摘要：【中文】本发明公开了一种基于神经网络的超声背散射零差K模型参数估算方法，是根据已知的超声背散射零差K模型的参数μ和k，利用蒙特卡洛仿真产生零差K模型独立同分布的超声背向散射信号样本；基于这些样本计算特征参数；利用计算得到的特征参数和已知的μ和k，训练得到反向传播神经网络模型；对于待测的超声背向散射信号样本，首先计算其特征参数，再将特征参数输入训练得到的反向传播神经网络模型，就可得到待测样本的零差K模型参数μ和k的估算结果。本发明方法相比现有技术，具有更高的估算精度和更快的估算速度。 【EN】The invention discloses an ultrasonic backscattering homodyne K model parameter estimation method based on a neural network, which is characterized in that according to parameters mu and K of a known ultrasonic backscattering homodyne K model, Monte Carlo simulation is utilized to generate ultrasonic backscattering signal samples with homodyne K models independently distributed; calculating feature parameters based on the samples; training to obtain a back propagation neural network model by using the characteristic parameters obtained by calculation and the known mu and k; for an ultrasonic backscatter signal sample to be detected, firstly, the characteristic parameters of the ultrasonic backscatter signal sample are calculated, and then the characteristic parameters are input into a back propagation neural network model obtained through training, so that the estimation results of homodyne K model parameters mu and K of the sample to be detected can be obtained. Compared with the prior art, the method has higher estimation precision and higher estimation speed.

摘要：【中文】本发明公开了一种Web中面向服务环境中的基于张量的信任评估方法，基于张量机制，服务消费者根据服务提供者提供的的历史服务报告，对服务提供者的服务提供表现进行建模。从服务提供者接收到所有相关的历史服务报告后，服务消费者为所有潜在服务提供者的所有历史服务构建稀疏张量模型。基于CP分解的稀疏张量补全方法补全张量中的缺失条目，从而服务消费者可以在当前时间段中评估所有潜在服务提供者在所需求服务上的潜在表现，并选择表现最好的服务提供者来提供所需服务。该发明可以在开放和动态环境中有效且准确地实现信任评估，尤其是当提供者的相关参考报告稀疏时。 【EN】The invention discloses a trust evaluation method based on tensor in a service-oriented environment in Web, which is based on a tensor mechanism and enables a service consumer to model service providing performance of a service provider according to a historical service report provided by the service provider. After receiving all relevant historical service reports from the service providers, the service consumer builds a sparse tensor model for all historical services of all potential service providers. The sparse tensor completion method based on CP decomposition completes the missing entries in the tensor, so that the service consumer can evaluate the potential performances of all potential service providers on the required service in the current time period and select the service provider with the best performance to provide the required service. The invention can effectively and accurately realize trust evaluation in an open and dynamic environment, especially when the related reference report of a provider is sparse.

摘要：【中文】本发明公开了基于深度学习的端到端乐谱音符识别方法，该方法一共分为三步：(1)数据预处理：需要从MuseScore中下载对应的数据集，重新编码音高和时值标签。(2)数据增强：对重新编码后的乐谱数据进行数据增强，本发明提出了4种不同的增强方法。(3)端到端模型：应用于端到端乐谱音符识别的深度卷积神经网络模型，将增强后的数据输入的模型，模型的输出为音符时值和音高。本发明在于针对打印体乐谱提出一个基于深度学习的乐谱音符识别模型，即输入整张乐谱图像到该模型，直接输出乐谱上音符的时值和音高，该模型完全端到端，能够精准识别多声部乐谱图像。 【EN】The invention discloses an end-to-end music score note identification method based on deep learning, which comprises the following three steps: (1) data preprocessing: the corresponding data set needs to be downloaded from musesecore and the pitch and duration tags re-encoded. (2) Data enhancement: the invention provides 4 different enhancing methods for data enhancement of the recoded music score data. (3) End-to-end model: the deep convolution neural network model is applied to end-to-end music score note identification, the enhanced data is input into the model, and the output of the model is a note duration and a pitch. The invention aims at the music score of a printed body to provide a music score note recognition model based on deep learning, namely, the whole music score image is input to the model, the duration and pitch of notes on the music score are directly output, the model is completely end-to-end, and the polyphonic music score image can be accurately recognized.

摘要：【中文】本发明公开了一种嵌入式双面互连功率模块封装结构和制作方法，由IGBT功率芯片，二极管芯片，上DBC基板，下DBC基板，中间转接板，介电填充层，焊料层，再布线层，过孔导电金属和功率端子组成。本发明通过焊料层将IGBT功率芯片及二极管芯片和下DBC基板连接。同时在中间转接板上制作矩形框架，并通过填充介电材料，将IGBT功率芯片和二极管芯片嵌入在中间转接板内。芯片和转接板的上表面覆有导电金属层，中间转接板的上下表面分别和上下DBC基板互连，各功率端子分别从上下DBC基板的导电覆铜层引出，得到嵌入式双面互连功率模块。该发明可以实现IGBT功率模块的双面散热，提高了散热效率。而且不使用键合线，减小了模块的寄生电感。 【EN】The invention discloses an embedded double-sided interconnection power module packaging structure and a manufacturing method thereof. The IGBT power chip and the diode chip are connected with the lower DBC substrate through the solder layer. And meanwhile, a rectangular frame is manufactured on the middle adapter plate, and the IGBT power chip and the diode chip are embedded in the middle adapter plate by filling dielectric materials. And conductive metal layers are coated on the upper surfaces of the chip and the adapter plate, the upper surface and the lower surface of the middle adapter plate are respectively interconnected with the upper DBC substrate and the lower DBC substrate, and each power terminal is respectively led out from the conductive copper-clad layers of the upper DBC substrate and the lower DBC substrate to obtain the embedded double-sided interconnection power module. The invention can realize double-sided heat dissipation of the IGBT power module and improve the heat dissipation efficiency. And moreover, no bonding wire is used, so that the parasitic inductance of the module is reduced.

摘要：【中文】本发明公开了一种基于数控负载信息的机床运动部件接触刚度退化率监测方法，为了提高加工精度和机床寿命有必要对机床运动部件的接触刚度进行监测。为了方便快捷监测运动部件接触刚度的退化过程，本发明通过定期监测机床空载运行在标准的测试程序下的负载信息，确定驱动电机驱动运动部件匀速运动所需克服的摩擦力矩，进而根据摩擦力矩与接触刚度的关系计算运动部件接触刚度退化率。该监测方法可以在不影响机床正常工作的情况下准确便捷地评估运动部件的磨损和接触刚度退化趋势，并且监测成本低、简单易操作。 【EN】The invention discloses a method for monitoring the degradation rate of contact rigidity of a moving part of a machine tool based on numerical control load information, which is necessary to monitor the contact rigidity of the moving part of the machine tool in order to improve the processing precision and prolong the service life of the machine tool. In order to conveniently and rapidly monitor the degradation process of the contact rigidity of the moving part, the invention determines the friction torque which needs to be overcome by driving the moving part to move at a constant speed by the driving motor by regularly monitoring the load information of the machine tool in no-load operation under a standard test program, and then calculates the degradation rate of the contact rigidity of the moving part according to the relationship between the friction torque and the contact rigidity. The monitoring method can accurately and conveniently evaluate the wear and contact rigidity degradation trend of the moving part under the condition of not influencing the normal work of the machine tool, and has the advantages of low monitoring cost, simplicity and easy operation.

摘要：【中文】本发明提出了一种在边缘场景下的节省功耗的调度算法，优化了边缘场景下的设备的续航。本发明区别于其他边缘计算资源调度的特征是加入了边缘服务器的休眠机制，通过任务调度进一步降低边缘服务器的功耗。本方法将边缘计算服务器分为主服务器和从服务器，主服务器用来接收数据和处理数据，从服务器用来处理数据，当从服务器无数据处理时，进入休眠状态。方法提出了两种策略：激进的策略：用于在优先保障计算能力的情况下防止延迟增长；保守的策略，用于实现可容忍延迟状态下的最低功耗。最终，通过多个服务器的负载均衡实现了优化延迟和功耗的目标。 【EN】The invention provides a power consumption-saving scheduling algorithm in an edge scene, and the endurance of equipment in the edge scene is optimized. The method is different from other edge computing resource scheduling in that a dormancy mechanism of an edge server is added, and the power consumption of the edge server is further reduced through task scheduling. The method divides the edge computing server into a main server and a slave server, wherein the main server is used for receiving data and processing the data, the slave server is used for processing the data, and when the slave server has no data processing, the slave server enters a dormant state. The method proposes two strategies: radical strategies: for preventing delay growth in the case of priority for computing power; a conservative strategy for achieving the lowest power consumption in a tolerable delay state. Finally, the goal of optimizing latency and power consumption is achieved through load balancing of multiple servers.

摘要：【中文】本发明公开了一种自动识别柴油喷雾破碎过程自动实现喷雾连续计算的方法，从而自动判断柴油喷射过程中液膜破碎到液滴这个过程，实现喷雾连续计算的SD‑ELSA算法。SD‑ELSA算法以计算网格中液相的球形度和平均粒子直径为判断依据，使用拉格朗日方法计算流场中的粒子，使用欧拉模型计算全流场内网格节点信息中属于连续液相的网格，对符合两种判据的欧拉团块，将其转化为粒子应用拉格朗日法重新进行计算。该方法实现了欧拉‑拉格朗日模型的动态转化耦合，能够自动获得流场离散相与连续相的完整信息并进行迭代计算，获得柴油射流的液柱、一次破碎以及二次破碎三个阶段，完整表现柴油喷雾过程。 【EN】The invention discloses a method for automatically identifying the diesel oil spray crushing process and automatically realizing continuous spray calculation, thereby automatically judging the process that a liquid film is crushed into liquid drops in the diesel oil spraying process and realizing the SD-ELSA algorithm of continuous spray calculation. The SD-ELSA algorithm takes the sphericity and the average particle diameter of a liquid phase in a calculation grid as judgment basis, uses a Lagrange method to calculate particles in a flow field, uses an Euler model to calculate the grid belonging to a continuous liquid phase in grid node information in a full flow field, and converts Euler agglomerates meeting two criteria into particles and applies the Lagrange method to calculate again. The method realizes dynamic conversion coupling of the Euler-Lagrange model, can automatically obtain complete information of a flow field discrete phase and a continuous phase, performs iterative calculation, obtains three stages of liquid column, primary crushing and secondary crushing of diesel jet, and completely expresses a diesel spraying process.

摘要：【中文】一种两步扩散法制备高性能双主相烧结混合稀土铁硼磁体的方法，属于稀土磁性材料制备技术领域。两种主相合金的成分分别为RE‑Fe‑B(RE为Nd或Pr)和(Nd,MM)‑Fe‑B，MM为混合稀土。本发明工艺，先以PrHoFe合金速凝片为扩散源，在(Nd,MM)‑Fe‑B氢破碎的粉末颗粒表面均匀地包覆一层富PrHo的化合物，利用Pr₂Fe₁₄B、Ho₂Fe₁₄B较高的各向异性场来提高矫顽力；然后以ZrCu合金速凝片为扩散源，在经第一步扩散后的粉末颗粒表面均匀地包覆一层富Zr层，阻止烧结过程中含MM主相晶粒的长大以及抑制与双主相中另一主相之间的互扩散，从而获得高矫顽力。 【EN】A method for preparing a high-performance double-main-phase sintered mixed rare earth iron boron magnet by a two-step diffusion method belongs to the technical field of rare earth magnetic material preparation. The two main phase alloys respectively comprise RE-Fe-B (RE is Nd or Pr) and (Nd, MM) -Fe-B, and MM is mixed rare earth. The process of the invention comprises the steps of taking a PrHoFe alloy rapid hardening sheet as a diffusion source, uniformly coating a layer of PrHo-rich compound on the surface of (Nd, MM) -Fe-B hydrogen crushed powder particles, and utilizing Pr₂Fe₁₄B、Ho₂Fe₁₄B, the coercivity is improved by a higher anisotropy field; and then, taking a ZrCu alloy rapid hardening sheet as a diffusion source, uniformly coating a Zr-rich layer on the surface of the powder particles subjected to the first-step diffusion, preventing the growth of MM-containing main phase grains in the sintering process and inhibiting mutual diffusion with the other main phase in the double main phases, thereby obtaining high coercivity.

摘要：【中文】本发明公开一种基于混合长短时记忆神经网络的水质指标预测方法，首先，将获取到的水质指标历史数据依照时间序列进行排序，并对该水质历史数据采用SG滤波平滑预处理。然后，再进行水质数据的归一化处理，将水质时间序列数据按照预设的滑动窗口大小划分为多个子序列作为特征序列，也就是转为有监督的数据后，输入基于编码‑解码器的长短时记忆ED‑LSTM神经网络模型，预测未来多个时间点的水质指标值，最终获取精准度较高的水质指标预测结果。 【EN】The invention discloses a water quality index prediction method based on a mixed long-time memory neural network. Then, normalization processing of the water quality data is carried out, the water quality time series data are divided into a plurality of subsequences according to the size of a preset sliding window and serve as characteristic sequences, namely after the characteristic sequences are converted into supervised data, an ED-LSTM neural network model based on a coder-decoder is input, water quality index values of a plurality of time points in the future are predicted, and finally a water quality index prediction result with high accuracy is obtained.

摘要：【中文】本发明涉及一种基于深度神经网络的电子产品异步自适应价值评估方法，用于解决二手电子产品自动定价的问题。具体包括估计产品本身价值的估价模块和为应对市场变化进行定价调整的自适应调价模块，自适应调价模块使用双深度Q网络，其模型结构由两个结构相同、参数不同的深度Q学习网络构成，即行为网络和目标网络；估价模块根据影响电子产品价格的内部属性得到t时刻的产品基本估价由估价模块输出的估计价格经过自适应调价模块选择的调价动作a_t调整之后得到最终价格本发明充分利用市场交易的信息，自适应调整定价以适应快速变化的市场，保证价格的合理性，提高交易成交率。 【EN】The invention relates to an asynchronous self-adaptive value evaluation method for electronic products based on a deep neural network, which is used for solving the problem of automatic pricing of second-hand electronic products. The system specifically comprises an evaluation module for estimating the value of a product and an adaptive pricing module for pricing and adjusting the market change, wherein the adaptive pricing module uses a double-depth Q network, and a model structure of the adaptive pricing module is composed of two depth Q learning networks with the same structure and different parameters, namely a behavior network and a target network; the evaluation module obtains basic evaluation of the product at the time t according to the internal attributes influencing the price of the electronic productEstimated price output by valuation modulePrice adjusting action a selected by self-adaptive price adjusting module_tAdjusted to obtain final priceThe invention makes full use of market trading information, adjusts price in a self-adaptive manner to adapt to the market with rapid change, ensures the rationality of the price and improves the trading success rate.