Fault-Tolerant Fractional Control and Active Reconfiguration of Redundant Interleaved DC–DC Converters for UAV and Avionic DC Networks

In the sky, the difference between a safe flight and a catastrophe often comes down to the reliability of electricity. Modern aircraft and unmanned drones are increasingly "more electric," meaning that systems once powered by hydraulic fluid or compressed air—such as flight controls, pumps, and sensors—are now driven by electric motors and electronics. These systems rely on a steady flow of power from high-voltage sources to lower-voltage buses that feed critical components. If the power conversion hardware fails, the consequences can be immediate and severe. The challenge for engineers is not just to build strong components, but to create systems that can survive a failure. When a part breaks, the system must instantly recognize the problem, isolate the broken piece, and reorganize the remaining healthy parts to keep the power flowing, all while the voltage remains within a safe range. This requires a delicate balance between rapid detection, physical energy storage, and intelligent control.

A team of researchers has developed a new mathematical framework to solve this specific problem for aircraft power systems. They focused on a type of electrical converter that acts like a bridge, stepping down a high voltage of 270 volts to a regulated 28 volts needed by sensitive avionics. Their design uses four parallel paths, or "phases," to deliver this power. Under normal conditions, three of these paths work together to share the load, while the fourth sits idle as a spare. The researchers created a system that can mathematically predict what happens when one of the active paths suddenly stops working, or when a sensor that measures the current begins to give false readings. By combining a method to detect these faults with a specialized controller that uses "fractional-order" logic—a way of managing power that looks at the history of errors in a non-standard, more flexible way—they demonstrated how to keep the power stable even when multiple things go wrong at once.

The core of their work involves a three-step process that happens in the blink of an eye. First, the system watches the electrical current in each path. If a switch inside a path fails and stops conducting electricity, the system notices a mismatch between the command it sent and the actual current flowing. This mismatch acts as a clear signal that a hardware fault has occurred. Second, the system must distinguish between a broken switch and a broken sensor. If a sensor is lying and reporting a wrong number, the total current measured by all sensors will not match the actual flow of energy required by the load. The researchers designed a check that compares the sum of all sensor readings against the physical laws governing the circuit. If the sum is off, the system knows a sensor is biased; if the sum is correct but one path is dead, it knows a switch has failed. This distinction is crucial because the fix for a broken sensor is different from the fix for a broken switch.

Once the fault is identified, the system moves to the third step: reconfiguration. If a switch has failed, the system immediately cuts off the broken path and wakes up the spare fourth path to take its place. However, there is a critical window of time between the moment the fault happens and the moment the spare path is fully active. During this brief interval, the load is momentarily heavier than the remaining healthy paths can handle. The researchers calculated that the output capacitor, which acts as a small battery, must be large enough to hold the voltage up during this gap. They derived a specific limit for how long the system can take to switch over before the voltage drops too low to be safe. In their simulations, the system detected the fault in 0.08 milliseconds and engaged the spare path by 0.20 milliseconds, well within the safety limits they calculated.

After the spare path is engaged, the system must smooth out the voltage and return it to the perfect 28 volts. This is where the researchers introduced their fractional-order controller. Unlike standard controllers that react to errors in a rigid, step-by-step manner, this controller uses a more nuanced approach that weighs past errors differently. In their tests, this method proved superior. When they simulated a complex scenario where a switch failed, a sensor began to lie, the load suddenly increased, and the input voltage dropped, the fractional controller kept the voltage above 27.35 volts. A standard controller, by comparison, allowed the voltage to dip slightly lower and took longer to return to a stable state. Specifically, the new method reduced the total amount of voltage error over time by approximately 21.6 percent compared to the standard approach.

The researchers also tested how robust this system is when the physical components are not perfect. In the real world, inductors and resistors vary slightly from their ideal values. They ran a series of tests where they changed these component values by up to twenty percent and varied the load current. Even in these difficult conditions, the system kept the voltage above 27.28 volts, demonstrating that the design is not just a theoretical curiosity but a resilient architecture. It is important to note that these results come from a detailed computer simulation, not from a physical flight test. The authors are careful to state that while the mathematical guarantees are solid, real-world hardware would still need to be tested to account for noise, heat, and other unpredictable factors.

The significance of this work lies in its transparency and its unified approach. Instead of treating fault detection, reconfiguration, and control as separate problems, the researchers showed how they are deeply connected. They proved that no amount of clever software can compensate for a delay that is too long; the physical energy stored in the capacitor sets a hard limit on how fast the system must react. However, once the hardware is reconfigured, the choice of control strategy matters greatly. By using a controller that remembers the history of errors in a flexible way, the system recovers faster and more smoothly. This research provides a clear, mathematically explicit blueprint for building power systems that can survive component failures, a capability that is becoming increasingly vital as aircraft and drones rely more heavily on electricity to stay airborne. The study does not claim to have solved every problem for the aviation industry, but it offers a rigorous, reproducible method for designing systems that can gracefully degrade and survive when things go wrong.