Fuzzy Logic and Fuzzy Hybrid Techniques for Construction Engineering and Management

Construction tasks, such as scheduling and project site layout, often have multiple objectives (e.g., cost, time, resource use, quality, and safety) that must be met simultaneously. These tasks are therefore formulated as nonlinear optimization problems, but because of their complexity, the optimum solution cannot always be determined analytically. In such cases, merely good solutions are sometimes sought through heuristic and metaheuristic optimization. In construction problems, however, subjective uncertainty must frequently be dealt with, and traditional heuristic and metaheuristic optimization cannot account for it. When fuzzy logic is combined with optimization, the resulting fuzzy optimization is able to process fuzzy variables and/or fuzzy restrictions (Haghighi and Ayati 2016). The optimization algorithm can be improved when fuzzy logic is applied to the optimization method itself (Cheng and Prayogo 2017). Fig. 1 shows examples of optimization that have been combined with fuzzy logic for use in construction.

With the advent of new technologies for automated data collection, storage, and analysis, construction organizations have been increasing the amount and types of data they collect. This change has led to greater interest in machine learning, which is a field of computer science wherein systems are developed that can learn from data for different applications [e.g., pattern recognition (Lu et al. 2018), predictive model development (Zuo and Xiong 2018), and data classification (Li et al. 2017)]. These systems cannot, however, process subjective variables and reasoning, which are often present in construction problems. Fuzzy logic systems based purely on expert knowledge, on the other hand, are not able to learn from data and adapt to new contexts (e.g., different construction sectors or different projects in the same sector). Models that integrate fuzzy logic and machine learning (i.e., models that combine knowledge- and data-driven approaches) are more reflective of reality than those that use only one kind of approach; they can learn from data, and they are appropriate for use in construction (Hüllermeier 2015). Fig. 1 shows examples of machine learning that have been combined with fuzzy logic and applied in construction.

In construction, expert knowledge is often used in decision making. Multicriteria decision making (MCDM) can help a group of experts select and rank alternative solutions according to various conflicting criteria. While MCDM has been used successfully, it does not capture the vagueness and subjective uncertainty that often accompany construction problems, which are further complicated by incomplete information, imprecise data, multiple criteria, and multiple decision makers. In addition, some criteria are not naturally evaluated numerically (i.e., assigned a crisp number) by experts. In these cases, it is more appropriate for experts to use linguistic assessments and natural language to evaluate criteria. To make it possible for experts to use natural language in their evaluations, fuzzy logic is integrated with MCDM to develop fuzzy MCDM (Chen and Pan 2018). Fig. 1 shows examples of MCDM that have been combined with fuzzy logic for use in construction.

Simulation is used in construction to develop and execute computer-based models of systems, such as construction processes and project management practices, in order to understand underlying behaviors, make predictions, and improve performance. The behavioral uncertainties of real-world construction systems fall into two categories: probabilistic or random uncertainties that can be modeled using numerical data, and nonprobabilistic uncertainties that include subjective or linguistically expressed information. Commonly used simulation techniques can address some of the complexities of construction systems, such as dynamism and the interactions between influencing factors, but they are not able to account for nonprobabilistic uncertainties, and they rely on the availability of sufficient numerical data. By integrating fuzzy logic with simulation techniques, researchers have developed fuzzy simulation techniques that can handle both probabilistic and nonprobabilistic uncertainties and overcome the lack of sufficient numerical data in modeling. Fig. 1 shows examples of simulation techniques that have been combined with fuzzy logic and applied in construction.

Productivity is a major concern of the construction industry, and methods to improve productivity were identified as an area of interest by several of the author’s industry partners. A multitude of factors and practices on both a project and an organizational level interact and affect productivity. Some of these variables are difficult to quantify and are best evaluated subjectively, while others are quantitative. Because of the large number of variables that affect productivity, it is challenging to use either purely expert knowledge or purely statistical methods to establish the relationships between variables. The author therefore combined fuzzy logic with clustering, machine learning, and genetic algorithms, an evolutionary optimization technique, to develop labor productivity prediction models for different project contexts. Her research team collected expert knowledge and field data and used fuzzy $c$-means (FCM) clustering to develop fuzzy inference systems (FISs) to model the complex relationships among variables that affect productivity. These FISs were used to predict concreting labor productivity for four different project contexts (industrial, warehouse, high-rise, and institutional buildings). The FISs were optimized using genetic algorithms to improve both accuracy and interpretability, the latter of which allows users to understand the reasoning behind the models and their results (Tsehayae and Fayek 2016). These techniques were also applied to develop labor productivity prediction models for electrical and shutdown activities. The context-specific FISs were then adapted for different project contexts using linear and nonlinear evolutionary-based transformation, thereby reducing data collection requirements for developing new models in different contexts (Tsehayae and Fayek 2018).

Construction organizations are seeking innovative approaches to measuring their internal competencies, so they can improve their project and organizational performance, which has traditionally been assessed using predefined performance measures. However, the relationships between competencies and performance measures are rarely defined. Competencies are both functional (i.e., practice–based) and behavioral (i.e., people–based). It is challenging for organizations to model and map competencies to performance because of the multidimensional nature of the variables and their relationships. The author and her research team used fuzzy machine learning and fuzzy MCDM to overcome this challenge (Omar and Fayek 2016). They integrated prioritized fuzzy aggregation, factor analysis, and fuzzy neural networks (FNNs) to develop a model that can identify the competencies that contribute most significantly to improvements in project key performance indicators (KPIs), as shown in Fig. 2. The model makes it possible for construction organizations to focus on improving the competencies that have the greatest effect on overall project performance. The author and her research team are now expanding these models to the construction organization level to determine what makes organizations competent and how to measure, predict, and improve construction organization performance (Tiruneh and Fayek 2019).

Many of the author’s research projects require that experts provide their opinions in multicriteria group decision making (MCGDM); either these opinions must be aggregated or consensus must be achieved among experts. Since better solutions are achieved when multiple experts participate in decision making, it is common to find a group of decision makers representing different points of view, backgrounds, and levels of expertise working together to solve construction problems. If the group of experts is heterogeneous (i.e., the importance of the experts’ opinions varies based on their experience and knowledge levels), it is common to assign qualitative or quantitative weights to the experts’ opinions based on how relevant their expertise is to the problem (i.e., based on an expert’s degree of importance).

The author has developed a variety of techniques for assigning weights to experts and aggregating their opinions or reaching consensus in applications that involve MCGDM. The systems developed using these techniques contain collective knowledge that is more heavily weighted toward greater expertise, thereby improving the MCGDM process. For example, Elbarkouky and Fayek (2011b) determined expert weights from a consensus weight factor for each expert based on the similarity of his or her opinions to the opinions of other experts, and used these weights in modified similarity aggregation. They used this method in a fuzzy similarity consensus model to define stakeholder roles and responsibilities in different project delivery systems.

To determine expert weights based on essential qualification attributes, Elbarkouky and Fayek (2011a) used an FIS incorporated in fuzzy preference relations consensus to reduce conflicts in the assignment of task responsibility between owners and contractors in different project delivery systems. Awad and Fayek (2012) determined expert weights based on experience measures using a multiattribute utility function combined with the analytic hierarchy process (AHP). They then used the weights in aggregating experts’ opinions to develop an FIS for predicting the risk of contractor default in surety bonding. Monzer et al. (2019) used fuzzy AHP to determine the relative importance of a set of attributes related to decision makers’ expertise in order to derive expert importance weights. They combined this technique with fuzzy aggregation operators to aggregate experts’ opinions on the probability and impact of construction project risks (i.e., threats) and opportunities in order to determine project contingency. Fayek and Omar (2016) used a fuzzy technique for order of preference by similarity to ideal solution (TOPSIS) to aggregate a group of decision makers’ views on the importance and levels of satisfaction of a set of interrelated criteria, which they used in prioritized fuzzy aggregation to evaluate the relative importance of construction project competencies that affect performance.

Although labor productivity is important to construction owners and contractors, overall project productivity is equally important in determining cost performance. Based on her industry partners’ input and changing needs, the author and her team expanded labor productivity prediction models to include other resources, such as equipment and materials. In order to identify the most significant drivers of productivity, they also explored relationships between factors that affect productivity and how those factors interact and change over the course of a project. They accomplished these tasks by developing fuzzy system dynamics (FSD) modeling approaches that process both fuzzy and deterministic inputs and capture the dynamic aspects of productivity, using them to create a predictive model of multifactor productivity for equipment-intensive activities (Gerami Seresht and Fayek 2018). The author and her team modeled relationships between system variables using either mathematical equations or FISs developed via FCM clustering. They used fuzzy arithmetic in mathematical equations containing subjective system variables, and explored different fuzzy arithmetic operations to determine the best approach for reducing uncertainty overestimation in the simulation results of the FSD model. They are now exploring integration of fuzzy machine learning techniques to define the relationships between system variables in order to increase model accuracy. Siraj and Fayek (2016) used these FSD modeling approaches to develop a model to assess construction project risks and opportunities and to determine project contingency (Siraj and Fayek 2016); this model is now being expanded to determine the most appropriate risk response strategies and to assess the impact of these strategies on project contingency.

Another important aspect of productivity is construction crew motivation and performance, which is of concern to many owners, contractors, and labor organizations. The overall performance of a crew is a function not only of its individual members, but also of the dynamics of the crew as a whole, and the situation or context in which the crew performs its tasks. The author and her team applied modern motivation theories, based on the concepts of efficacy, engagement, identification, and cohesion, and modeled these factors, not only at the individual level, but also at the crew level and in the project context (Raoufi and Fayek 2018). They also defined a number of metrics to measure the impact of crew motivation on performance, the latter of which has some subjective measures, such as counterproductive behavior. In order to model the attributes, behaviors, and interactions of crew members and crews with each other and their environment, and to derive overall crew behavior to predict performance, the author and her team developed a fuzzy agent–based modeling (FABM) method. This method enables agent-based modeling to incorporate subjective uncertainty in the form of membership functions, fuzzy rules, and an FIS to model agent attributes and behaviors. This model accurately predicts crew performance based on crew motivation levels, so that both motivation and performance can be most effectively improved. The author and her team are now expanding their FABM to include probabilistic uncertainty, and they are using Monte Carlo simulation to explore and compare the performance of crews in various scenarios.

Introduced by Zadeh (1975), fuzzy numbers are specific fuzzy set types that are used when exact values are not precisely measurable. The author has used fuzzy numbers and fuzzy arithmetic in several applications. For example, the author and her team applied fuzzy arithmetic methods to determine construction project contingency (Elbarkouky et al. 2016), and used them in FSD models to predict productivity (Gerami Seresht and Fayek 2018) and to determine construction project contingency (Siraj and Fayek 2016). There are two approaches for implementing fuzzy arithmetic. The first is the $\alpha$-cut method, which is simple to implement but results in overestimation of uncertainty that increases with each calculation step and reduces the interpretability of the results. The second uses the extension principle, which reduces overestimation but has been limited in implementation to computational methods that use only two $t$–norms (i.e., min and drastic product); the application of these $t$–norms creates challenges in implementing fuzzy arithmetic. The min $t$–norm produces the same result as that of the $\alpha$-cut, along with the same overestimation of uncertainty, but the drastic product $t$–norm produces results that are highly sensitive to changes in inputs, limiting its use in construction. The author and her research team developed novel computational methods for performing fuzzy arithmetic by the extension principle on triangular fuzzy numbers using the drastic product $t$–norm and the Lukasiewicz (i.e., bounded difference) $t$–norm. The drastic product and Lukasiewicz $t$–norms reduce overestimation of uncertainty and sensitivity to changes in inputs, thus making the use of fuzzy arithmetic in construction more effective (Gerami Seresht and Fayek 2018, 2019). These computational methods are being further developed to include trapezoidal and Gaussian fuzzy numbers and to incorporate other $t$–norms, such as the family of Yager $t$–norms, which are parametric.

The author’s industry partnerships have brought together owners, contractors, and labor groups, some of which are direct competitors, to collaborate on issues of industry-wide importance. Through these partnerships, some of which have spanned close to 23 years, she has contributed to a culture of university-industry collaboration where each party views the other as part of the fabric of doing business and increasing knowledge in the construction domain. The multiple and diverse perspectives of her industry partners have enabled her and her team to carry out research that is important to all parties in construction and to develop comprehensive solutions that can only be addressed through a multiparty collaborative effort. The long-term sustained funding from the NSERC IRC has ensured the continuity of her research program by providing uninterrupted financial support. This program has also helped ensure that industry partners provide meaningful input to the research by contributing to the definition of problems and providing extensive in-kind support, including access to expert knowledge and data. The consistency of her partners has also ensured that her research evolves over time to address the most relevant issues facing the industry.

For example, the author’s productivity studies have covered two IRC terms, and they have involved a cross section of her industry partners. Initially, these studies focused on labor productivity (Tsehayae and Fayek 2016, 2018), as this was identified as the major concern of the construction industry at the time. These studies have since been expanded to include multiresource project-level productivity (Gerami Seresht and Fayek 2018) and overall capital project productivity (Ayele and Fayek 2019), based on the changing needs of the construction industry and the author’s partners. These studies have involved significant expert knowledge and field data collection, and they have produced a number of data collection instruments, some of which are being used by industry partners. For example, PCL Construction, a leading contractor, has incorporated these data collection tools in a training manual for foremen. In each study, participating companies received confidential reports based on their data, and all data were pooled to identify the most significant factors affecting productivity in the Alberta construction industry.

Through her industry partnerships, the author has applied her fuzzy logic techniques to solve many real-world, practical problems. The in-kind support provided by industry partners has enabled her team to develop and validate their work using real-world data to ensure the quality of the developed models and the usefulness of the developed products. Students are often on site or in corporate offices, working alongside industry personnel to conduct research and implement their findings.

A representative example of the author’s ability to use fuzzy logic to solve practical construction problems in collaboration with industry partners is the development of the Fuzzy Contingency Determinator (FCD) software tool. Capital Power Corporation (CPC), one of the author’s industry partners, is a large owner organization that constructs power generation facilities using natural gas, coal, and wind. Risk management and contingency reserve assessment are critical to its operations, but the company was facing a number of challenges using Monte Carlo simulation for risk analysis. Since it often lacked adequate quality historical data, as is frequently the case in construction, the company was relying on subject matter experts to choose probabilistic distributions of costs (i.e., inputs to the simulation) based on experience. However, it is difficult to schedule meetings for experts; it is time-consuming to arrive at consensus; and experts would rather use natural language (e.g., high, medium, low) than probability distributions of costs to assess risks, making it necessary to calibrate each expert’s meanings for the linguistic terms. CPC also wanted approaches that would accurately capture uncertain events in terms of both risks (i.e., threats) and opportunities, and would provide the company with a reliable and accurate estimate of contingency. By applying fuzzy logic techniques, the author and her team developed a novel approach to risk analysis and contingency determination that overcomes the challenges associated with existing methods of risk analysis that rely on historical data and involve time-consuming consensus-reaching processes.

The author and her team developed FCD, which both allows experts to express the probability and impact of risks and opportunities using linguistic terms, and provides users with the ability to calibrate each expert’s meaning of the terms to suit different projects. To save CPC even more time, the author and her team developed a fuzzy MCGDM approach, based on fuzzy AHP and fuzzy aggregation operators, to weight individual experts based on their level of expertise so that their opinions could be aggregated rather than forcing them to reach consensus (Monzer et al. 2019). Using fuzzy arithmetic, FCD calculates the contingency allowance for a project, as shown in Fig. 4 (Elbarkouky et al. 2016). FCD not only gives contingency values comparable to those derived through Monte Carlo simulation, but also narrows the uncertainty range, freeing up contingency reserve for use on other projects. The author’s application of fuzzy logic has helped CPC significantly reduce the time and effort required for risk analysis. FCD has significantly improved the efficiency and accuracy of CPC’s risk assessment processes and has been fully integrated into its daily practices, replacing its existing Monte Carlo simulation methods. The author and her team continue to enhance FCD, and based on demand from other industry partners, they are developing a generalized risk analysis software tool, Fuzzy Risk Analyzer (FRA), that can be customized by any owner or contractor to suit different project types and construction sectors.

Through her collaborative research program, the author has completed many short-term projects that had immediate benefits and bridged the gap between academic theory and engineering practice. Developing solutions to practical problems that are identified as relevant and important by industry has helped build trust with and make an impact on the construction industry. The author has implemented these solutions in useful products and found effective ways to disseminate results so that her industry partners can use them in practice, thereby helping organizations better manage their projects and workforces and increase their competitiveness.

The author has found the following methods to be the most effective for research dissemination:

By adapting fuzzy logic and fuzzy hybrid techniques to suit the characteristics of the construction domain and implementing these techniques in tools that are readily accepted by practitioners, the author has contributed to more widespread acceptance of fuzzy logic across both academia and industry. The solutions she and her team have developed have provided industry practitioners with an intuitive approach for expressing their opinions and have made construction decision making more transparent. With more experienced people leaving the industry than entering it, there is a lack of continuity and proper transfer of knowledge and skills between construction personnel. The methods and tools the author has developed have helped document and capture this knowledge so that it can be used in future decision making and in the training of new personnel.

The focus of academic journals is to disseminate and preserve knowledge, rather than to instruct practitioners on how to apply research findings. The author proposes creating a companion journal for a top-tier academic journal that would allow researchers to publish a practice paper based on a published academic paper. This practice paper would be grounded in the theory and rigor of the published academic research, but it would focus on how the results can be applied in practice and would use language that is more readily understood by nonexperts in the field. In addition, creating efficient methods of training industry in the products researchers develop, whether new software, processes, or research findings, is critical if practitioners are to use the results. Applying research results and innovations to demonstration projects and sharing findings will increase impact and encourage other organizations to adopt research products. Sharing intellectual property developed through research partnerships is also important so that both academics and practitioners can benefit from it.

Real breakthroughs in construction research require an interdisciplinary approach and the ability to communicate, collaborate, and build relationships outside areas traditionally associated with the construction industry. Collaborations that work are often based on both a complementarity of skills and a shared set of values and objectives. To promote such collaborations, the academic construction community needs more multidisciplinary and interdisciplinary funding programs and avenues for dissemination, such as strategic networks that require researchers to be from different departments, faculties, or institutions. Funding mechanisms that support people rather than projects can help facilitate meaningful research, allowing researchers to focus on longer-term programs. Formal industry-academic partnerships that are underpinned by long-term sustained funding are also beneficial in creating impactful research.

One of the goals of construction research is to develop practical solutions that can be readily implemented to improve construction performance and competitiveness. Research and development (R&D) partnerships are one mechanism to help achieve this goal, and demonstrating the value of these partnerships is essential for encouraging investment in research. In addition, demonstrating the measurable impact of research products is important for encouraging industry to adopt them in practice.

However, measuring the impact of research on stakeholders, such as industry, government, and society, is difficult, as the impact may not be immediately evident. As part of the author’s efforts to measure the impacts of her own research program, she developed a framework to assess the impacts of R&D partnerships on university, industry, and government (Daoud et al. 2017). While the engineering domain lacks such a formal evaluation framework, some industries (e.g., healthcare, finance, education) have formal methods of measuring research impact through a logic model. The author adapted the logic model approach to suit the construction domain and applied it to university-industry-government partnerships. For each stakeholder, the framework identifies: research inputs in the form of resources, such as funding and access to data; outputs in the form of activities, such as publications, workshops, and training; and impact of short-, medium-, and long-term outcomes in the form of internationally recognized expertise, creation of innovative technologies, impact on codes and standards, and increased profitability and competitiveness of construction organizations. This framework can help partners in a research partnership understand how their investment of resources and their activities affect targeted outcomes, and subsequently, it can help them increase the impact of their efforts. Additionally, by demonstrating the measurable value of construction R&D, this framework may encourage industry and government to implement research findings and further invest in R&D programs.