Why fluid dynamics matters for display design in process control: commentary on Bennett and Malek.

Abstract

INTRODUCTION As its name implies, cognitive engineering requires attention to properties of both human cognition and engineering systems. Bennett and Malek (this issue) have done an exemplary job of investigating properties of human cognition that are pertinent to design of animated mimic displays for process control systems. In so doing, however, they may have inadvertently overlooked some potentially vital properties of real engineering systems (e.g., how fluid flows in a piping network under normal and, especially, abnormal situations). A closer look at fluid mechanical factors reveals that using animation to depict flow rate will sometimes provide operators with misleading feedback that could negatively affect plant safety. FLOW PATTERNS IN A PIPING NETWORK Implications for Animated Mimic Displays Bennett and Malek (this issue) motivated their research on animated mimic displays by need to support fault management behavior. In their discussion authors return to this issue, stating that the inclusion of animation in mimic displays could ... improve detection and diagnosis of faults (p. 448). However, two studies conducted evaluated performance on a quantitative psychophysical task, not a fault management task. It is not clear how results from an elemental task of quantitative judgments of velocity generalize to more complex relational task of fault management. Therefore, it is of interest to consider implications of using an animated mimic display in a complex piping network, typical of that found in real-world applications. Although usability of a design can only be assessed empirically, its usefulness can be evaluated analytically by identifying control requirements associated with a problem (Rouse, 1990). In this case these requirements can be examined by reviewing how fluid flows in a piping network. We initially consider simplest case (shown in Figure la) of fluid flowing through a pipe controlled by a valve (VA), resulting in a sensed flow rate (FA). Breakdown of Normal Expectations about Flow Rate Conservation of mass requires that instantaneous flow rate of an incompressible fluid in a rigid pipe be same at every location along a pipe segment. A pipe segment is defined as any continuous length of pipe uninterrupted by branch or feeder pipes. Clearly, flow rate will change across a branch or feeder point, as fluid will leave or enter pipe at such a location. Under normal operating conditions, it is therefore appropriate to represent flow in an entire pipe segment based on output of a single flow sensor. A display of type advocated by Bennett and Malek (this issue) would be effective in this case. However, consider more critical and demanding case of a pipe break or leak. Depending on location and severity of leak or break, flow rate along pipe could change drastically as a function of spatial location. For example, flow rate downstream of flow sensor location could be much less than that at flow sensor itself because of leaking fluid. However, an animated mimic display like that shown in Figure 1b would erroneously suggest that fluid is flowing at same flow rate all along pipe segment. It would do so because display extrapolates from flow datum collected at a single location to create a very compelling and attention-grabbing animated representation of what is normally true (i.e., constant flow rate all along pipe). This normal relationship is an inference because we do not have flow sensors all along pipe. During some faults, this inference is incorrect, and animated display may discourage operators from entertaining valid hypotheses about where pipe break or leak might be. The resulting situation is analogous to that observed in Three Mile Island (TMI) control room (Rubinstein, 1979). …