(330c) Path Flow Indicators As Optimization Drivers In a Systematic Retrofit Methodology for Chemical Batch Processes

In a competitive globalized economy retrofit design of chemical plants is typically motivated by economic incentives to meet dynamic market requirements in cost-effective ways subjected to environmental and societal constraints. Especially for batch processes producing low volume-high value chemicals retrofitting is of high importance due to inherent process layout flexibility, uncertainty in product demand and shorter time-to-market periods leading to limited time for process development and operation under non-optimal conditions. It is therefore highly important to develop systematic methodologies, conceptually robust yet easily applicable in industrial practice, in order to enhance the often empirically based retrofitting. In this perspective, a good starting point can be systematic methods originally developed for continuous process design (i.e., heat and mass integration, superstructure optimization, flowsheet decomposition etc.) with proper adjustments for the intrinsic special features of chemical batch processes.

The present study applies a source-sink oriented path flow decomposition method, originally developed for continuous processes, to the field of batch process retrofitting. In particular, it is intended to enhance the concept of path flow indicators targeting at process improvement potential by revealing the extent to and the conditions under which they can serve as multi-objective optimization drivers. The study also focuses on the boundary conditions of the batch process retrofitting problem formulation, and in particular on the batch-to-batch variability effect for base case definition and interpretation of the retrofitting results.

The systematic retrofit methodology consists of three stages summarized below. In the first stage, the base case scenario is defined. First, process boundaries, objectives and path flow indicators are selected. Then, the real process performance is investigated based on historical process data. After this evaluation, it has to be decided whether process performance is sufficiently described by an average batch or whether batch-to-batch variability has to be included. For instance, batch-to-batch variability can be represented as a set of reference batches that describe the base case scenario. In a next step, process modeling is conducted and mass and energy balances of the base case(s) have to be derived. At the end of this stage, the objective functions of the base case(s) are calculated, which serve as reference in the next stages.

The second stage comprises a procedure that systematically generates promising retrofit alternatives for the respective base case(s). The procedure decomposes the whole flowsheet of the base case into component path flows, for which various path flow indicators are calculated that quantitatively or qualitatively describe the process performance. In this work, a new path flow indicator, called “renting cost indicator”, is introduced that enlarges the scope of batch process retrofit alternatives by considering the occupancy times of the component path flows in the process units. All path flow indicators are linked to heuristics which in combination with sensitivity analysis result in parametric and/or structural retrofit alternatives of the base case(s) targeting at improved process performance with respect to these indicators.

In the last stage of the methodology, the retrofit alternatives are classified and prioritized. The objective is not restricted to economic process improvement as ecological aspects related to process energy consumption are also investigated and Pareto-front analysis is conducted. More importantly, the generated retrofit alternatives are also classified according to path flow indicator differences compared to the base case(s). This enables a more detailed analysis of the diverse effects a retrofit alternative can have. In particular, it is investigated in what extent the indicator values are shifted after applying the respective retrofit alternative. Since the indicator values of the base case(s) are typically used as drivers to define retrofit alternatives, the quantification of the effect of retrofit alternatives on the indicators is crucial for the robustness of the methodology. In this context, the objective function differences of the retrofit alternatives are compared to the path flow indicator differences in order to analyze the trends. For the analysis of the trends a new approach is proposed that considers batch-to-batch variability to cluster similarly performing retrofit alternatives, providing a more robust interpretation of the retrofitting results.

The presented retrofit methodology was applied to an industrial mono-product batch process for a specialty chemical. Based on the process data treatment results the methodology was first evaluated for an average batch, and then a set of reference batches was selected to describe batch-to-batch variability. For all scenarios, the path flow indicator matrix of the base case(s) was calculated and with the help of applicable heuristics and sensitivity analysis on process parameters promising retrofit alternatives were defined. The newly introduced renting cost indicator identified the potential for unit occupancy time reduction and was demonstrated to be a valuable complement to the rest of the path flow indicators. Moreover, trends have been revealed between the impact of retrofit alternatives on well-defined sections of the path flow indicator matrix  and their effect on single objectives and multi-criteria Pareto-fronts. This is an important step towards a better understanding of the conditions under which a retrofit action originally inspired by path flow indicator values will be indeed successful from an overall system perspective. Finally, batch-to-batch variability innovatively coupled with a fuzzy clustering approach to provide a differentiation measure for retrofit alternatives was shown to smoothen the trendlines between changes in path flow indicator matrix sections due to retrofit alternatives and their multi-objective Pareto front relative positions. Overall, the application of the method has been able to reveal retrofit alternatives leading to 5-10% reduction in the environmental impact of an industrial batch process already running for a significant time period.