针对测量系统本身导致的脉冲截断给脉冲高度分析带来的挑战,本研究提出一种复合神经网络模型,用于预测产生了截断的脉冲高度。该模型将长短期记忆模型(Long and Short-term Memory,LSTM)嵌入UNet结构,采用模拟脉冲数据集对模型进行训练...针对测量系统本身导致的脉冲截断给脉冲高度分析带来的挑战,本研究提出一种复合神经网络模型,用于预测产生了截断的脉冲高度。该模型将长短期记忆模型(Long and Short-term Memory,LSTM)嵌入UNet结构,采用模拟脉冲数据集对模型进行训练,使用相对误差指标对模型性能进行评估。结果显示:在对模拟脉冲序列进行脉冲高度估计时,UNet-LSTM模型的平均相对误差约为2.31%,相较于传统的梯形成形算法的平均相对误差降低了1.91%;在粉末铁矿样品和粉末岩石样品的实际测量中,不同截断率的实测脉冲序列也进一步验证了UNet-LSTM模型的脉冲高度估计性能,在两种样品、8组离线脉冲序列的高度估计中得到的平均相对误差为2.36%,表明该模型可以准确估计截断脉冲的高度。展开更多
The continuously growing esthetic awareness for the facial appearance and the spreading of information about the possibilities of adult treatment by public media result in an increase of adult patients which seek orth...The continuously growing esthetic awareness for the facial appearance and the spreading of information about the possibilities of adult treatment by public media result in an increase of adult patients which seek orthodontic treatment to improve their facial esthetics. In general, these patients show such a severe skeletal deformity that it is detectable even by non-experts because of its extraoral manifestation, which is the main motivation for treatment. Because of the nature of these deformities and because of the lacking growth usable for therapy the only promising treatment for these patients is the combined orthodontic-surgical approach. Besides a stable and functional occlusion with physiologic position of the condyle, the goals of treatment are the improvement of the dental and, above all, facial esthetics since the patient judges the success of treatment mostly by the extraoral appearance. The dentofacial appearance must be defined prior to treatment to plan the individual right approach in knowledge of the different treatment possibilities for Angle Class II deformities and thus be able to reach for both sides—patient and orthodontist—satisfying result. With this article, a systematic therapy concept to treat patients with Class II deformities and skeletal deep bite with a short lower face (short face syndrome) under consideration of the soft tissue analysis is presented.展开更多
不同海区的近岸海浪浪高具有明显差异性。针对当前大部分时间序列预测模型缺乏对不同地区(多源)浪高预测的适应性难题,提出了一种基于局部加权回归的多周期趋势分解(Seasonal and Trend decomposition using Loess,STL)和两级融合策略...不同海区的近岸海浪浪高具有明显差异性。针对当前大部分时间序列预测模型缺乏对不同地区(多源)浪高预测的适应性难题,提出了一种基于局部加权回归的多周期趋势分解(Seasonal and Trend decomposition using Loess,STL)和两级融合策略的浪高预测模型,简称为MSTL-WH(Multiple STL-Wave Height)。结合多源近岸浪高时间序列的多周期性、非线性和非平稳性的特点,首先利用周期图法提取多源近岸浪高数据集中的4个主要周期,并基于主要周期进行多次STL分解,将复杂的原始浪高序列分解为周期项、趋势项和余项;然后利用长短期记忆网络(Long Short Term Memory,LSTM)并结合两级融合策略,搭建近岸浪高预测网络;最后使用自注意力机制重新调整权重并输出未来12 h的浪高值。通过与当前主流时间序列预测方法对比,验证了所提方法在多源近岸浪高序列预测中具有较好的实用性和更低的预测误差。展开更多
文摘针对测量系统本身导致的脉冲截断给脉冲高度分析带来的挑战,本研究提出一种复合神经网络模型,用于预测产生了截断的脉冲高度。该模型将长短期记忆模型(Long and Short-term Memory,LSTM)嵌入UNet结构,采用模拟脉冲数据集对模型进行训练,使用相对误差指标对模型性能进行评估。结果显示:在对模拟脉冲序列进行脉冲高度估计时,UNet-LSTM模型的平均相对误差约为2.31%,相较于传统的梯形成形算法的平均相对误差降低了1.91%;在粉末铁矿样品和粉末岩石样品的实际测量中,不同截断率的实测脉冲序列也进一步验证了UNet-LSTM模型的脉冲高度估计性能,在两种样品、8组离线脉冲序列的高度估计中得到的平均相对误差为2.36%,表明该模型可以准确估计截断脉冲的高度。
文摘The continuously growing esthetic awareness for the facial appearance and the spreading of information about the possibilities of adult treatment by public media result in an increase of adult patients which seek orthodontic treatment to improve their facial esthetics. In general, these patients show such a severe skeletal deformity that it is detectable even by non-experts because of its extraoral manifestation, which is the main motivation for treatment. Because of the nature of these deformities and because of the lacking growth usable for therapy the only promising treatment for these patients is the combined orthodontic-surgical approach. Besides a stable and functional occlusion with physiologic position of the condyle, the goals of treatment are the improvement of the dental and, above all, facial esthetics since the patient judges the success of treatment mostly by the extraoral appearance. The dentofacial appearance must be defined prior to treatment to plan the individual right approach in knowledge of the different treatment possibilities for Angle Class II deformities and thus be able to reach for both sides—patient and orthodontist—satisfying result. With this article, a systematic therapy concept to treat patients with Class II deformities and skeletal deep bite with a short lower face (short face syndrome) under consideration of the soft tissue analysis is presented.
文摘不同海区的近岸海浪浪高具有明显差异性。针对当前大部分时间序列预测模型缺乏对不同地区(多源)浪高预测的适应性难题,提出了一种基于局部加权回归的多周期趋势分解(Seasonal and Trend decomposition using Loess,STL)和两级融合策略的浪高预测模型,简称为MSTL-WH(Multiple STL-Wave Height)。结合多源近岸浪高时间序列的多周期性、非线性和非平稳性的特点,首先利用周期图法提取多源近岸浪高数据集中的4个主要周期,并基于主要周期进行多次STL分解,将复杂的原始浪高序列分解为周期项、趋势项和余项;然后利用长短期记忆网络(Long Short Term Memory,LSTM)并结合两级融合策略,搭建近岸浪高预测网络;最后使用自注意力机制重新调整权重并输出未来12 h的浪高值。通过与当前主流时间序列预测方法对比,验证了所提方法在多源近岸浪高序列预测中具有较好的实用性和更低的预测误差。
