Optimization of an Ammonia Synthesis Reactor using Simultaneous Approach

Daison Yancy-Caballero, Lorenz T. Biegler, Reginaldo GuirardelloSHILAP 2015 · DOI

TL;DR

This paper presents a novel approach to optimize the design of ammonia synthesis reactors for maximum economic return. By discretizing the complex, non-linear differential-algebraic equations governing the reactor and solving them simultaneously with an optimization algorithm, the authors achieve reliable and efficient results, demonstrating a robust method for tackling similar chemical engineering problems.

What Problem Does This Paper Address?

Ammonia is a crucial chemical feedstock for fertilizers, plastics, and pharmaceuticals. Its industrial production relies heavily on the Haber-Bosch process, with the ammonia synthesis reactor being a critical component. Optimizing this reactor's design, specifically its length, to maximize economic return is a well-studied problem. However, existing methods often simplify the reactor's complex, non-linear behavior, particularly by assuming constant thermodynamic properties. This paper aims to address this by developing and applying a more rigorous optimization methodology that accounts for temperature and pressure-dependent properties, leading to a more accurate and potentially more profitable reactor design.

Methods

The core of this paper's contribution lies in its optimization methodology. The authors model the ammonia synthesis reactor using a system of non-linear differential-algebraic equations (DAEs) that capture mass and energy balances, along with a modified Temkin-Dyson-Simon kinetic expression that accounts for fugacities and thus non-ideal behavior.

The key innovation is the simultaneous approach for solving this DAE-constrained optimization problem. Instead of solving the DAEs separately and then optimizing, they discretize the DAEs using orthogonal collocation on finite elements (OCFE). This transforms the continuous DAEs into a set of algebraic equations. These algebraic equations are then incorporated directly as equality constraints within a non-linear programming (NLP) problem.

The NLP problem is then solved using the IPOPT solver within the GAMS optimization modeling platform. To mitigate the risk of local optima with IPOPT, multiple initial guesses were employed. The paper also highlights the importance of including an effectiveness factor to account for intraparticle mass transfer resistance, a factor often omitted in previous studies.

Datasets

Not discussed in the paper. The paper uses a set of parameters for the simulated reactor, which are summarized in Table 1 and taken from a previous study by Murase et al. (1970).

Key Insights and Ideas

  • The power of simultaneous optimization: By treating the discretized reactor model equations as constraints within the optimization problem, the authors achieve a more holistic and accurate optimization compared to sequential approaches.
  • Accounting for non-ideal behavior matters: Using a modified kinetic expression that considers fugacities and temperature/pressure-dependent properties leads to a more realistic model and, consequently, potentially better optimization results.
  • Effectiveness factor is crucial: Omitting the effectiveness factor (which accounts for mass transfer limitations within the catalyst) can lead to suboptimal designs, as it can range from 0.3 to 0.5 for ammonia synthesis.
  • OCFE is robust: The orthogonal collocation on finite elements method proved to be a robust discretization technique, yielding consistent results even with a small number of finite elements.

Key Results

  • The proposed simultaneous optimization approach, using OCFE and IPOPT, successfully determined an optimal reactor length and other operating parameters to maximize economic return.
  • When comparing simulations with constant versus temperature/pressure-dependent properties, the latter resulted in a 9% increase in profit and a 5% increase in optimal reactor length.
  • The optimal reactor length obtained in this work (around 7.012 m) is comparable to recent studies (e.g., Yusup et al., 2006; Borges et al., 2012; Carvalho et al., 2014), suggesting the validity of their approach.
  • The OCFE method was found to be computationally efficient, with solutions obtained in a few CPU hours.

Limitations and Future Work

  • Local Optima: The authors acknowledge that IPOPT is a local solver and does not guarantee a global optimum. They mitigated this by using multiple initial guesses, but a truly global optimization strategy could be explored.
  • Decision Variables: The study focused on optimizing only the reactor length. The authors suggest that future work could expand the optimization to include other decision variables such as inlet feed temperature, feed rate, and the hydrogen-to-nitrogen ratio.
  • Model Complexity: While the model accounts for temperature and pressure-dependent properties and a modified kinetic expression, further refinements could include more detailed catalyst deactivation models or more complex reactor configurations.

Summary updated

Abstract

Ammonia is a valuable chemical produced in the industry and is wide used in the manufacture of fertilizers, plastics and pharmaceuticals. On the other hand, some important tools in chemical processes are modelling, simulation and optimization, and these types of studies have attracted considerable attention of many researchers. Several models of ammonia synthesis have been developed with the purpose of design and optimization. The typical design problem in this process requires obtained an optimal reactor length with maximum economic return and many works in literature have focused on the optimization of this reactor using different optimization techniques. In this work, an alternative approach to those presented in literature for solving this boundary value problem, and determine the optimal solution, is presented. The model of the reactor along with the kinetic form a non-linear differential-algebraic system. These differential-algebraic equations are discretized using orthogonal collocation on finite elements with continuous profiles approximated by Lagrange polynomials. In this manner, the resulting algebraic collocation equations are written as equality constraints in the optimization problem. Thus, the optimization problem is solved by implementing the IPOPT solver within the GAMS optimization-modeling platform. Although IPOPT solver here implemented is a local solver that does not guarantee the global optimum, multiple initial guesses were used to solve the problem and obtain an optimal solution of the problem. The obtained results are compared with other results available in literature and indicate that the used approach is a good alternative for dealing with this type of problems.
BibTeX
@inproceedings{yancycaballero2015optimization,
  author = {Daison Yancy-Caballero and Lorenz T. Biegler and Reginaldo Guirardello},
  title = {{Optimization of an Ammonia Synthesis Reactor using Simultaneous Approach}},
  booktitle = {SHILAP},
  year = {2015},
  doi = {10.3303/cet1543217},
}
Figure