Identifying potential moisture damage is crucial for maintenance practices and assurance of well-being of oc cupants. However, due to limited information availability and standardization, assessing damage prevalence on the building stock scale remains understudied. By combining investigation records and building databases, this study leverages data analytic techniques and machine learning modeling to characterize damage pathology and predict its occurrence in Swedish buildings. The interrelationships between damage-specific attributes and their associations with building parameters of several damage types were analyzed using feature selection, forming the basis for developing predictive models. Results show that multilabel classifiers outperform binary classifiers for every damage type, with lead tree ensemble models achieving minimum average AUCPR and F2 of 0.85 for microbial growth, 0.87 for deformation, 0.91 for odor, and 0.95 for water leakage. The identified patterns were interpreted and verified against descriptive statistics. The binary relevance models estimate that one-third of school buildings, 20 % of commercial and office buildings, and 15 % of residential dwellings in regional building stock contain moisture damage. These findings advance the quantification of moisture damage by providing new knowledge and approaches for appraising moisture damage likelihood at aggregated and individual building levels, thereby aiding in moisture safety evaluations and preventive maintenance efforts

Predicting modeling and analysis of moisture damages in Swedish buildings Moisture damage in buildings poses significant challenges, leading to costly repairs and negative impacts on indoor environments. Understanding the patterns and prevalence of such damage is crucial for implementing effective preventive and mitigative measures. This report synthesizes findings from the comprehensive study on moisture damage in Swedish buildings, highlighting the complex interplay of multidimensional variables that contribute to these issues. Analyzing 2,100 moisture-related damage records from empirical damage investigations from 2014 to 2020, the study employs data-driven analytical techniques to identify primary moisture damage profiles and evaluate their prevalence. Multivariate analyses reveal diverse associations among factors related to moisture damage, with significant variations in damage types across different building components, phases, and actors responsible. The findings indicate that moisture damage is most prevalent in buildings constructed between 1960-1980 and 2000-2020, with common issues including microbial growth, deformation of building envelopes and roofs, odors caused by humidity-related problems and wind-driven rain. Buildings with non-ventilated crawlspaces and non-ventilated attics are particularly susceptible. By characterizing moisture damage patterns and their associated factors, these findings enhance our understanding of moisture damage occurrences and informs strategies for relevant actors to improve their practices on moisture damage prevention along building lifecycles. Furthermore, various predictive modeling techniques and machine learning algorithms were explored to predict the presence of prevalent moisture damage types, including microbial growth, deformation, odor, and water leakage. The binary and multilabel classification models achieve high prediction rates of over 0.9 average AUCPR and F2 scores by training with building-related features in cross-validation and validation. The lead binary relevance models estimate that approximately one-third of school buildings, 20% of commercial and office buildings, and 15% of residential dwellings in the uninvestigated Swedish building stock may contain unreported moisture damage. These findings advance quantitative research on moisture damage, providing new insights and tools for assessing the extent and patterns of in-situ moisture damage, thereby aiding existing moisture safety evaluations, and building maintenance strategies. Specifically, the results can help building owners, insurance companies, and damage investigators predict and address potential moisture damage in buildings. They also support developers, designers, and contractors in making informed decisions about construction solutions, helping to avoid designs with a high risk of damage. By preventing moisture damage in both existing and new buildings, we can significantly reduce costs and the environmental impact of remediation and material replacement, which currently result in substantial expenses and CO2 emissions.

The undesirable legacy of radioactive concrete (blue concrete) in post-war dwellings contributes to increased indoor radon levels and health threats to occupants. Despite continuous decontamination efforts, blue concrete still remains in the Swedish building stock due to low traceability as the consequence of lacking systematic documentation in technical descriptions and drawings and resource-demanding large-scaled radiation screening. The paper aims to explore the predictive inference potential of learning Bayesian networks for evaluating the presence probability of blue concrete. By integrating blue concrete records from indoor radon measurements, pre-demolition audit inventories, and building registers, it is possible to estimate buildings with high probabilities of containing blue concrete and encode the dependent relationships between variables. The findings show that blue concrete is estimated to be present in more than 30% of existing buildings, more than the current expert assumptions of 18–20%. The probability of detecting blue concrete depends on the distance to historical blue concrete manufacturing plants, building class, and construction year, but it is independent of floor area and basements. Multifamily houses and buildings built between 1960 and 1968 or nearby manufacturing plants are more likely to contain blue concrete. Despite heuristic, the data-driven approach offers an overview of the extent and the probability distribution of blue concrete-prone buildings in the regional building stock. The paper contributes to method development for pattern identification for hazardous building materials, i.e., blue concrete, and the trained models can be used for risk-based inspection planning before renovation and selective demolition. © 2023 The Author(s)

Indoor radon represents a health hazard for occupants. However, the indoor radon measurement rate is low in Sweden because of no mandatory requirements. Measuring indoor radon on an urban scale is complicated, machine learning exploiting existing data for pattern identification provides a cost-efficient approach to estimate indoor radon exposure in the building stock. Extreme gradient boosting (XGBoost) models and deep neural network (DNN) models were developed based on indoor radon measurement records, property registers, and geogenic information. The XGBoost models showed promising results in predicting indoor radon intervals for different types of buildings with macro-F1 between 0.93 and 0.96, whereas the DNN models attained macro-F1 between 0.64 and 0.74. After that, the XGBoost models trained on the national indoor radon dataset were transferred to fit building registers in metropolitan regions to estimate the indoor radon intervals in non-measured and measured buildings by regions and building classes. By comparing the prediction results and the statistical summary of indoor radon intervals in measured buildings, the model uncertainty and validity were determined. The study ascertains the prediction performance of machine learning models in classifying indoor radon intervals and discusses the benefits and limitations of the data-driven approach. The research outcomes can assist preliminary large-scale indoor radon distribution estimation for relevant authorities and guide onsite measurements for prioritized building stock prone to indoor radon exposure. 

Hazardous materials in buildings cause project uncertainty concerning schedule and cost estimation, and hinder material recovery in renovation and demolition. The study aims to identify patterns and extent of polychlorinated biphenyls (PCBs) and asbestos materials in the Swedish building stock to assess their potential presence in pre-demolition audits. Statistics and machine learning pipelines were generated for four PCB and twelve asbestos components based on environmental inventories. The models succeeded in predicting most hazardous materials in residential buildings with a minimum average performance of 0.79, and 0.78 for some hazardous components in non-residential buildings. By employing the leader models to regional building registers, the probability of hazardous materials was estimated for non-inspected building stocks. The geospatial distribution of buildings prone to contamination was further predicted for Stockholm public housing to demonstrate the models’ application. The research outcomes contribute to a cost-efficient data-driven approach to evaluating comprehensive hazardous materials in existing buildings.

The presence of hazardous materials inhibits material circularity. The existing residential buildings are exposed to the risk of the unforeseen presence of asbestos-containing materials during the demolition or renovation process. Estimating the potential occurrence of contaminated building components can therefore facilitate semi-selective demolition and decontamination planning. The study aims to investigate the prediction possibility of seven frequently detected asbestos-containing materials by using artificial neural networks based on a hazardous material dataset from pre-demolition audit inventories and national building registers. Through iterative model evaluation and careful hyperparameter tuning, the prediction performance for each asbestos-containing material was benchmarked. A high level of accuracy was obtained for asbestos pipe insulation and ventilation channel, yet barely any patterns were found for asbestos floor mats. Artificial neural networks show potential for classifying specific asbestos components and can enhance the knowledge of their detection patterns. However, more quality data are needed to bring the models into practice for risk assessment for not yet inventoried residential buildings. The proposed screening approach for in situ asbestos-containing materials has high applicability for the quality assurance of recycled materials in circular value chains.

Identifying in situ hazardous materials can improve demolition waste recyclability and reduce project uncertainties concerning cost overrun and delay. With the attempt to characterize their detection patterns in buildings, the study investigates the prediction potential of machine learning techniques with hazardous waste inventories and building registers as input data. By matching, validating, and assuring the quality of empirical data, a hazardous material dataset for training, testing, and validation was created. The objectives of the explorative study are to highlight the challenges in machine learning pipeline development and verify two prediction hypotheses. Our findings show an average of 74% and 83% accuracy rates in predicting asbestos pipe insulation in multifamily houses and PCB joints or sealants in school buildings in two major Swedish cities Gothenburg and Stockholm. Similarly, 78% and 83% of recall rates were obtained for imbalanced classification. By correlating the training sample size and cross-validation accuracy, the bias and variance issues were assessed in learning curves. In general, the models perform well on the limited dataset, yet collecting more training data can improve the model's generalizability to other building stocks, meanwhile decreasing the chance of overfitting. Furthermore, the average impact on the model output magnitude of each feature was illustrated. The proposed applied machine learning approach is promising for in situ hazardous material management and could support decision-making regarding risk evaluation in selective demolition work. © 2022 The Author(s)

The presence of hazardous materials hinders the circular economy in construction and demolition waste management. However, traditional environmental investigations are costly and time-consuming, and thus lead to limited adoption. To deal with these challenges, the study investigated the possibility of employing registered records as input data to achieve in situ hazardous building materials management at a large scale. Through characterizing the eligible building groups in question, the risk of unexpected cost and delay due to acute abatement could be mitigated. Merging the national building registers and the environmental inventory from renovated and demolished buildings in the City of Gothenburg, a training dataset was created for data validation and statistical operations. Four types of inventories were evaluated to identify the building groups with adequate data size and data quality. The observations’ representativeness was described by plotting the distribution of building features between the Gothenburg dataset and the training dataset. Evaluating the missing data and the positive detection rates affirmed that reports and protocols could locate hazardous materials in the building stock. The asbestos and polychlorinated biphenyl (PCB)-containing materials with high positive detection rates were highlighted and discussed. Moreover, the potential inventory types and building groups for future machine learning prediction were delineated through the cross-validation matrix. The novel study contributes to the method development for assessing the risk of residual hazardous materials in buildings.

Assessment of the presence of hazardous materials in buildings is essential for improving material recyclability, increasing working safety, and lowering the risk of unforeseen cost and delay in demolition. In light of these aspects, machine learning has been viewed as a promising approach to complement environmental investigations and quantify the risk of finding hazardous materials in buildings. In view of the increasing number of related studies, this article aims to review the research status of hazardous material management and identify the potential applications of machine learning. Our exploratory study consists of a two-fold approach: science mapping and critical literature review. By evaluating the references acquired from a literature search and complementary materials, we have been able to pinpoint and discuss the research gaps and opportunities. While pilot research has been conducted in the identification of hazardous materials, source separation and collection, extensive adoption of the available machine learning methods was not found in this field. Our findings show that (1) quantification of asbestos-cement roofing is possible from the combination of remote sensing and machine learning algorithms, (2) characterization of buildings with asbestos-containing materials is progressive by using statistical methods, and (3) separation and collection of asbestos-containing wastes can be addressed with a hybrid of image processing and machine learning algorithms. Analysis from this study demonstrates the method applicability and provides an orientation to the future implementation of the European Union Construction and Demolition Waste Management Protocol. Furthermore, establishing a comprehensive environmental inventory database is a key to facilitating a transition toward hazard-free circular construction

Hazardous materials encountered during building renovation or demolition processes not only result in uncertainty in cost estimation and the lead time but also hampers material recyclability and reuse. Therefore, the paper discusses the possibility of predicting the extent of the hazardous materials, including asbestos, PCB, mercury, and CFC, through data mining techniques based on registered records. Pre-demolition audits contain observation data that can be used as a sample for statistical prediction through careful processing. By developing an innovative approach of merging data from environmental inventories with building registers, the positive ratio of remaining hazardous materials in the Gothenburg building stock can be estimated. The study highlights the challenges of creating a training dataset by completing information from the existing environmental inventory, providing new insight into digital protocol development for enhancing material circularity.