Abstract
Artificial intelligence techniques are at the centre of a major shift in business today. They have a very broad array of applications within businesses, including that of optimisation for risk reduction in civil engineering projects. This is an active area of research, which has started to see real-world applications over the last few decades. It is still hindered by the extreme complexity of civil engineering problems and the computing power necessary to tackle these, but the economic and other benefits of these emerging technologies are too important to ignore. With that in mind, this chapter reviews the current state of research and real-world practice of optimisation techniques and artificial intelligence in risk reduction in this field. It also examines related promising techniques and their future potential.
TopIntroduction
Civil engineering comprises the design, construction, operation, and maintenance of buildings and infrastructures including a variety of works such as residence, bridges, and roads (Zavala et al., 2014). Since the second World War, with the rapid advances made in computational methods, optimisation techniques based on mathematical programming have been increasingly deployed in the field of civil engineering (Topping, 1983). Optimisation refers to acquiring the best outcome under specific conditions (Rajput & Datta, 2019), and optimization problems are evident in many disciplines, including operations research, computing, engineering and economics. Optimisation techniques consist of a powerful set of tools that can be deployed to help the effective management of a company’s resources, and can be seen as an artificial intelligence (AI) tool. Kolter and Procaccia (2017) noted that “one of the most significant trends in AI in the past 15 years has been the integration of optimization methods throughout the field” (p.5). More specifically, in the field of civil engineering, optimisation can be executed in each step of a project life cycle, from design and construction to operation and maintenance, but can also be applied more generally to risk estimation and reduction (Dede et al., 2019; Mei & Wang, 2021).
Optimisation algorithms can be used to identify solutions to correctly formulated optimisation problems. This requires the formulation of an equation, or objective function, which calculates a measure of performance. Variables in the problem being optimised can then be represented as combinations of parameters for this function. For example, a problem such as drainage network optimisation can be represented as an optimisation problem by the use of an equation for calculating the cost of making changes to an existing drainage network. The parameters for this equation involve a mobilization cost (M) representing an initial cost of making change. Additionally, a combination of a cost for each pipe altered (I), the length of pipe requiring alteration (L) and the cost of purchasing pipes of a particular cross-section (c). Finally, storage tank alteration costs (S), area of the storage tank (a) and a base cost (b) as shown in equation 1 (Sayers, 2015).
(1)Importantly for most algorithms to function well, this objective function must return values which are differentiable and represent the problem space well. This means that identifying a suitable objective function is in many cases one of the most complex parts of formulating a problem as an optimisation problem. Making this more challenging, is the fact that in most optimisation algorithms this objective function will be called a very high number of times, as different solutions are tested through the optimisation process. This is because of an effect known as “combinatorial explosion”, where the number of parameters has an polynomial effect upon the size of the search space (dramatically increasing it in response to a small increase in parameters, or parameter values). Because of this, if an objective function takes even a few seconds to complete, this will result in extremely time-consuming algorithm runs when trying to be sure to undertake a reasonable exploration of the search space. Unfortunately, the majority of civil engineering optimisation problems are extremely complex and sometimes multi-objective optimisation problems, meaning they are generally computationally intractable, as exhaustive, or even mostly exhaustive, searches. Because of this, optimisation algorithms for these applications are generally non-deterministic, and are based around the use of heuristics. These heuristics allow them to be applied to problems where an exhaustive search for solutions with a computer is impractical, and still identify a reasonable solution. If this problem is even more pressing, then meta-heuristics and artificial intelligence techniques can be used to further alleviate this, such as machine learning regressors trained to approximate the above equation without the need to complete the processing associated with a full evaluation of this equation.
The objectives of this chapter are to first present a review of academic literature on optimisation techniques, including an outline of optimisation techniques and how they work and may be applied to engineering related problems. This is followed by an assessment of some examples of risk reduction and risk assessment optimisation in civil engineering, and a discussion of emerging issues. Finally, the concluding section offers some reflection on research directions, prospects for the field, and the practicalities of applications.
Key Terms in this Chapter
Non-Dominated Front/Set: Equivalent to the Pareto front/set, non-inferior front/set.
Optimisation Problem: A problem consisting of a number of parameters which can be manipulated to result in changes to the objective functions result, and one or more objectives which are quantifiable and differentiable.
Heuristic: A process by which an approximate answer to a solution can be derived.
Fitness Function: Equivalent to objective function, evaluation function.
Hyper Parameter: A value which can be modified, resulting in some change to the function of a computer model/process which is manipulating parameters relating to an objective function.
Optimisation Algorithm: An algorithm which can be applied to a properly structured optimisation problem, identifying one or more solutions which meet defined criteria of suitability.
Parameter: A value which can be modified, resulting in some change to an objective functions value.
Meta-Heuristic: A heuristic process which is an addition to a process which is already in itself heuristic.
Non-Inferior Front/Set: Equivalent to the Pareto front/set, non-dominated front/set.
Evaluation Function: Equivalent to objective function, fitness function.
Pareto Front/Set: The Pareto front/set is a set of all solutions in a multiple-objective search space, for which there is no way to improve performance in any objective, without simultaneously decreasing performance in one or more other objectives. Equivalent to non-dominated front/set, non-inferior front/set.
Objective Function: A function which when supplied a number of parameters relating to an optimisation problem, will return a value identifying the suitability of those parameters as a solution to the problem. Equivalent to fitness function, evaluation function.