Peng Wu’s research while affiliated with Curtin University and other places

Road transport infrastructure is a crucial component of the entire infrastructural network. Timely and efficient maintenance of roads requires accurate and effective evaluation of pavement health, of which cracking is an important aspect. However, accurately assessing pavement cracks across large-scale road networks remains challenging due to spatial variations, which diminish the effectiveness of traditional machine learning methods. This study developed a novel spatial machine learning (SML) model and employed laser scanning data and satellite remote sensing images to assess road segment-based crack severity across the state-level road network in the Wheatbelt of Western Australia. Geocomplexity is introduced to measure the complexity of local patterns and spatial dependence among neighboring road segments. Results showed that SML can accurately and effectively predict pavement cracks on a large spatial scale with an accuracy (AC) from 0.524 to 0.701. In the SMLs, laser-scanning pavement variables contributed 34.15% to 43.33% of the total explainable variations, and geocomplexity variables also contributed significantly, ranging from 27.35% to 49.92%. The SML model exhibited the highest coefficient of determination ( $R^2$ ) and AC for crack prediction compared with Multiple Linear Regression (MLR), Generalized Additive Model (GAM), Bayesian Regularized Neural Network (BRNN) and Support Vector Regression (SVR). The findings provided a deep insight into large-scale crack deterioration by considering the spatial characteristics and achieved high-resolution crack assessment to support road maintenance decision-making. The spatial machine learning approach and the concept of geocomplexity can be widely applied to address large-scale spatial tasks in road engineering.

Built assets are responsible for 50 per cent of resource consumption, 36 per cent of energy consumption and 38 per cent of energy-related carbon emissions. Therefore, innovative solutions are critical to accurately calculate emissions and identify reduction opportunities. Key objectives of this project include: Digital transformation maturity assessment: To assess the definitions and levels of digital transformation for stakeholders with various needs in the various life-cycle stages of built assets. This will enable organisations to assess their maturing levels when adopting digital twin technologies and support the identification of resources and strategies for industry partners to prepare for each level of digital transformation implementation. Life-cycle assessment integration: To embed life-cycle assessment approaches into digital twin platforms to assess life-cycle emissions of built assets including housing, building and infrastructure assets, and evaluate different design, construction and operation scenarios. Artificial intelligence (AI)-driven decision making: To demonstrate how digital twin and relevant AI technologies, including LiDAR and linked data, can help achieve automatic and informed decision-making in the postconstruction phases of built assets, such as operational and maintenance phases.

Access to urban greenspace is a fundamental requirement in providing critical ecosystem services, improving health and well-being across all ages, fostering social cohesion, and addressing prevalent health disparities in an increasingly urbanised society. Access refers to the availability of urban greenspace, while accessibility indicates the ease of reaching and enjoying these greenspaces and the quality of greenspace services. However, quantitative studies to interpret such difference between accessibility and access are limited. To contribute to this gap, this study developed a Spatial Delta Model (SDM) to quantify the difference between accessibility and access to greenspace and assess its spatial characteristics. The study examines the block-level access, accessibility, and their difference in Perth, Australia, using the SDM with a series of high-resolution greenspace and socio-economic spatial data. Access was calculated as the total greenspace near residential blocks and accessibility was derived using a modified Gaussian two-step floating catchment area (MG2SFCA) approach. Once they were quantified, a set of residential, morphological, and greenspace related factors were utilised to explain the spatial patterns of the difference between accessibility and access using a machine learning geographical detector model. The findings on the measure of the greenspace usability and the user experience further contribute to develop a green city classification system (GCCS), which is useful to informing urban planning and greenspace management.

The U.S. is the world’s second largest contributor to carbon emissions, with its road transport sector being one of the most significant emission sources. However, few studies have been conducted on factors influencing the emissions changes for the U.S. from the perspective of passenger and freight transport. This study aims to evaluate the carbon emissions from the U.S. road passenger and freight transport sectors, using a Logarithmic Mean Divisia Index approach. Emissions from 2008 to 2017 in the U.S. road transport sector are analysed and key findings include: 1) passenger transport contributes over 70% to total transport carbon emissions, with cars and light trucks contributing the highest share; 2) energy intensity and passenger transport intensity are critical for reducing emissions from road passenger transport, and transport structure change is causing shift of emissions between different passenger transport modes; and 3) the most effective strategies to reduce carbon emissions in the road freight transport sector are to improve energy intensity and reduce freight transport intensity. Several policy recommendations regarding reducing energy and transport intensity are proposed. The results and policy recommendations are expected to provide useful references for policy makers to form carbon emissions reduction strategies for the road transport sector.

Machine learning models have seen widespread application in predicting building thermal loads. Yet, these existing models generally predict the thermal load for a single thermal zone or an entire building, overlooking buildings with multiple thermal zones. Moreover, interactions between thermal zones are seldom accounted for these methods. Addressing this gap, our study introduces a graph-based approach that employs a Temporal Graph Attention Network (TGAT) model for building thermal load prediction. This model accommodates both the spatial and temporal dependence across different thermal zones. Specifically, the TGAT model initially utilizes graph attention neural networks (GATs) to identify the spatial relationships between thermal zones. The spatial representation thus obtained is subsequently processed via Gated Recurrent Units (GRUs) to ascertain temporal dependencies. Lastly, a fully connected layer is engaged to decode the representation for the final output. Our model incorporates 11 features as inputs to depict the characteristics of outdoor environment and thermal zones. Testing the proposed TGAT model in a small office virtual environment for heating load prediction revealed that a configuration of four GAT attention heads achieved optimal result for predicting the ensuing hour’s heating load with a six-hour lookback. Additionally, we scrutinized and discussed the efficacy of feature selection and the GAT module. This research thus establishes a novel groundwork for forthcoming studies on zone-based building energy management optimization.