01 October 2002
Take a Lickin', Keep on Tickin'!
Raymond Sepe Jr., Chris Morrison: Electro Standards Laboratories
Fault-resilient electric motors in critical path applications such as automotive systems require designing for continuous operation and performance.
Electric motors in automotive systems for ancillary operations and steering systems, or for power generation and traction in hybrid electric vehicles, promise to improve vehicle performance and efficiency. This improved performance is often obtained using modern control techniques that depend on the control system's sensor inputs operating properly. For example, high-performance induction motor controllers often employ an optical rotor position encoder.
When sensors fail, the control system needs to compensate for that failure to continue operating. Techniques used to develop fault-resilient controllers for induction motors can also be extended to other motor technologies.
Control techniques include the following:
- Indirect field-oriented current controllers (IFOCs) are commonly used for high-performance induction motor drive systems over the entire speed range. They typically incorporate rotor position feedback and motor current feedback.
- Sensorless vector controllers eliminate the need for position feedback but operate poorly at very low speeds.
- Scalar control methods have been used when rotor flux orientation cannot be maintained. However, they do not allow for decoupled torque and flux control.
- Volts/Hertz open loop induction motor control can be used without feedback sensors. However, results include slow dynamics and sluggish response. Furthermore, operating volts/hertz is characterized by large transient behavior in motor currents and torque.
Selecting the control method that best fits the particular application is a difficult design decision in motor controller development.
Rather than developing a controller based on a single control technique, the system Electro Standards Laboratories developed adaptively changes control techniques in the event of sensor loss or sensor recovery. This reorganizing controller, comprising failure detection and fallback strategy, adopts the best control methodology according to available feedback and operational hardware.
While the first part monitors system components status, such as sensors, motor, inverter, etc., the latter engages the most appropriate control strategy based on a hierarchical basis.
Figure 1: Control hierarchy and sensor requirements
Figure 1 defines a control hierarchy with sensors each control method requires. A vector represents each part's operational status. The constituents of the status vector are updated via a failure detection system. A healthy operation of a specific component corresponds to 1, 0 shows a faulty operation, and X indicates a particular sensor is not needed for operation.
In the present report, fault-tolerant performance in the presence of failures in measurement and instrumentation devices is addressed. Accordingly, the following status vector is defined:
Note that S4 might represent physical voltage sensor(s) or a priori knowledge of a constant inverter bus voltage. Depending on the functionality of the sensors, the most appropriate control strategy will be employed. All available sensor information to conduct motor control is used.
Transition between control strategies may depend on a number of factors, including availability of needed sensors. But the transition strategy includes an assessment of performance as well as operating stability.
Figure 2: Logic structure illustrating sensor controller requirements
Figure 2 shows a logic structure indicating which sensors are needed for each type of control.
Performance quality in induction motor drives depends on their control strategy.
Indirect vector control method
This offers the best transient and steady state performance over the entire speed range. The main principle behind successfully implementing this technique is to match the stator excitation with the rotor flux angle. This method relies on the availability of position and current measurements.
For high-performance applications, sophisticated adaptive and identification techniques may need to be integrated into the control algorithm to compensate for the nonlinear and time-varying nature of the induction motor drives. These effects are mainly due to dependency of the rotor time constant on temperature rise and the mutual inductance dependence on saturation effects. If the position sensor becomes unusable, the indirect vector current controller will not function properly. This has resulted in development of a range of position sensorless techniques.
Position sensorless techniques
Estimating rotor flux components is the main building block in most position sensorless techniques developed for induction motor drives. In fact, normalized rotor flux components are directly used as trigonometric functions of rotor flux angle. This assumption largely depends on the accuracy of the flux estimation scheme.
Flux estimation involves calculations that depend on knowledge of voltages being applied to motor and numerical calculations susceptible to offset errors and numerical stability issues. Motor voltage can be measured with voltage sensors or calculated based on knowledge of voltage commands and voltage on the DC bus. The existing trade-off between mitigating measurement noise, limited sampling time, and speed prohibits an accurate performance over the entire speed range.
Indeed, the induction motor drive's performance will be substantially deteriorated outside the targeted speed range. This, along with the need for external voltage sensors in the event of imperfection in inverter operation, has called for yet another alternative that can substitute the sensorless technique-namely, scalar control of current magnitude.
Scalar control of current magnitude
Having lost the access to rotor flux angle, an emulated rotor flux position can be used to control the induction motor drive. This internal angle generator maintains the electrical velocity of the rotor flux and continues to control the magnitude of stator current.
As a direct consequence of scalar control, a load-dependent torque production, along with a deteriorated dynamic performance, is expected. Despite a reduction in performance quality, the motor can still operate and deliver torque. This controller can be used with an incipient pullout detection method to enhance its effectiveness across a wide range of operation. Implementing current magnitude scalar control requires properly operating the current sensors. In the event of current sensing failure, scalar control in terms of volts/hertz can be engaged.
This is considered the last step in the hierarchical classification of control strategies for induction motor drives. Although a significant deterioration of transient performance and a loss in torque/ ampere is expected, this method depicts one of the most direct methods for operating a voltage source inverter. Although voltage sensors are typically not used, the application of a desired voltage to the motor without the need for voltage sensors implies the inverter bus voltage is known.
Figure 3: Fault-resilient control using layered approach to controller architecture
The concept of fault-resilient control lends itself to a layered architecture. The highest performance control that requires sensor feedback is at the innermost layer, and the lowest performance control least dependent on sensor feedback is at the outmost layer (Figure 3).
A major source of sensor failure in motor control systems is loss of the rotor position sensor.
The encoder-based controller performs well over the entire speed range, as long as the encoder is functioning properly. Although the sensorless controller does not require rotor position feedback, it has limitations, including inaccuracy at lower speeds.
The encoder-based controller and the sensorless controller differ in their sensitivity to variations in the bus voltage. The encoder-based system does not deteriorate when the bus voltage varies, even when it is not known. The sensorless controller develops large torque errors, assuming no measurement of the bus voltage is available.
Encoder Failure Detection
Statistical characteristics of position data are used to detect encoder malfunctions. A drastic change in the average speed over a short period of time indicates a mechanical or electronic encoder system breakdown. However, in some situations, mechanical slipping of the encoder or cable problems can also cause inaccurate position information that will finally contribute to field-oriented control malfunction. This type of anomaly will not cause an immediate change in the average speed but will result in a larger calculated variation in speed.
When operating the system under encoder-based control, the sensorless flux angle estimator is active. The flux angle generator runs in parallel and integrates motor voltages to determine the machine flux. This calculated machine flux is not used by the controller but is available in the event a switchover to sensorless control is required.
Insofar as the estimator achieves the magnitude and phase characteristic of a perfect integrator, and assuming knowledge of the machine parameters, the correct flux angle is generated. In this case, a switch from encoder-based control to sensorless control is a smooth transition that does not produce any torque transients.
Once running under closed loop sensorless control, the encoder-based angle generator continues to run in parallel in the event a switchover is desired. The angles drift in time with respect to each other. They slowly move in and out of phase. Therefore, a command to switch from the sensorless controller back to the encoder-based controller is likely to result in a large angular discontinuity as the system switches from the sensorless angle curve to the encoder-based angle curve. This discontinuity, when fed into the indirect field-oriented current controller, results in a sharp torque transient.
Figure 4: Plot of torque transient when switching from sensorless control to encoder-based control
The torque transient at controller switchover was observed in the laboratory and in simulation. Figure 4 shows a plot of a simulation experiment that illustrates the torque transient.
At t=1.5 sec, a switching command from sensorless to encoder-based control is issued. The control system engages the encoder-based system and immediately places its angle generator into the closed loop system. Due to its error tracking the actual rotor flux angle, an immediate increase in the torque occurs. This results in an unwanted impact to the mechanical system.
A smooth transition technique from sensorless to encoder-based vector control was designed and implemented in the control algorithm. The idea is to force synchronization between the encoder-based angle generator and the sensorless controller's angle at the instant of controller switchover. Once engaged within the closed loop system, the encoder-based angle generator then properly maintains synchronization with the actual rotor flux angle.
In our lab, experiments were designed and performed on an 8-kilowatt induction motor starter/ alternator for hybrid electric vehicles. The lab's modular motor test platform incorporating a TMS320C32 floating-point digital signal processor was used to implement the control algorithm.
In implementing such systems, transients can be generated at the instant of controller switchover. With proper synchronization, the smooth transition between encoder-based IFOC and sensorless IFOC has been achieved.MC
Raymond Sepe, Ph.D., vice president of R&D services at Electro Standards Laboratories, has more than 10 years' experience developing electric motor systems and power electronics. Contact him at rsepe@ElectroStandards.com; www.ElectroStandards.com. Chris Morrison, a design engineer at Electro Standards Laboratories, has over 15 years' experience developing data communications products and electric motor systems.