Tìm kiếm nhà trọ - Ở ghép

    {xen:phrase loading}

Increase your Control System Performance

Thảo luận trong 'Labview' bắt đầu bởi bmnhy, 19 Tháng một 2007.

  1. bmnhy Giảng Viên

    Số bài viết: 914
    Đã được thích: 0
    Điểm thành tích: 0
    Increase your Control System Performance with More Accurate I/O

    If a control system is to be effective, it must first be capable of accurately representing the physical quantity that is to be controlled. Assuming the sensors and transducers are adequate, the next step is making sure the control system itself can effectively read the transducers. Regardless of the type of control (PID, fuzzy logic, model-free adaptive), a control system can only be as accurate as its inputs and outputs.

    Table of Contents
    Improvements without Compromise
    Analog-to-Digital Converter (ADC) Considerations
    Noise Considerations
    A Final Note about Accuracy
    Improvements without Compromise

    It is possible to improve controller performance simply through tuning, but doing so usually involves certain trade-offs. It is generally accepted that modifying gain parameters in a PID control system produces the effects shown in Table 1.

    Rise Time Settling Time Overshoot Steady State Error
    Increase Kp Decrease N/C Increase Decrease
    Increase Ki Decrease Increase Increase Decrease
    Increase Kd N/C Decrease Decrease N/C

    Table 1. Effects of gain parameters on a PID control system

    As you can see, an improvement in one area often means a compromise in another. For example, increasing the proportional gain will improve settling time, but will have an adverse effect on the amount of overshoot. If, on the other hand, we address the I/O capabilities of the system, it is possible to achieve performance improvements in all aspects, as Table 2 indicates.

    Rise Time Settling Time Overshoot Steady State Error
    ADC Resolution N/C Decrease N/C Decrease
    Signal Conditioning Decrease Decrease Decrease Decrease
    Calibration N/C Decrease N/C Decrease

    Table 2. Effects of improving various input parameters for a PID control system.

    Accurate inputs give the best possible representation of the physical system. As you can see from the table, the importance of accurate I/O in a control system cannot be overstated.
    Analog-to-Digital Converter (ADC) Considerations

    When a signal is converted from analog to digital there are a finite number of levels with which the signal can be represented, according to the resolution of the ADC. A higher resolution means that the device is capable of representing more discrete levels. Many black-box PID controllers use an ADC with 12 or even as few as 10 bits of resolution. With 10 bits there are 1024 discrete levels that can be used to represent the signal. If the input range is +/- 1 V, this translates to roughly 2 mV of code width. In the case of a simple PID controller, if the setpoint corresponds to an input voltage of 3 mV, the signal will be represented as either 2 or 4 mV. This difference results in an inaccuracy known as quantization error, which can cause the controller to continuously hunt for the setpoint, thereby increasing settling time, as seen in Figure 1.

    The maximum steady state error due to quantization is inversely related to the resolution of the ADC. An improvement in resolution of just 2 bits will result in up to a 4X decrease in steady state error. Likewise, an improvement in resolution of 4 bits will result in up to a 16X decrease in steady state error. This means that a PID controller with a 10-bit ADC could exhibit 256 times the steady state error of a PID controller with a 16-bit ADC. For the ultimate in accuracy, National Instruments offers M Series devices with 18 bits of resolution. This corresponds to more than 262,000 discrete levels with which to represent a signal.

    Figure 1. Inadequate ADC resolution leads to quantization error, which causes poor controller performance.
    Noise Considerations

    Noise is an inherent part of many systems, and can be the source of much frustration when trying to tune a controller. There are many control algorithms and tuning strategies that seek to minimize the effects of noise. Although this can be effective in certain situations, it is a very reactive approach to problem solving. This is similar to taking a pain reliever when you have the flu – it might make you feel better, but does nothing to treat the illness itself. A better, more proactive solution would be to employ signal conditioning to help reduce or eliminate the noise before it reaches the controller.

    Even a small amount of noise can present a significant problem for a control system. When working with low voltage transducers like thermocouples, just 30 uV of noise can result in an error of 5 degrees C. A common problem is 60 Hz noise from ordinary A/C power, which could easily cause unstable thermocouple readings. This is an ideal use case for a low-pass filter, such as the 2 Hz filter found on the SCXI-1102 thermocouple input module. This low-pass filter will remove all frequency components higher than 2 Hz, eliminating the unwanted noise from A/C power. For high-frequency noise, the high-accuracy M Series devices are equipped with an onboard low-pass filter that can be programmatically enabled or disabled based on the requirements of the system. Figure 2 shows the step response for a control system with a noisy input signal. The reponse with the filter enabled shows an improvement in all areas of performance (decreased rise time, less overshoot, faster settling time, and substantially less steady state error).

    Figure 2. A lowpass filter can help reduce the effects of high frequency noise and improve the response of a control system.

    Digital signals are susceptible to noise as well. Noise on a digital line can cause a low-logic signal to be interpreted as a high-logic signal. Let’s say, for example, that a quadrature encoder is being used in conjunction with a motor in a position control system. Inductive noise caused by the motor could be present on the encoder signal. This in turn could be perceived as a high pulse, causing the system to sense the position incorrectly. Digital filters, such as the debouncing filters found on M Series devices, present an excellent solution to this problem. Figure 3 shows how digital filters keep unwanted noise from being interpreted as a digital pulse.

    Figure 3. Debouncing filters help eliminate noise on a digital signal.

    Additionally, noise on an input signal can cause the controller to continuously hunt for the setpoint, thereby increasing settling time. Not only does this hurt the performance of the control system, but it can also cause physical damage to actuators. Consider a situation in which a stepper motor is changing its output very rapidly in either direction in an attempt to maintain a given setpoint and compensate for noise. This puts an unnecessary load on the motor that could induce premature failure.

    We have examined both systems that are noisy by nature and systems that are susceptible to noise from outside sources. If it is not addressed properly, noise can very easily be the limiting factor in a control system’s effectiveness. However, when appropriate signal conditioning measures are employed, noise will not pose a threat to controller performance.
    A Final Note about Accuracy

    It is important to recognize the benefits of accurate I/O hardware, as it serves as the link between the physical system and the control algorithm. You may spend hours tuning a controller in an attempt to correct an underlying problem that is a direct result of inadequate signal representation. All other things being equal, more accurate I/O hardware will always result in better control system performance.
    Related Links:

Chia sẻ trang này