Pan TC, Kao JJ (2009) GA-QP Model to Optimize Sewer System Design, JOURNAL OF ENVIRONMENTAL ENGINEERING, 135(1) 17-24
Summary
Sanitary sewers aren’t rocket science. Upstream elevations must be greater than downstream elevations. The diameters of downstream pipes must be larger than their upstream counterparts. A line must be sufficiently large to carry a design flow. These are usually the main concerns, which are often accompanied by other sets of secondary limitations. If viewed at the level of an individual sewer line, its optimal design is a fairly simple problem to solve. When optimizing the sewer design for a large network, however, the process becomes exceedingly complex. In this study, the authors attempted to tackle the latter.
The problem in this article was so complex, in fact, that the authors had to break the optimization process into two separate parts: a genetic algorithm and a quadratic program. In the genetic algorithm, a set of pipe diameters for the entire system constituted a chromosome of possible solutions. These chromosomes were then passed to the quadratic program, in which the cost was optimized by modifying pipe slopes and buried depths. Optimal solutions in the quadratic program were limited by an array of constraints commonly encountered in sewer design. The result of the quadratic program was then fed back to the genetic algorithm, where the fitness of a chromosome was evaluated as the inverse of its cost (as cost increased, the fitness of a chromosome became worse, and vice versa). Upon running the GA-QP model, results were fed into an MGA function, which was used to save “good alternative” solutions which were not mathematically optimal. When applied to a case study, the authors’ method was successful in reducing the cost of a sewer design from $87 million to $10 million – a reduction of nearly 90%.
Discussion
This article was an interesting implementation of a GA, but I found two major setbacks.
First and foremost, I feel that the authors’ exclusion of particular design factors and “unquantifiable issues” would render any optimal or near-optimal solution infeasible. Geology is a major factor in any project involving cut and fill, and can cause unexpected changes in design in order to minimize costs. Although the inclusion of “unquantifiable issues” is nearly impossible in a model, I feel that the constraints posed by these issues would result in a significantly more expensive final design than what the authors advertise.
Secondly, I feel that the authors’ method is fairly limited in applicability. In order for this GA-QP scheme to be implemented, a new, large sewer network must be installed at a single point in time. New residential neighborhoods are obviously a good application, but the method offers no suggestions of improvement to existing networks, or small expansions upon such networks.
Nevertheless, I found the authors’ ideas to be interesting and worthwhile. Even if real-world constraints reject optimal and near-optimal solutions generated by the model, these solutions may encourage new approaches to design which could potentially save millions.
Summary
Sanitary sewers aren’t rocket science. Upstream elevations must be greater than downstream elevations. The diameters of downstream pipes must be larger than their upstream counterparts. A line must be sufficiently large to carry a design flow. These are usually the main concerns, which are often accompanied by other sets of secondary limitations. If viewed at the level of an individual sewer line, its optimal design is a fairly simple problem to solve. When optimizing the sewer design for a large network, however, the process becomes exceedingly complex. In this study, the authors attempted to tackle the latter.
The problem in this article was so complex, in fact, that the authors had to break the optimization process into two separate parts: a genetic algorithm and a quadratic program. In the genetic algorithm, a set of pipe diameters for the entire system constituted a chromosome of possible solutions. These chromosomes were then passed to the quadratic program, in which the cost was optimized by modifying pipe slopes and buried depths. Optimal solutions in the quadratic program were limited by an array of constraints commonly encountered in sewer design. The result of the quadratic program was then fed back to the genetic algorithm, where the fitness of a chromosome was evaluated as the inverse of its cost (as cost increased, the fitness of a chromosome became worse, and vice versa). Upon running the GA-QP model, results were fed into an MGA function, which was used to save “good alternative” solutions which were not mathematically optimal. When applied to a case study, the authors’ method was successful in reducing the cost of a sewer design from $87 million to $10 million – a reduction of nearly 90%.
Discussion
This article was an interesting implementation of a GA, but I found two major setbacks.
First and foremost, I feel that the authors’ exclusion of particular design factors and “unquantifiable issues” would render any optimal or near-optimal solution infeasible. Geology is a major factor in any project involving cut and fill, and can cause unexpected changes in design in order to minimize costs. Although the inclusion of “unquantifiable issues” is nearly impossible in a model, I feel that the constraints posed by these issues would result in a significantly more expensive final design than what the authors advertise.
Secondly, I feel that the authors’ method is fairly limited in applicability. In order for this GA-QP scheme to be implemented, a new, large sewer network must be installed at a single point in time. New residential neighborhoods are obviously a good application, but the method offers no suggestions of improvement to existing networks, or small expansions upon such networks.
Nevertheless, I found the authors’ ideas to be interesting and worthwhile. Even if real-world constraints reject optimal and near-optimal solutions generated by the model, these solutions may encourage new approaches to design which could potentially save millions.
1 comment:
I agree that applications for this GA method seem very limited- pretty much only for new, large developments. I agree the the geological properties could play a role in economic feasibility. It would be interesting to see whether the different excavation costs could somehow be quantified to yield more realistic results.
Post a Comment