Early fault detection is crucial to avoid catastrophic flight accidents caused by unplanned downtime of equipment. Aimed at the shortcomings of feature selection and early fault detection, this paper proposes a method...
详细信息
Early fault detection is crucial to avoid catastrophic flight accidents caused by unplanned downtime of equipment. Aimed at the shortcomings of feature selection and early fault detection, this paper proposes a method combining two -stage multi -view featureoptimization and improved support vector data description for aeroengine bearing early fault detection. First, a comprehensive evaluation index is constructed via combining monotonicity, correlation, robustness index and entropy weight method to be used for initial screening sensitive degradation features from multi -view candidate degradation feature. After that, the hierarchical clustering algorithm and local preserving projection algorithm is introduced to realize sensitive degradation features clustering and dimensionality reduction. On this basis, the Euclidean distance sample reduction and the sparrow search algorithm is utilized to complete the training sample selection and model parameter optimization of different clusters and further use the verification samples to filter the optimal features and model. Finally, based on the optimal degradation features and models, a fusion performance degradation index is built to achieve intelligent detection of early fault. Comparisons with some existing fault detection methods, the proposed method improves degradation trend sensibility and demonstrates potential of good early fault detection engineering applications.
Most traditional recommendation systems(RS) improve the model structure to obtain valid user behavior representation and make better recommendations. However, limited by the scale of the model, the model cannot accomm...
详细信息
ISBN:
(数字)9789887581581
ISBN:
(纸本)9798350366907
Most traditional recommendation systems(RS) improve the model structure to obtain valid user behavior representation and make better recommendations. However, limited by the scale of the model, the model cannot accommodate an extended range of user historical data and cannot effectively extract long-term preferences. Reinforcement learning(RL) aims to maximize the cumulative return. Introducing RL can smoothly solve the long-term and short-term profit balance. However, there are specific problems with the direct combination of the two. Training RS that relies on offline user data will cause the strategy to suffer from an overestimation of action and distribution shift. Unlike the traditional RL environment, the ample space of the RS will also cause problems of insufficient exploration and sparse rewards, making the policy more difficult to learn. In response to the above issues, this work proposes an offline RL two-stageoptimization framework. The framework consists of an offline RL method based on unknown behavior to alleviate the problem of action overestimation and a two-stage feature optimization training method to reduce the impact of large spaces. Experimental results prove that the method proposed in this article can effectively improve the long-term recommendation ability of the model and more effectively improve the model performance than the general offline RL algorithm.
暂无评论