Search

Safety is the foremost important issue on the construction site. Wearing a safety helmet is a compulsory issue for every individual in the construction area, which greatly reduces the injuries and deaths. However, though workers are aware of the dangers associated with not wearing safety helmets, many of them may forget to wear helmet at work, which leads to significant potential security issues. To solve this problem, we have developed an automatic computer-vision approach based on Convolutional Neural Network (YOLO) to detect wearing condition. We create a safety helmet image dataset of people working on construction sites. The corresponding images are collected and labeled and are used to train and test our model. The YOLO based model is adopted and parameters are well tuned. The precision of the pro-posed model is 78.3% and the accuracy rate is 20 ms. The results demonstrate that the proposed model is an effective method and comparatively fast for the recognition and localization in real-time helmet detection.

In order to meet the required heat supply of each household, save energy, as well as avoid wasting excessive heat, it is necessary to accurately predict the thermal load (TL) in a community. In this paper, we focus on cross-floor prediction over a week for three thermal users located on different floors. Three machine learning methods including Back Propagation Neural Network (BPNN), Extreme Learning Machine (ELM), and Support Vector Machines (SVM) are applied for indoor temperature (Tin) and thermal load prediction. The experimental results show that Extreme Learning Machine outperforms the other two methods in terms of thermal load prediction.

Plate-fin heat exchanger (PFHE) is a popular and featured compact heat exchanger, which transfers heat at relatively low-temperature differences and has a high heat transfer surface area to volume ratio. Due to the compact size and light weight, it has been widely used in various engineering areas such as energy, transportation and aerospace. The optimal design of PFHE aims to minimize the economic cost or maximize the efficiency, formulating a mixed integer nonlinear optimization problem which challenges the optimization tools. In this paper, the number of entropy generation units is formulated as the objective functions. Given the strong non-linear behaviors, a recent proposed meta-heuristic algorithm named sine cosine algorithm (SCA) is adopted to solve the problem of the design and optimization of the PFHE and compared with the results of several classical and new algorithms. Test results show that the sine cosine algorithm is the a competitive solver for PFHE optimal design problems.

Environment of construction site is becoming more complicated and risky than ever before, due to the rapidly development of society. The traditional site management system is facing the challenges of manual supervision negligence as well as the inflexible attendance system, which cannot guarantee the life safety and legitimate interests of all employees. To address the above problem and reduce the risks, many construction sites utilize intelligent approaches such as effective safety helmet detection and face recognition. This paper propose a hybrid approach combining the popular YOLOv3 and Facenet model to detect the safety helmet wearing of construction workers and help them with attendance checking through camera simultaneously. At first, this method apply YOLOv3 to implement safety helmet detection. Then, Facenet is used to achieve face recognition with face detected by MTCNN model. Finally, combined with the above two modules, helmet detection and personnel information identification can be realized in site supervision system. The experimental results show the effectiveness of the proposed combined approach. As a result, database established by video and image process results can ensure the reasonable salary payment of construction workers, further improving the safety assurance measures and the management efficiency of the construction site.

In data mining and machine learning area, features targeting and selection are crucial topics in the real world applications. Unfortunately, massive redundant or unrelated features significantly deteriorate the performance of learning algorithm. This paper presents a novel classification model which combined grey wolf optimizer (GWO) and spectral regression discriminant analysis (SRDA) for selecting the most appropriate features. The GWO algorithm is adopted to iteratively update the currently location of the grey wolf population, while the classification algorithm called SRDA is employed to measure the quality of the selected subset of features. The proposed method is compared with genetic algorithm (GA), Jaya, and three recent proposed Rao algorithms also with SRDA as the classifier over a set of UCI machine learning data repository. The experimental results show that the proposed method achieves the lower classification error rate than that of GA and other corresponding methods generally.

This paper proposed a Big Data platform solution for city power grid status estimation, with an in-memory distributed computing frame for power grid heterogeneous data analysis. A big data management system is proposed for managing high volume, heterogeneous and multi-mode power data. In addition, a high efficiency data-computing engine, a parallel computing model for power engineering and a 3D display system are consisted. Based on the platform, three real-system application cases are given in order to illustrate the practicability of the proposed platform, including high efficiency power grid load shedding computing and analysis, voltage sag association rule mining with 3D display, and real-time dynamic evaluation of operation status of wind farm units. As a conclusion, big data platform is a key technology to sufficiently realize the potential advantages of power data resources, and it can provide solid technological bases for the development of modern Smart Grid.