Abstract. With the rapid advancement of artificial intelligence (AI) in industrial applications, this paper proposes a novel intelligent optimization method based on a hybrid data-model-driven approach to address traditional challenges in loader excavation trajectory optimization, including high data acquisition costs and computational complexity. First, a kinematic model of the loader working mechanism is established using the Denavit-Hartenberg (D-H) method to enable workspace mapping. Subsequently, an EDEM-RecurDyn co-simulation platform is developed to efficiently generate training datasets. Then, a linear trajectory ensuring the required excavation volume is then planned to define the end-effector pose constraints. The core contribution lies in the deep integration of physics-based modeling and data-driven intelligence. Specifically, an AI-based prediction model is developed to evaluate trajectory performance, and a corresponding fitness function is formulated. The IVY optimization algorithm is then employed for efficient autonomous optimization, with the optimal trajectory generated via spline fitting. Experimental results demonstrate that, compared with conventional metaheuristic algorithms, the proposed method improves convergence speed by approximately 88% and fitness value by approximately 26%. This research overcomes the limitations of purely data-driven or model-driven approaches, providing an efficient and cost-effective solution for intelligent construction machinery operation, and demonstrates the potential of AI technology in industrial trajectory optimization.
Volume 1, Issue 1
Published online articles for EII Volume 1, Issue 1. Each title below links to a dedicated article page with the full abstract and a downloadable PDF, while this issue page keeps the table of contents and article summaries visible in plain HTML.
Abstract. Under the background of "five aspects of education in parallel", integrating aesthetic education into engineering training is crucial for cultivating all-round engineering innovation talents. The engineering training center, as the core carrier of practical teaching, provides natural support for the implementation of aesthetic education. This article takes the aesthetic education teaching in university engineering training centers as the research object, explains its core connotation and educational value, analyzes the current predicaments, and combines the practical experience of Jilin University's engineering training center to explore feasible paths from five aspects: concept innovation, curriculum reconfiguration, faculty building, evaluation optimization, and atmosphere creation. The research shows that focusing on engineering practice to explore aesthetic elements and constructing a collaborative closed-loop aesthetic education system is the key to enhancing the humanistic literacy and innovation ability of engineering talents, and can provide reference for related teaching reforms in universities.
Abstract. Accurately predicting the sound insulation performance of a system is of great significance for the development of automotive noise, vibration, and harshness (NVH) performance. However, traditional numerical simulation methods are computationally expensive, while purely data-driven models often suffer from stability issues and lack of physical consistency when dealing with complex structures. To address this issue, this paper proposes a sound intensity-guided gated recurrent unit (SI-GRU) model for predicting the sound insulation of automotive floor systems by embedding sound intensity, a key indicator of sound insulation performance, as prior knowledge into the gated recurrent unit (GRU) network architecture. This approach enhances the model's stability and robustness during the frequency-domain learning process. Experimental results indicate that various deep learning models can effectively capture the overall trend of sound insulation performance as a function of frequency. Compared to benchmark models such as GRU, LSTM, and 1D-CNN, the proposed SI-GRU achieves superior results across evaluation metrics including RMSE, MAE, and MedAE, with a 2.5% reduction in RMSE prediction error relative to the GRU network. In the mid-to-high frequency range, where NVH performance is most critical, the model can stably control the relative error within approximately 6%-7%. The results demonstrate that combining domain knowledge with data-driven models can effectively improve the reliability and engineering applicability of sound insulation prediction, providing an efficient predictive method to replace high-cost numerical simulations.
Abstract. Although Maritime Autonomous Surface Ships (MASS) can detect and avoid collisions autonomously, they should be supervised or intervened by shore control center operators (SCCOs) if necessary. SCCOs may operate with a dynamic level of human control (LoHC) and their mental workload (MWL) increases with the LoHC. In addition, SCCOs represent the ultimate safety barrier for collisions, while a high MWL (HMWL) may cause human errors and increase collision risks. This paper proposes a fault-tree-based collision risk analysis method that explicitly models SCCOs' HMWL, introduces an HMWL-performance dependency gate, and analyzes critical events under dynamic LoHC conditions.