The introduction of WirelessHART transmitters enables new measurements to be made that previously could not be justified because of the high installation cost associated with installing a wired transmitter. Most of the initial installations of WirelessHART transmitters have been in monitoring applications. However, there are many applications where WirelessHART transmitters may be preferred, but there is a requirement that the measurement be used in closed-loop control. There are significant differences in the frequency and manner in which a new measurement value is updated by a wired transmitter vs. a wireless transmitter. Thus, it is natural to question what impact this has when a wireless transmitter is used in closed-loop control. In this article we examine the difference in a wired vs. a wireless installation and what changes in the PID are needed to address control applications using wireless devices.
Challenge – Control Using Wireless
Since most wireless transmitters are battery-powered, minimizing how often a measurement value is sensed and communicated to reduce transmitter power consumption is desirable. However, most multi-loop controllers used in DCS systems today are designed to over-sample the measurement by a factor of 2x to 10x to avoid the restrictions of synchronizing the measurement value with the control. Also, to minimize control variation, the typical rule of thumb is that feedback control should be executed 4x to 10x times faster than the process response time, which we will define as the process time constant plus process delay. Also, the conventional PID design used in DSC controllers assumes that a new measurement value is available each execution and that control is executed on a periodic basis. The measurement update and control execution that are typically assumed in a traditional control application using wired transmitters is illustrated below.
The conventional PID design (based on difference equation, z-transform) used in DSC controllers assumes that a new measurement value is available each execution and that control is executed on a periodic basis. When the measurement is not updated on a periodic basis, the calculated reset action may not be appropriate. If control is only executed when a new measurement is communicated, this could result in a delayed control response to setpoint changes and feed-forward action on measured disturbances that occur between measurement updates. Also, as the PID execution period is increased, the basic assumptions made in the PID design of the reset and derivative calculation may no longer be valid. For example, two common ways of implementing PI control are shown below.
In many cases, the reset contribution of the PID is realized using a positive feedback network (top implementation) in which the time constant of the filter in this network defines the reset time in seconds per repeat. This approach is often taken, since it supports the implementation of external reset for use in cascade and override applications. When the reset is implemented as an integrator (bottom approach), then logic is used to avoid reset windup. However, using either approach, as the period of execution becomes significant compared to the process response time, then the approximations in the reset implementation often break down and negatively impact control performance.
It may at first appear that there is no technical solution that minimizes how often a measurement is communicated without compromising control performance. In fact, both requirements can be met using a combination of the communication capability of WirelessHART devices and a modification in the way the PID is implemented. The key to understanding how the PID must be modified is to realize that when the PID reset is implemented using a positive-feedback network (top implementation above), the filter-time constant is a direct reflection of the process dynamic response. For example, when the Lambda PID tuning rules are used, then the reset (filter-time constant) is set equal to the process time constant plus the process dead time.
The standard communication techniques that are supported by a WirelessHART device are defined in the HART 7 specification that has been adopted as an international standard, IEC 62591Ed. 1.0. The device may be configured to communicating new measurement values using one of five defined burst message-triggered modes. For control applications, the two communication techniques that best fit control applications are
Continuous – The device wakes up at a configured update period, senses the measurement and then communicates the value.
Window – The device wakes up at a configured update period, senses the measurement and then communicates the measurement if the specified trigger value is exceeded.
Window communication is the preferred method of communications for control applications, since for the same update period window communications will always require less power that continuous communications. When window communication is selected, a new value will be communicated only if:
- the magnitude of the difference between the new measurement value and the last communicated measurement value is greater that a specified trigger value;
- or if the time since the last communication exceeds a maximum update period.
Thus, the measurement is communicated only as often as required to allow control action to correct for unmeasured disturbances or response to setpoint changes. For windowed mode you must specify an update period, a maximum update period and a trigger value. The update period and the maximum update period are separate configured values. For best control performance process disturbances, the update period configured for window communications should be one-fourth the process response. If disturbance response is not critical, then a slower update period may be used.