Knowledge Extraction and Fuzzy Expert System Modeling for IoT-Enabled Asset Management in Construction and Infrastructure Projects: Insights from a Practitioner Survey

Abstract

IoT technologies have the potential to provide real-time visibility of labour, materials, and plant and machinery on construction and infrastructure projects, but the knowledge gained from practical experience is rarely turned into reusable decision-making knowledge. Most existing studies offer only descriptive observations or technical pilot projects and rarely convert survey findings into clear, rule-based models that can support asset management decisions. In this study, knowledge is drawn from a survey of practitioners and a fuzzy expert-system model is developed which links the perceived benefits of IoT for labour, materials, and plant and machinery to the perceived project performance. The questionnaire was completed by 75 construction professionals (96% of them working in India) and included eleven five-point Likert items, which were analysed using the relative importance index, Cronbach's alpha, exploratory factor analysis, non-parametric tests, and clustering. A Wang–Mendel Mamdani rule base and a Takagi–Sugeno–Kang model were trained and compared with linear regression, random forest, and a mean-only baseline using leave-one-out cross-validation. Perceived benefits were high (mean 3.96 of 5; relative importance index 0.79), and 97.3% of respondents would recommend IoT-based solutions. Materials management (mean 4.19) and plant and machinery management (mean 4.17) ranked highest, whereas labour productivity ranked lowest (mean 3.57). Three factors explained 63.1% of variance, and internal consistency across all items was high (alpha = 0.89). Plant and machinery benefits were the dominant predictor of perceived project performance (standardised beta = 0.56, p < 0.001). Cross-validated mean absolute error was 0.49 for regression, 0.53 for the Takagi–Sugeno–Kang model, 0.59 for the Mamdani model, and 0.64 for the baseline, although differences between models were not statistically significant. The fuzzy approach yielded 15 interpretable, data-derived rules. Practitioners are strongly positive about IoT-enabled asset management, with plant and machinery the most credible entry point. The rule base provides transparent decision knowledge, but the small, India-centred sample and perception-based outcomes limit generalisation; validation with sensor data from live projects is required.