【摘要】 在现代制造业中,精确预测刀具磨损状态对于提升生产效率和产品质量至关重要。然而,传统磨损预测方法往往难以有效处理高频、复杂的时序数据,无法满足实际应用需求。为此,本文提出了一种基于Informer与时序卷积网络(temporal convolutional network, TCN)的混合模型,以提高刀具磨损状态的识别精度。该模型融合了TCN的长序列依赖建模能力与Informer处理大规模不规则时序数据的优势,能够在兼顾低计算成本的同时,实现对多时间步长磨损状态的准确预测。在真实加工数据集上的实验结果表明,与传统模型相比,本文方法在磨损状态分类任务中具有更高的识别精度和更强的泛化能力。更多还原

【Abstract】 In modern manufacturing, accurately predicting tool wear status is crucial for improving production efficiency and product quality. However, traditional wear prediction methods struggle to handle high-frequency and complex time-series data, making them inadequate for practical applications. Thus, the paper proposes a hybrid model based on Informer and temporal convolutional network(TCN) to enhance the recognition accuracy of tool wear states. The proposed model integrates TCN’s ability to model long-sequence dependencies with Informer’s strength in handling large-scale irregular time-series data, enabling accurate predictions of multi-step wear states with low computational cost. Experiments results on a real-world machining dataset, demonstrate that compared to traditional models, the proposed method achieves higher recognition accuracy and better generalization in tool wear state classification tasks.更多还原