No abstract

No abstract

in Wiley InterScience (www.interscience.wiley.com).When only single bypasses and utility duties are used as manipulations, optimal operation of heat exchanger networks (HENs) can be categorized as an active constraint control problem. This work suggests a simple split-range control scheme to implement the optimal operation. For a given HEN and information about the disturbances, the corresponding control structure can be found by solving an integer-linear programming (ILP) problem with two objective functions providing optimal split-range pairs (for tracking active constraints during the operation) and appropriate control pairings (for fast control action). A HEN case study is used to demonstrate the application of the proposed design technique. Dynamic simulation shows the ability to provide the optimal operation of the obtained control structure.

The fundamental problem in optimal sensor network design is choosing a set of important or strategic process variables to be measured. An optimization formulation for sensor network design that relates process economics and data reconciliation is proposed. To address this, an economic quantity is defined to quantify the loss of operational profit caused due to measurement uncertainty. The resulting analytical expression that quantifies the loss is shown to be the sum of weighted error variances of the reconciled estimates obtained from reconciliation. The final formulation is a mixed integer cone program that can be solved to obtain a globally optimal sensor network. The effect of the process economics, capital cost, and marginal utility of additional sensors is illustrated using case studies. ■ BACKGROUNDThe selection of measured variables is an indispensable task for effective control, monitoring, and safe operation of a chemical process. Several hundred variables exist in a typical chemical process. However, because of the nature of the process and the high cost of measuring instruments, only a subset of these variables can be measured.Typically, sensor network design procedures are conducted to address a specific process activity. From the viewpoint of fault detection and diagnosis, sensors are selected such that the faults are observable and resolvable. Several algorithms based on the digraph and signed digraph representation of the process model have been developed. 1,2 However, in contrast to these qualitative objectives, quantitative approaches do exist. Ali 3 introduced the concept of reliability, which uses the available sensor failure probability information to quantify the network performance. Later, Bhushan and Rengaswamy 4,5 formulated an optimization problem that minimizes the capital cost subject to maximal reliability and vice versa. In control applications, the problem is usually that of selecting a suitable candidate set of controlled variables. 6,7 Given the set of measurements, data reconciliation procedures adjust the measurement to improve the estimation accuracy. From a data reconciliation perspective, Mah 8 adopted estimation accuracy as a measure for choosing sensor networks and also showed that adding redundant measurements improved the estimation accuracy. For efficient process monitoring, a sensor network should be capable of providing precise information about the state of the process in the presence of random and gross errors. Specifications of precision, error detectability, and resilience can be enforced as constraints to obtain minimal cost networks. 9−11 Traditional cost-based approaches for the design of sensor networks are primarily based on the capital cost of the hardware element. Other cost factors such as maintenance cost for sensors have been discussed by Nguyen and Bagajewicz. 12 Also, operational profit-based studies have become increasingly important in recent years. Profit-based metrics are easily decipherable and also of direct use to the end user. From a control vie...

No abstract

No abstract

No abstract

No abstract