Abstract:This paper aims to propose a novel deep learning-integrated framework for deriving reliable simulation input models through incorporating multi-source information. The framework sources and extracts multisource data generated from construction operations, which provides rich information for input modeling. The framework implements Bayesian deep neural networks to facilitate the purpose of incorporating richer information in input modeling. A case study on road paving operation is performed to test the feasibility and applicability of the proposed framework. Overall, this research enhances input modeling by deriving detailed input models, thereby, augmenting the decision-making processes in construction operations. This research also sheds lights on prompting data-driven simulation through incorporating machine learning techniques.
What problem does this paper attempt to address?
The problem that this paper attempts to solve is: in construction simulation, how to improve the reliability of input models by integrating multi - source information, thereby enhancing the decision - making process. Specifically, traditional input models are usually based on only a single type of input data (such as weather conditions), which may lead to the model being unable to accurately reflect the complexity and uncertainty in actual situations. For example, in truck transportation time modeling, if the influence of weather conditions is ignored, it may lead to sampling random variables from non - ideal conditions, thus reducing the accuracy of the model.
To solve these problems, the author proposes a new framework based on Bayesian Deep Neural Networks (BDNN). This framework aims to improve input modeling in the following ways:
1. **Integrating multi - source information**: Use data from different sources (such as weather, traffic conditions, mechanical equipment status, etc.) to provide more abundant information for input modeling.
2. **Handling uncertainty**: Through BDNN, it can not only integrate multiple inputs, but also quantify the uncertainty in the output, especially the aleatoric uncertainty that represents the inherent randomness. This is very important for reflecting the inevitable random changes in the construction process.
Through this method, researchers hope to generate more detailed and reliable input models in construction simulation, and then improve the quality of decisions made based on these models. In addition, this study also shows how to promote the development of data - driven simulation methods through machine - learning techniques.
### Formula Explanation
In Bayesian Deep Neural Networks, the probability distribution of the output can be represented by the mean \(\mu\) and variance \(\sigma^2\):
\[ p(y|x)=\mathcal{N}(\mu(x),\sigma^2(x)) \]
where \(y\) is the output and \(x\) is the input feature vector. Here, \(\mu(x)\) and \(\sigma^2(x)\) represent the mean and variance of the predicted value respectively, which are determined by the network parameters.
In this way, BDNN can not only give deterministic prediction results, but also provide the confidence interval of the prediction results, so as to better reflect uncertainty and risk.