基于预训练模型和多视角循环神经网络的电力文本匹配模型

针对传统方法未能考虑词向量的动态性及句子间交互不充分等问题,提出基于BERT预训练模型及多视角循环神经网络的文本匹配模型。通过BERT-whitening方法对BERT输出的句向量进行线性变换优化,并利用多视角循环神经网络将两句子不同位置的BERT动态词向量进行双向交互计算;将句向量与词粒度交互向量进行融合后计算结果。实验结果表明,提出的模型相较于对比模型有明显性能提升,实用性良好。     

超高压气井井口节流工艺选择与优化

超高压气井的节流系统对气井的生命周期起到至关重要的作用,而合理的节流工艺,不但可以延长气井开采寿命,还可降低成本。对国内超高压气井使用的节流工艺进行了介绍,并基于计算流体力学对川西北气矿产气井不同节流工艺内流场进行仿真,运用流场分析软件对单级节流和两级节流工艺系统内部的压力、流速等问题进行了分析,确定了不同节流工艺系统合理CV值及开度,得到了不同节流工艺系统的最高流速,提出了现场不同节流工艺下的节流参数选择及使用建议。研究表明,3种节流工艺均可满足节流要求和现场生产情况,其中,单级节流工艺简单,但流速均在850 m/s以上,阀门冲蚀严重;一级固定二级可调节流的最大流速在550~600 m/s,低于单级节流,但最大产量受限;两级可调节流的最大流速与一级固定二级可调节流相仿,可实现按需调产,但成本相对较高。     

计及位置不确定的喷雾机喷杆系统仿形姿态渐近跟踪控制北大核心CSCD

针对喷雾机喷杆仿形系统中同时存在负载变化、未建模不确定项、物理参数摄动以及外部干扰等问题,提出了一种基于小波网络逼近的具有自适应性和鲁棒性的反步控制方法.首先,将含有不确定、未知和非线性项的喷杆仿形系统建立为完整的数学模型,将其等价转化为具有严格反馈的状态空间形式;其次,采用设计的小波基元去构造神经网络,在满足最优误差有界条件下逼近反步法中虚拟等效控制部分,选取自适应更新律估计系统中存在的未知参数,引入鲁棒补偿项减小复合干扰对系统的不利影响,降低了输入指令信号的阶次要求;最后,通过构造合适的Lyapunov函数,应用稳定性理论证明了闭环系统位置跟踪误差渐近收敛到原点.仿真结果表明,所提控制方法可实现喷雾机喷杆位置姿态快速升降机动调整,有效地增强了喷杆系统的鲁棒稳定性和控制精度.     

Anti-Controlling Codimension-2 Bifurcation of Discrete Dynamical Systems in 1 : 2 Resonance北大核心CSCD

从分岔反控制的角度设计了一套非线性反馈控制策略,来实现离散动力系统1∶2共振情形下余维二分岔的各种分岔解。首先,针对传统分岔准则在确定高余维分岔点时存在的局限性,建立了一个1∶2共振情形下的余维二分岔的新显式准则,基于这个显式准则通过设计线性控制增益来确保此类余维二分岔的存在性。然后,推导了1∶2共振的中心流形,并基于范式方法通过设计非线性控制增益,分析了1∶2共振情形下余维二分岔解的类型和稳定性。最后,以一个Arneodo-Coullet-Tresser映射为例,在指定的参数点处通过控制实现了具有1∶2共振分岔特性的各种分岔解,进一步验证了理论分析。     

Preservation of transform orders under a semiparametric model and its mixture

In this paper, sufficient conditions for preservation of several transform orders under a typical family of semiparametric distributions are made. The preservation properties are developed to compare mixture semiparametric distributions. Possible applications of the achieved results to compare scale family of distributions and also compare coherent systems with dependent but identical components and series and parallel systems with randomized number of components are provided.     

Traveling waves for a four-compartment lattice epidemic system with exposed class and standard incidence

This paper is considering the problem of traveling wave solutions (TWS) for a susceptible-exposed-infectious-recovered (SEIR) epidemic model with discrete diffusion. The threshold condition for the existence and nonexistence of TWS is obtained. More specifically, such kind of solutions are governed by the threshold number ?₀. We can find a critical wave speed c^? if ?₀ > 1, by employing the Schauder's fixed point theorem, limiting argument and two-sided Laplace transform, we confirm that there exists TWS for c > c^?, while there exists no TWS for c < c^?. We also obtain the nonexistence of TWS for ?₀ ≤ 1. At last, we give some biological explanations from the epidemiological perspective.     

A Note on the Indirect Controls for a Coupled System of Wave Equations

By a procedure of successive projections, the authors decompose a coupled system of wave equations into a sequence of sub-systems. Then, they can clarify the indirect controls and the total number of controls. Moreover, the authors give a uniqueness theorem of solution to the system of wave equations under Kalman’s rank condition.     

Homology and cohomology intersection numbers of GKZ systems

We describe the homology intersection form associated to regular holonomic GKZ systems in terms of the combinatorics of regular triangulations. Combining this result with the twisted period relation, we obtain a formula of cohomology intersection numbers in terms of a Laurent series. We show that the cohomology intersection number depends rationally on the parameters. We also prove a conjecture of F. Beukers and C. Verschoor on the signature of the monodromy invariant hermitian form.     

Sampling Consensus of 2nd-Order Multi-Agent Systems Based on Time-Varying Topology北大核心CSCD

基于速度一致位移差保持不变的一致性概念,研究了二阶多智能体系统在时变拓扑下的采样一致性问题。首先,引入虚拟领导者,将具有时变拓扑结构的多智能体系统的采样一致性问题转换为误差系统的采样控制稳定性问题。其次,通过预估采样误差,研究采样误差对系统达到一致性的影响。最后,应用Lyapunov稳定性理论,分析所构造的误差系统的稳定性,并给出该误差系统最终稳定的充分条件。数值仿真结果验证了理论分析的有效性和正确性。     

Adaptive Constrained Control of Unknown Strict Feedback Nonlinear Systems With Dead Zone Input北大核心CSCD

研究了具有死区输入的预设约束未知高阶严格反馈非线性系统的控制问题,提出了一种基于免疫函数的自抗扰预设漏斗约束自适应控制策略。首先,针对系统内部的未知问题,采用免疫函数与扩张状态观测器结合对系统内部未知项进行观测;其次,通过Lyapunov方法与漏斗控制相结合设计控制器,使得跟踪误差能够维持在预先设定的漏斗约束范围内;同时,利用双曲正切函数速率变化快这一特性设计自适应控制律,引入指令滤波器避免反步法中重复求导问题,分析证明了闭环系统所有信号的有界性。仿真实例表明了控制方法的有效性。