With the development of power electronics and digital control technology, more and more control systems are adopting digital control methods. Currently, in CNC lathes, textile machinery, and vehicle drives, the adoption of fully digital control is an inevitable trend. Compared with analog control, digital control not only has advantages such as convenient control, stable performance, and low cost, but also opens up development space for networked and intelligent control of the system. A fully digital speed control system can not only conveniently realize motor control, but also achieve various additional functions through software programming, making the speed control system more user-friendly and intelligent, which is something analog control cannot achieve. Systems using fully digital control can achieve fault tolerance through software programming. References [1-2] proposed some software remediation strategies for faults caused by PWM inverters. This paper proposes a software fault-tolerant scheme for phase sequence connection faults existing in the application of brushless DC motors (BLDC motors, BLDCM). Normally, when the motor output phase sequence is mismatched, it is often necessary to manually determine the cause of the fault and match the phase sequence of the motor and controller. This requires human intervention and changes to the motor's cable connections. This paper proposes a fault-tolerant scheme for phase sequence faults. The faulty phase can be easily and quickly identified through software methods. At the same time, the control software adjusts the output phase sequence of the driver, thus possessing a self-recovery function for faults. Experiments show that the phase sequence fault can be accurately identified using this method, and the driver can start normally after the fault is cleared. The whole process does not require human intervention or any changes to the wiring. System Principle The BLDCM rotor uses permanent magnet excitation, which has high power density, simple control, and good speed regulation performance. Therefore, it is widely used in AC drives. Currently, a large number of documents have studied the control technology of BLDCM [3] and constant power field weakening [4-6]. The following is a brief introduction to the entire system and the control principle of BLDCM. Figure 1 is the main circuit diagram of the entire system. In this system, the BLDCM drive adopts a buck + full_bridge circuit structure. Unlike the conventional three-phase bridge drive method, the output current of the buck circuit, i.e. the current on the inductor, is controlled to enable the BLDCM to obtain DC current, thereby obtaining the best possible torque control effect. Figure 2 (a), (b), and (c) are the current waveforms of the inductor, capacitor, and motor bus terminal, respectively. [align=center] Figure 1 Main circuit of drive system[/align] [align=center] (a) Current waveform on inductor[/align] [align=center] (b) Current waveform on capacitor[/align] [align=center] (c) Current waveform of motor bus[/align] Figure 2 Current waveform of each component under constant current control[align=center] Figure 3 Switching transistor signal, three-phase back EMF and current waveform[/align] Figure 3 shows the back EMF, phase current and three-phase Hall signal status and switching transistor status of bldcm. As shown in Figure 3, the back EMF of bldcm is a pure trapezoidal wave. When current is passed through the maximum of each back EMF, a constant electromagnetic torque can be generated. Its expression is as follows: td＝（ea×ia＋eb×ib＋ec×ic）/ω (1) where td is the electromagnetic torque of the motor, ea, eb, ec are the back EMF of each phase, ia, ib, ic are the current of each phase, and ω is the angular velocity of the motor. Therefore, BLDCM control is relatively simple, requiring only commutation control based on Hall signals. Figure 3 shows each conducting switch and the current Hall signal state. Commutation control is performed by acquiring the Hall signal transitions, including rising and falling edges, and then according to a specific switching sequence. During startup, the BLDCM initializes the switching of the transistors by reading the Hall element states, thereby generating starting torque. However, when the motor output phase sequence differs from the controller's, the motor may fail to start or start abnormally. This will be analyzed below. Starting Torque Analysis In actual commutation control, the control signals for the switching transistors are output by reading the states of the three-phase U, V, and W Hall signals. For Hall elements installed at 120°, the effective states are 1–6, while for Hall elements installed at 60°, the effective states are 0, 1, 3, 4, 6, and 7. The following analysis uses the 120° installation case to examine starting faults caused by incorrect phase sequence connection. Table 1 lists six motor-driver connection configurations. For the six connection methods listed in Table 1, Mode 1 is the correct connection, while in Modes 2 to 4, only one phase is connected correctly, and the other two phases are connected in reverse. This is called the first type of fault. In the remaining two connection methods, all phases are connected incorrectly, which is called the second type of fault. [align=center]Table 1 Connection Methods of Motor and Driver[/align] Whether a permanent magnet synchronous motor driven by a sinusoidal current can start normally requires the following two conditions to be met simultaneously: i. At startup, the direction of the motor torque is consistent with the given direction of rotation; ii. After rotation, the angle between the stator composite current vector and the rotor shaft remains unchanged. When condition ii is met but condition i is not met, the motor will reverse; when condition ii is not met, the motor cannot rotate. Since the bldcm uses square wave current drive, the starting torque of the motor cannot be analyzed by coordinate transformation. Here, the average torque of each commutation region is used for analysis. Therefore, the conditions for normal motor startup mentioned above become: i. At startup, the direction of the average torque of the motor is consistent with the given direction of rotation; ii. After rotation, the direction of the average torque remains unchanged. When condition ii is met but condition i is not met, the motor will reverse; when condition ii is not met, the motor will not rotate. The following analysis addresses the two fault scenarios mentioned above. According to the BLDCM control method, at any given time (ignoring commutation time), only two phases are energized, and their back electromotive forces (EMFs) have opposite polarities. Therefore, if the energized phases are swapped, the generated torque will be reversed. For the first type of fault, from a temporal perspective, there are two cases: First, the phase with zero EMF is connected incorrectly, resulting in current flowing through it. For mode 2, in the 90°–150° range of Figure 3, since phases a and b are swapped, the average torque of phase b is zero. At this time, the average torque is determined by phase c, and its direction is consistent with the given direction of rotation. Second, when phase c experiences zero EMF, the currents in phases a and b are in opposite directions to the back EMF, resulting in an average torque opposite to the given direction of rotation, as shown in the 210°–270° range of Figure 3. Therefore, condition ii mentioned above cannot be met, and the motor cannot start. For the second type of fault, since all three phases are connected incorrectly, there is current in the phase with zero back EMF at any given time, and the average torque in that sector is zero. The average torque of the motor is determined by the other current phase. For mode 5, in the commutation interval of 30° to 90°, the back EMF of phase c is zero, and its average torque is zero. However, phase b receives a current opposite to its back EMF, thus generating a reverse average torque. Similar to the 30° to 90° interval, for any commutation interval, since current is received in the phase with zero back EMF, and its average torque is zero, the average torque generated by the other current phase due to the incorrect phase sequence must be a reverse torque, thus satisfying condition ii, and not satisfying condition i, the motor rotates in reverse. It should be noted that in the second type of fault, since the phase with zero back EMF does not generate average torque, the motor current will increase significantly under the same load. Fault Detection Principle Analysis: For permanent magnet motors, when the direction of the stator synthesized current vector does not coincide with the direction of the rotor magnetic field, electromagnetic torque must exist. This torque forces the motor to steer in the direction of the resultant current vector. Therefore, during motor rotation, the direction of the stator output current vector needs to be changed in real time to keep the electromagnetic torque constant. For BLDCM, when the output current vector of the three-phase full-bridge circuit remains constant, the motor rotor will rotate to that position. As shown in Figure 3, when transistors T3 and T6 are always on, the motor rotor will rotate to the midline of the opposite axis of the A-phase winding and the B-phase winding, and the state of the Hall element can be obtained from this. The actual connection sequence can be determined based on the different Hall signal states. Figure 4 shows the correspondence between the six output current vector directions and the Hall signal states. The inner numbers in Figure 4 represent the Hall signal states, and the outer numbers indicate the conduction states of the switching transistors. [align=center] Figure 4 Six current states of the stator and the states of the Hall element[/align] Here, the corresponding current vector is obtained by controlling the output voltage vector. When a phase sequence fault occurs, in this paper, switching transistors T3 and T6 are turned on, and the output current vector is located in sector 4. If there is no phase sequence fault, the Hall signal state is 4. Table 2 lists the Hall signal states corresponding to the six different connection methods. [align=center]Table 2 Hall State under Different Connection Methods[/align] Software Implementation Method The entire control system uses the F2407A DSP, which integrates a capture unit to capture the transitions of the three-phase Hall signals. Alternatively, the signal state can be obtained by reading the high and low levels, and the motor connection method can be determined by referring to Table 2. Secondly, once the motor connection method is determined, phase sequence faults can be eliminated without changing the electrical connections by adjusting the control sequence of the three-phase bridge, thus achieving fault tolerance in the control software. Figure 5 is the software flowchart for phase sequence fault detection and self-recovery. [align=center]Figure 5 Software Flowchart[/align] Experimental Results The entire system uses the TI F2407A DSP control chip, with a BLDCM rated power of 400W, rated speed of 450r/min, rated current of 12A, and rated voltage of 36V, powered by a battery pack. The switching transistor operates at a frequency of 50kHz. Inductor L1 = 650μH, capacitor C0 = 100μF. Figure 6 shows the phase current waveform obtained under connection mode 2. It can be seen that the current distortion is large, and the current amplitude is also large, which is consistent with the analysis above. Figure 7 shows the phase current waveform when correctly connected. Compared with Figure 6, its current waveform is more stable. [align=center] 5a/div, 5ms/div Figure 6 Phase current waveform under connection mode 2[/align] [align=center] 2a/div, 5ms/div Figure 7 Phase current waveform when the motor is correctly connected[/align] Conclusion This paper analyzes the starting fault caused by incorrect motor phase sequence by analyzing the average torque during BLDCM startup. With the help of three-phase Hall signals, by controlling the driver to output a certain voltage vector, the connection mode of the motor at fault can be accurately determined. Furthermore, the normal starting of the motor can be achieved through software adjustment without human intervention. Experimental results show that this method can accurately determine the cause of the fault and automatically eliminate the fault to complete the startup when judging the motor's starting fault. This method not only has the advantage of simple software control but also has certain practical value.