yaw angle jumps, calibration failure	Electromagnetic interference, hard-iron effects

---

Application Scenarios

Fault-Tolerant Control Algorithm Validation

• Fault Detection: Validate the flight control system’s ability to detect various faults and its response speed |
• Control Reconfiguration: Test the system’s ability to reconfigure control and implement degraded control after fault occurrence |
• Safe Landing: Validate emergency landing strategies under severe fault conditions |

Sensor Fusion Algorithm Testing

• Fault Identification: Test the capability of multi-sensor fusion to identify single-sensor faults |
• Weight Adjustment: Validate adaptive sensor weight adjustment under fault conditions |
• State Estimation: Evaluate state estimation accuracy under fault conditions |

Flight Safety Assessment

• Fault Coverage: Assess the flight control system’s coverage capability across different fault modes |
• Fault Propagation: Analyze fault propagation paths and impact scope within the system |
• Safety Boundary: Determine safe flight boundaries under various fault conditions |

---

Usage Considerations

Fault Parameter Configuration

1. Fault Intensity: Fault intensity parameters must be set within a reasonable range; excessively large values may cause simulation divergence or crash |
2. Gradual Injection: It is recommended to inject faults gradually to avoid simulation instability caused by instantaneous abrupt changes |
3. Injection Timing: Select appropriate fault injection timing to avoid injecting severe faults during critical control phases (e.g., takeoff or landing) |
4. Parameter Consistency: Ensure fault parameters match the corresponding fault type to avoid incompatible parameter combinations |

Simulation Safety

1. Boundary Protection: Set reasonable flight boundaries; automatically terminate the simulation if the UAV’s state exceeds safe limits |
2. Fault Recovery: When designing fault recovery tests, ensure sufficient time and space for recovery |
3. Multiple Faults: Exercise caution when testing simultaneous multiple faults to avoid unpredictable interaction effects |
4. Logging: Enable detailed simulation logging for post-analysis and fault reproduction |

Interaction with Flight Control System

1. Message Frequency: The frequency of fault injection messages must match the flight control processing cycle to avoid message accumulation or loss |
2. Protocol Compatibility: Ensure fault parameter formats are consistent with those expected by the flight control system to prevent parsing errors |
3. Permission Management: Thoroughly validate fault injection logic in simulation before conducting real-flight tests |
4. Emergency Stop: Ensure a reliable emergency stop mechanism is available during real-flight tests to immediately take over control in case of fault loss of control |

---

Fault Modeling Principles

The PHM fault model divides fault injection locations into actuators, environment, payload, sensors, and state output links. The modeling goal is not simply to cause simulation anomalies, but to ensure faults are observable by the controller, diagnostic algorithms, and health assessment modules.

Motion Model Injection Relationships

Except for sensors, most faults ultimately enter the rigid-body motion model through total force, total torque, mass, inertia, or external disturbances.

\[ \begin{aligned} {}^{e}\dot{p} &= {}^{e}v \\ {}^{b}\dot{v} &= -[{}^{b}\omega]_{\times}{}^{b}v + gR_{be}^{T}e_3 + \frac{F}{m} \\ \dot{q} &= \frac{1}{2} \begin{bmatrix} 0 & -{}^{b}\omega^T \\ {}^{b}\omega & -[{}^{b}\omega]_{\times} \end{bmatrix}q \\ J{}^{b}\dot{\omega} &= -{}^{b}\omega \times (J{}^{b}\omega) + \tau + G_a \end{aligned} \]

Fault Category	Primary Influenced Quantities
Power System Faults	Total thrust F, three-axis torque tau
Environmental Faults	External force, external torque, relative wind speed and world-frame velocity
Load Faults	Mass m, inertia matrix J, additional torque
Sensor Faults	Scale factor, constant bias, impact, delay, random walk noise

Power System Faults

Fault Type	Injection Location	Typical Manifestations	Key Parameters
Battery Fault	Voltage, current, remaining capacity model	Voltage drop, insufficient power, reduced endurance	Internal resistance, capacity, SOC, discharge rate
ESC Fault	Between PWM and motor command	Single-channel output saturation, delay, failure	Gain, bias, dead zone, response time
Motor Fault	Motor speed and thrust model	Reduced thrust, stalling, slower response	Thrust coefficient, speed limit, time constant
Propeller Fault	Thrust/torque generation stage	Reduced propeller efficiency, imbalance, additional vibration	Efficiency coefficient, damage ratio, torque deviation

The power system can be modeled according to the signal chain of battery, ESC, motor, and propeller:

Sub-model	Relationship	Description
Battery	U_b = k_Ub * f_Ub(C_dch_effective, U_b0)	k_Ub is the battery fault coefficient, affecting output voltage
ESC	U_m = k_Um * sigma * U_b	sigma is the normalized throttle, k_Um is the ESC fault coefficient
Motor	omega = k_omega / (k_Tm * T_m * s + 1) * omega_ss	Describes response speed and execution efficiency faults
Propeller	T = k_CT * c_T * omega^2, M = c_M/T * T	k_CT changes thrust efficiency

Environmental and Load Faults

Fault Type	Modeling Method	Impact
Wind Fault	Add external force in geographic or body frame	Position drift, velocity error, increased attitude compensation
Wind Torque Fault	Add additional roll/pitch/yaw torque	Attitude oscillation, long-term control bias
Real-time Obstacle	Inject obstacle via UE or collision model	Path planning, obstacle avoidance, and collision detection
Mass Change	Modify mass parameter	Simulate payload release, load change, battery replacement
Moment of Inertia Change	Modify inertia matrix	Affects attitude response and angular acceleration
Center of Gravity Shift	Modify center of gravity or add additional torque	Produces sustained attitude deviation and control compensation

Load faults should simultaneously update gravity, inertia, and moment arm; otherwise, inconsistent dynamic results will occur.

The environmental wind field can be composed of atmospheric turbulence, constant wind, wind shear, and gusts:

\[ {}^{e}v_{wind} = {}^{e}v_{turb} + {}^{e}v_{cons} + {}^{e}v_{sheer} + {}^{e}v_{gust} \]

The drag and additional torque caused by wind disturbance can be written as:

\[ F_{wind} = -C_d v_{wind,b,relative}\lvert v_{wind,b,relative}\rvert \]

\[ \tau_{wind}=p_{wind}\times F_{wind} \]

Where p_wind is the position vector of the wind force application point in the body frame. Real-time obstacles are typically dynamically instantiated by the 3D engine at the fault trigger time and can be superimposed with the wind field to construct complex mission environments.

Load mass shift changes the inertia matrix and introduces additional torque:

\[ J = J_{health} + J_{body} \]

\[ \tau_{body}=p_{var}\times R_{be} \begin{bmatrix} 0\\0\\g \end{bmatrix} m_{var} \]

Sensor Faults

Sensor	Common Faults	Observed Manifestations
Gyroscope	Bias, increased noise, drift, stalling	Attitude estimation drift, abnormal angular velocity
Accelerometer	Bias, scale factor, impact anomaly	Abnormal attitude and velocity estimation
Magnetometer	Bias, interference, loss of lock	Yaw angle jump or long-term deviation
Barometer	Bias, random walk, step change	Incorrect altitude estimation
GPS	Loss of satellites, delay, noise, spoofing, position jump	Abnormal position and velocity estimation

Sensor faults are typically injected before SensorOutput or HIL output. After injection, ensure that data type, unit, and sampling frequency still meet the flight control input requirements.

Common sensor faults can be uniformly represented as the measured quantity after passing through scale factor, constant bias, drift, and noise from the true quantity:

Sensor	Simplified Model	Main Parameters
Gyroscope	omega_m = k_g * (omega + c_g) + b_g + n_g	Scale factor, constant bias, drift, white noise
Accelerometer	a_m = k_a * (a + lever_arm_terms - gravity_terms) + b_a + n_a	Scale factor, lever arm terms, drift, noise
Magnetometer	m_m = k_m * (R_be * m_e + c_m) + b_m + n_m	Geomagnetic vector, installation attitude, bias
Barometer	d_baro = -k_baro * (p_z + c_baro) + b_baro + n_baro	Altitude bias, scale factor, drift
GPS	p_gps = k_gps * (p + c_gps) + b_gps + n_gps	Position bias, scale factor, random walk

Fault Injection Protocol

The PHM module receives fault identifiers and fault parameters via inSILInts and inSILFloats. Each fault uses a unique FaultID to locate the type, and parameters are written into a floating-point array according to the fault module convention.

Variable Name	Type	Length	Description
inSILInts	int32[]	8	Fault identifier, used for type selection and channel localization
inSILFloats	float32[]	20	Fault parameters, used to set fault intensity, ratio, noise, or wind speed, etc.

FaultID	Fault Type	Fault Parameters
123450	Motor efficiency fault	Efficiency ratio for motors 1 to x, range 0-1
123451	Propeller fault	Efficiency ratio for propellers 1 to x, range 0-1
123452	Battery failure fault	None
123453	Low voltage fault	Voltage failure ratio, range 0-1
123454	Low battery fault	Battery failure ratio, range 0-1
123455	Load fault	Weight leakage ratio, range 0-1
123456	Load drift fault	Weight leakage ratio and x/y/z leakage factors, range 0-1
123457	Load leakage fault	Weight leakage ratio and leakage factor, range 0-1
123458	Constant wind fault	Wind speed on x/y/z axes
123459	Gust wind fault	Gust intensity and arrival time
123540	Turbulence wind fault	Turbulence intensity
123541	Shear wind fault	Shear wind intensity
123542	Accelerometer noise interference	Noise gain
123543	Gyroscope noise interference	Noise gain
123544	Magnetometer noise interference	Noise gain
123545	Barometer noise interference	Noise gain
123546	GPS fault	Noise gain, 3D mode, and number of satellites

Single fault injection example: 4 motors with efficiencies of 1, 0.8, 0.5, 1.

inSILInts   = [123450, 123450, 0, 0, 0, 0, 0, 0]
inSILFloats = [1, 0.8, 0.5, 1, 0, 0, ..., 0]

Multi-fault joint injection example: Inject motor efficiency fault and accelerometer noise fault simultaneously.

inSILInts   = [123450, 123450, 123544, 0, 0, 0, 0, 0]
inSILFloats = [1, 0.5, 0.2, 0.7, 5, 0, ..., 0]

Fault Parameter Management

Parameter	Description
faultType	Fault category or module number
startTime	Fault start time
duration	Fault duration
severity	Fault intensity or damage ratio
channel	Actuator or sensor channel
bias	Additive bias
scale	Multiplicative scale
noiseStd	Noise standard deviation

When creating a PHM dataset, it is recommended to record fault parameters, flight controller logs, ground truth states, control inputs, and sensor outputs simultaneously. Saving only flight controller logs is usually insufficient to replay the fault location and intensity.

Module Page Index

Topic	Page
Battery fault	Battery Fault
Motor fault	Motor Fault
Propeller fault	Prop Fault
Wind and environmental fault	Wind Fault
Load fault	Load Fault
GPS fault	GPS Fault
General sensor fault	Sensor Fault
PHM six-degree-of-freedom model	PHM 6DOF
Fault parameter extraction	Fault Params Extract

---

Related Resources

• PX4 Official Documentation - Fault Detection
• RflySim PHM Development Guide
• A Survey on UAV Fault-Tolerant Control
• IEEE PHM Standards

---

Note: This document serves as the index for the RflySim PHM library. For detailed usage instructions for each module, please refer to the corresponding standalone module documentation pages.