STRATEGIES FOR DESIGN AND SIMULATION OF HEAT EXCHANGER NETWORKS

Abstract

In a chemical industry, process streams require utilities to heat or cool them to required target temperatures. To conserve energy and utilities, the heat exchange is usually achieved among process streams directly through a Heat Exchanger Network (HEN). Therefore, the optimal design ofHEN is very important for overall process economy. There exist two types ofHEN problems; i) creating new designs (grassroots design) for the new plants, and ii) revamping existing designs (retrofit). The HEN problem, even in the simplified form, i.e., counter-current units, no phase change, one cost correlation and neglecting pressure-drop in the HEN, is too complicated to solve. In addition, many a times, the process demands HEN to be flexible enough to accommodate the fluctuations in process parameters. Thus, the final decision of selecting a HEN rests, in addition to the cost, on its flexibility. To understand the flexibility of HEN, its sensitivity to input parameter variations is required to be studied. At present, there are two approaches to solve HEN problems; i) the heuristics based evolutionary approach (Pinch Technology, PT) and ii) the algorithmic approach (Mathematical Programming, MP). Even though the HEN synthesis problem is inherently combinatorial, evolutionary PT has become popular in the industry. This work has been carried out with the aim to identify the deficiencies in current PT based approaches and to suggest new strategies and tools to overcome these. In the present work, new analysis tools, viz. The ER diagram, the Topology Table, the Topology equations, the expression and graphs for FT estimation accounting for temperature cross, and equations and plots for estimating number of shells in a network have been presented. Using these tools and on the basis of Global Minimum Number of Units (GMNU) and Minimum Number of Shells (MNS), a strategy for HEN design and screening of near-optimal HEN comprising multipass shell and tube exchangers have been proposed. Astep-wise retrofit strategy based on shut down, stream matching, by passing etc. has also been devised. For the performance evaluation of the HEN through sensitivity analysis, steady state simulation model for ideal and real exchangers having shell side leakage has also been presented. In order to implement the strategies, algorithms with their computer programs in Chave been developed. The Enthalpy Rectangle (ER) diagram has been found to be more useful to the existing HEN presentation scheme (grid diagram). The ER Diagram facilitates for the input parameters for determination of the number of shells in a HEN. The proposed Topology Table can be used in problem characterisation (viz. pinched, hybrid or threshold), in locating pinch swap and in setting Asymptotic Energy Targets. The Topology Table provides an understanding ofthe thermodynamics of HEN problem and are useful in designing an optimal HEN. In addition, it helps in deriving topology equations for utility targets, which are traditionally computed by repeated use of Problem Table Algorithm. The Topology equations are useful in setting absolute bounds on Heat Recovery Approach Temperature (HRAT), from "no network" case to HRAT=0. The optimum HEN always calls within their bounds ofHRAT and hence defines the solution space. Conventionally, optimal HEN is obtained by evolution starting from maximum energy recovery (MER) network. Since, many researchers have indicated that optimal network will be one of the GMNU networks, the present study recommends optimisation for GMNU networks only. This saves the designer from a lot of design effort and topology traps. It has also been established that in an exchanger, the variation in number-of-shells with temperature approach follows a similar trend as that of area with temperature approach. These results have been found to be applicable to HEN as well. The variation of sum total of shells in a HEN with HRAT follows the same trend as that ofarea target with HRAT. An analysis of 25 HEN problems proved that these results are general in nature. It has, therefore, been proposed that by replacing area with the total number-of-shells in HEN optimisation strategy, uncertainties associated with area targeting due to unavailability of exact value of heat transfer coefficients at design stage can be avoided. The shells based strategy does not rely on any of the simplifying assumptions involved in estimating area and capital cost. Hence, the strategy based on number-ofshells provides a fast and reliable tool to quickly screen near-optimal HENs for detailed analysis. Since the number of shells in a shell and lube exchanger depends on LMTD correction factor (Fj), new plots and correlations have been proposed in this work to facilitate fast and accurate determination of Ft even in those zones were conventional Ft determination methods fail to give accurate results. For computational purposes, linear Ft equations within ± 1 % accuracy have been developed for exchangers containing up to 5 shell passes. These equations directly account for temperature cross. After obtaining near-optimal HEN designs, the final selection is generally made after performance evaluation and sensitivity analysis. Therefore, multi-shell simulation models have been developed for ideal as well as real exchangers which account for baffles and shell-side leakage. The simulation models incorporates the effect of parameters, such as number of shell-side and tube passes, number of baffles, heat exchanger configuration etc. The above developed model has been used for sensitivity analysis. A simplified Artificial Neural Network(ANN) model was developed which spans the input parameter fluctuation encountered in real plant scenario, and accordingly provides the output parameters. The outlet temperatures predicted from the trained ANN has been compared with those obtained from simulation model. The results are within ± 0.5 °C. Such ANN model may find a use with HEN operations.