Cooling water intake systems in nuclear power plants are increasingly susceptible to operational disruptions due to marine organism intrusions. Traditional optimization methods based on static numerical simulations and evolutionary algorithms lack adaptability to complex, dynamic operational conditions and uncertainties. This study introduces a cognitive intelligence-integrated optimization framework, combining discrete element method (DEM) and finite element method (FEM) simulations with deep reinforcement learning (DRL), fuzzy analytic hierarchy process, and expert knowledge to optimize filtration net parameters adaptively and robustly.

The cascade filtration system of nuclear power plants has long faced challenges such as high operational uncertainty and complex human decision-making in response to the interception of marine hazards. In order to break through the dual limitations of the lack of real-time performance of traditional physical models and the lack of interpretability of data-driven methods, this study proposes ' CogFiltrate ' -a dynamic decision-making framework based on human-machine collaborative cognition. The framework pioneered the intelligent operation and maintenance paradigm of deep integration of fuzzy reasoning and uncertainty analysis : by constructing a fuzzy evidence network to uniformly quantify the randomness and cognitive uncertainty factors; on this basis, a two-way confidence negotiation mechanism is developed to achieve collaborative optimization of artificial intelligence recommendations and artificial experience in dynamic risk scenarios. After the system is deployed on the full-scale filtering platform of the nuclear power plant, the emergency response efficiency is significantly improved, and the reliability of the framework is successfully verified under extreme disaster conditions. This study provides an example of combining theoretical innovation with engineering practice for the intelligent operation and maintenance of critical infrastructure, and opens up a new path for cognitive intelligence to drive industrial safety.

Under the background of complicated fault detection scenarios and diversified data in the current power distribution grid, it is a significant challenge to deploy high-performance fault detection models for efficient multi-task collaboration processing on lightweight edge intelligent devices. A collaborative multi-teacher distillation for multi-task fault detection in the power distribution grid(CMT-KD) is introduced, which enhances multi-task fault detection by redefining Residual Network (ResNet). We further implement model compression and acceleration by adopting the weight quantization method, while enhancing the performance of the multi-task fault detector via dynamically adapting the knowledge weights from multiple teachers. The proposed method allows for the detection of multiple power distribution grid faults with a single model, significantly reducing the number of model parameters, floating point operations, and training time, while also demonstrating favorable performance in terms of improving accuracy and reducing prediction errors.

Oriented object detection has a very wide range of application scenarios. In recent years, a lot of rotation detectors have been designed to achieve high-performance oriented object detection. Intersection-over-Union (IoU) is the commonly used indicator to evaluate the accuracy of detection performance. Many methods introduce IoU into the bounding box regression loss to achieve the aligned training and evaluation process for better performance. However, in this paper, we demonstrate several drawbacks of rotated IoU loss through both experiments and theoretical derivation: 1) There is a negative correlation between the loss gradient and the angular error. 2) The optimization process of rotated IoU loss suffers from scale sensitivity, which is not conducive to the model convergence. To solve the problems, we propose a Gradient Calibration Loss (GCL) that optimizes the rotated IoU loss via gradient analysis and correction. We construct the optimized gradient in GCL to avoid IoU loss oscillation and scale sensitivity, thereby accelerating model convergence. Models supervised by GCL have a more stable training process, faster convergence, and better performance. Moreover, GCL can be easily introduced into the existing rotation detectors to achieve performance gains without extra inference overhead. Extensive experiments on multiple oriented object detection datasets and models demonstrate the superiority of our method. Our method achieves state-of-the-art performance on the mainstream benchmark datasets. The source code and models are available at https://github.com/ming71/GCL.

How to represent the object more appropriately in oriented object detection is an essential problem to be solved, there are many solutions for the object represented. It is a relatively novel approach to represent objects as a number of sample points useful for both localization and recognition. However, the current point-set-based representation methods do not effectively supervise all points for learning, and the internal information of the convex hull in the point set cannot be effectively learned. Therefore, this letter proposes point set distance (PSD) loss, which learns set-to-set supervision of objects to effectively represent objects. Besides, most of the current sample selection strategies are based on the Intersection over Union (IoU), but these methods cannot comprehensively measure candidate samples quality. To select high-quality point sets, we propose to use the probability distribution of point sets to select the positive samples. Our probabilistic point set sample selection (PPSS) scheme effectively exploits the classification information, regression information, and the distribution characteristics of the point set. Experimental results on remote sensing image datasets including DOTA, DIOR-R, and HRSC2016, demonstrate the proposed method for arbitrary-oriented object detection achieves consistent and substantial improvements.

Most streaming applications, such as multimedia and digital signal processing (DSP) application, are iterative in nature, so pipelined implementation can be introduced into streaming application for high throughput. In this case, as a communication-centric design approach, NoC is capable of solving communication bottleneck incurred by throughput increment. In this paper streaming application with pipelined implementation was mapped onto the NoC architecture by an energy-aware mapping algorithm proposed focusing on the pipelining mechanism. This algorithm performs the task allocation, scheduling and communication scheduling simultaneously and minimizes the energy consumption. The result generated by this algorithm was verified by a cycle-accurate simulator written in SystemC. Experimental results show that 7 times throughput increments can be achieved and energy consumption is also reduced by 12%, compared with one that doesn't involve pipelined implementation.