سال:0 | دوره: | شماره: |تعداد مقالات:6

As delays in construction projects escalate costs, timely project completion stands as a pivotal criterion for success in construction endeavors. Accurate scheduling duration estimates play a vital role in averting additional expenses and mitigating the risk of disputes among employers, contractors, and clients. Experts assert that delays are a common occurrence in the majority of civil engineering projects, emphasizing the critical role of time management in these endeavors. Project scheduling often faces constraints related to activity precedence relationships, project completion time, budget, and various resources like tools, equipment, machinery, or limited human resources. In the realm of construction project control, neural networks emerge as potent and innovative tools. Leveraging machine learning capabilities and analyzing intricate data, these tools contribute significantly to enhancing the management and control of construction processes. This article introduces a model for addressing project scheduling challenges, proposing a novel application of the Long Short-Term Memory (LSTM) neural network. Results demonstrate that LSTM outperforms other Recurrent Neural Networks (RNNs) in handling time series problems. Furthermore, this study advances our understanding of GPT models' application, offering insights into research prospects for implementing GPT models within the construction industry.

The prompt diagnosis of abnormalities in power transformers is of paramount importance. Dissolved Gas Analysis (DGA) serves as an essential and vital tool for identifying faults. This paper introduces a method based on a decision tree (DT) algorithm using DGA to assess the condition of transformer oil samples in two steps: Normal/Faulty and Fault Type. The DTs in this paper were trained using 80% of the 729-sample dataset and evaluated with the remaining 20%. The dataset includes concentrations of five gases dissolved in transformer mineral oil: H2, CH4, C2H2, C2H4, and C2H6. These key features, along with other necessary parameters for learning DTs, contribute to the analysis. By employing two separate and sequential DTs for diagnosing transformer oil samples, the proposed method significantly improves the accuracy of identifying the health status and the type of potential fault. In the test samples, the method achieved a precision of 95.5% for normal state detection and 78.3% for fault type identification.

Production of hydrogen from renewable energy (RE) such as wind and solar power in Iran is of significant importance due to the country's considerable wind and solar potential. For this reason, Tabriz and Jask, being the windiest and sunniest stations in Iran, respectively, have been selected for study. In the present work, the crucial impact of government subsidies on renewable hydrogen production in Iran is evaluated for the first time. Energy-economic-environmental analyses based on data over the 25-year project lifetime were conducted using HOMER V.2.81. The wind and solar electricity generated using four technologies- water electrolysis, steam methane reforming, biological method, and breaking water molecules through thermochemical processes-are converted into hydrogen. Results indicate that in Tabriz and Jask, electricity production from RE costs $0.145/kWh and $0.125/kWh, respectively. The RE system requires a 100kW diesel generator in both stations. The optimal wind turbine size is 20kW, while the solar cells are best at 140kW. The payback period in both stations, compared to the traditional diesel generator system, is less than 2 years. Renewable electricity production in these stations has led to a reduction in pollutants, approximately 161 tons in Tabriz and 263 tons in Jask. The study findings concerning hydrogen production suggest that the most significant solar hydrogen is generated through water electrolysis. Subsidies granted for renewable electricity production have a positive impact on the economies of wind and solar systems, reducing the payback period. Taking into account these subsidies, surplus electricity increases, and solar-produced hydrogen exceeds that generated by wind.

The escalating demand for electrical power propels the evolution of power systems from regional to national scales. However, this expansion introduces challenges such as congestion and transmission bottlenecks, compromising system reliability and stability. High Voltage Direct Current (HVDC) systems, particularly those employing Voltage Source Converter (VSC) technology, offer promising solutions due to their unique control capabilities. This paper proposes the utilization of supplementary control alongside VSC-based HVDC to mitigate low-frequency oscillations and enhance dynamic and transient stability in power systems. Through a comprehensive investigation, including linearization of nonlinear power system equations, the efficacy of different input signals for supplementary control is evaluated using techniques like Singular Value Decomposition (SVD), Relative Gain Array (RGA), and Damping Function. The design of a phase compensator as a supplementary controller, employing generator speed deviation as input, is presented based on the linearized model. Additionally, recognizing the limitations of linear controllers in nonlinear systems, an adaptive neural network-based damping controller is proposed to improve dynamical and transient stability. Results demonstrate the effectiveness of the adaptive neural network controller over the phase compensator, particularly in stabilizing the power system and damping oscillations, underscoring the significance of considering nonlinear dynamics in controller design for HVDC systems.

This paper proposes a novel method for restoring images corrupted by impulse noise. This new method is based on fuzzy cellular automata and an adaptive neural fuzzy inference system (Anfis). The proposed method consists of two phases: identifying and removing salt and pepper noise. In the first phase of the method proposed, salt and pepper noise pixels are identified in two steps. In the first step of the first phase, salt and pepper noises are detected by the average and minimum values of the pixels in the neighborhood of the center pixel. In order to improve the accurate rate of noise detection, pixels that are not detected as noise are re-evaluated by a new algorithm in the second step of the first phase. This new algorithm uses the measure of cosine similarity of Moore's neighborhood values around the central cell, which is based on four types of pixel placement patterns. The state of the pixels is re-evaluated by the fuzzy cellular automata. In the second phase of the proposed method, noisy pixels are restored using Anfis based on Moore neighborhood pixels around the central cell. The method proposed in this paper is evaluated using PSNR and SSIM. Also, the quantitative and qualitative results show that the new method proposed in this paper is robust in different noise levels from 10% to 90%, and image details such as edges are preserved better compared to other filters. .

The classification of Digital Surface Model (DSM) images derived from LiDAR sensors is a challenging task, particularly when distinct ground classes with identical height information must be distinguished. However, DSM images contain valuable spatial information that can be utilized to enhance classification accuracy. This paper proposes a novel strategy, called Multishape Morphological Two-Stage Convolutional Neural Network (MM2CNN), for DSM classification to achieve accurate classified land-cover maps. The proposed method combines the strengths of multishape morphological profiles (MMPs) and a two-stage CNN model as a smart algorithm to effectively discriminate between different land covers from a single-band DSM image. More precisely, the CNN, as a smart method, is used to learn hierarchical rich representations of the data, while the MMPs are used to extract spatial information from the DSM imagery. The approach involves generating MMPs with three structuring elements, training three parallel CNN models, and then stacking and feeding the probability maps to a second-stage CNN to predict the final pixel labels. Experimental results on the Trento benchmark DSM image show that the suggested technique achieves an overall accuracy of 97.32% in a reasonable time, outperforming some other DSM classification methods. The success of the MM2CNN technique demonstrates the potential of integrating MMPs with CNN for precise DSM classification, which has a wide range of applications in environmental investigations.