摘要：【中文】本发明公开了一种差速搅拌的化工反应釜，属于反应釜技术领域。本发明技术方案要点：通过电机带动转动轴套和位于转动轴套内部中心的转动内管进行转动，由于转动轴套和转动内管尺径的不同，和对应皮带轮的尺寸大小相异，转动轴套和转动内管呈现不同的线速度，促使主搅拌杆部分以及内旋转框和副搅拌杆在转动时中能够获得更高的搅拌效率，从而促进反应釜内溶液的反应效率，同时固定在主搅拌杆上下的过滤条随搅拌杆转动时，还可以将絮状杂质等缠绕固定；气泵通过连接管向转动内管中注气，气体通过固定分布在内旋转框底部的气嘴向上喷出，从而带动溶液翻腾，进一步促进溶液各组分的融合，同时还可以将较重的沉淀体向上推动促进其分散便于混合。 【EN】The invention discloses a differential stirring chemical reaction kettle, and belongs to the technical field of reaction kettles. The technical scheme of the invention is as follows: the rotating shaft sleeve and the rotating inner tube positioned in the center inside the rotating shaft sleeve are driven to rotate by the motor, the rotating shaft sleeve and the rotating inner tube are different in size from the corresponding belt pulley, the rotating shaft sleeve and the rotating inner tube have different linear velocities, so that a main stirring rod part, an inner rotating frame and an auxiliary stirring rod can obtain higher stirring efficiency during rotation, the reaction efficiency of a solution in a reaction kettle is improved, and meanwhile, flocculent impurities and the like can be wound and fixed when the filter strips fixed above and below the main stirring rod rotate along with the stirring rods; the air pump passes through the connecting pipe and injects the gas in to rotating the inner tube, and gaseous air cock including rotatory frame bottom through fixed distribution is spout upwards to drive solution and tumble, further promote the fusion of each component of solution, can also promote its dispersion to be convenient for to mix with heavier precipitate upwards promotion simultaneously.

摘要：【中文】本发明提供一种抗静电弹性体材料及其制备方法。该抗静电弹性体材料的构成原料按重量份数计包括100份弹性体和1‑150份高分子热塑性弹性体型抗静电剂，其中所述高分子热塑性弹性体型抗静电剂为聚醚的亲水性软链段共聚物，所述抗静电剂的共聚组分为聚酰胺硬段、聚乙烯硬段、聚酯硬段或聚丙烯硬段中的一种。本发明采用高分子热塑性弹性体型抗静电剂为原料，所得材料既不析出小分子物质，能够长效抗静电，又对原弹性体的弹性基本无影响，是一种高弹性、长效抗静电弹性体材料，适于集成电路、防尘等不同的应用。 【EN】The invention provides an antistatic elastomer material and a preparation method thereof. The antistatic elastomer material comprises the following raw materials, by weight, 100 parts of an elastomer and 1-150 parts of a high-molecular thermoplastic elastomer type antistatic agent, wherein the high-molecular thermoplastic elastomer type antistatic agent is a polyether hydrophilic soft segment copolymer, and a copolymerization component of the antistatic agent is one of a polyamide hard segment, a polyethylene hard segment, a polyester hard segment or a polypropylene hard segment. The invention adopts the macromolecule thermoplastic elastomer type antistatic agent as the raw material, the obtained material does not precipitate micromolecular substances, can resist static for a long time, basically has no influence on the elasticity of the original elastomer, is a high-elasticity and long-acting antistatic elastomer material, and is suitable for different applications such as integrated circuits, dust prevention and the like.

摘要：【中文】本发明公开了一种基于通道L1范数剪枝的神经网络压缩方法，属于计算、推算或计数的技术领域。该方法在外部数据集上训练卷积神经网络后获取该网络的初始化权值参数；在卷积核通道方向上根据参数group将待剪枝卷积层的权值参数分为n个组，并分别求取每个组的L1范数；根据一定的阈值对L1范数大于阈值的组内权值进行剪枝，对L1范数小于阈值的组内权值予以保留并参与再训练。本发明在保证参数的特征提取能力不被压缩的同时在一定程度上减小剪枝对最终目标检测准确率的影响，解决了剪枝后准确率下降过多以及硬件实现难度大的问题。 【EN】The invention discloses a neural network compression method based on channel L1 norm pruning, and belongs to the technical field of calculation, calculation or counting. The method comprises the steps of training a convolutional neural network on an external data set and then obtaining an initialization weight parameter of the network; dividing weight parameters of the convolutional layer to be pruned into n groups according to the parameter group in the direction of a convolutional kernel channel, and respectively solving an L1 norm of each group; pruning the intra-group weight with L1 norm larger than the threshold according to a certain threshold, and reserving the intra-group weight with L1 norm smaller than the threshold and participating in retraining. The method and the device ensure that the feature extraction capability of the parameters is not compressed, reduce the influence of pruning on the final target detection accuracy to a certain extent, and solve the problems of excessive reduction of the accuracy after pruning and high hardware realization difficulty.

摘要：【中文】本发明属于车辆设计技术领域，公开了一种驱动轴跳动校核方法，通过CATIA软件搭建悬架DMU模型；设定内节和外节的节型，确定理论内节点坐标和第一设定点坐标；将外节搭建到DMU模型中，跳动DMU模型，得到理论外节点坐标和第二设定点的坐标；求解得到多个不同的跳动工况下的实际内节点坐标；根据对应坐标求解得到多个不同的跳动工况下内节点的滑移量、内节夹角和外节夹角；调节轴杆长度，使得外节夹角在外节夹角许用值内，滑移‑摆角曲线位于滑移‑摆角安全边界线的中部。本发明中，采用CATIA软件对驱动轴进行跳动校核，只需搭建一次悬架模型，且无需逐个驱动轴长度进行模拟试验，能够快速准确的进行驱动轴校核。 【EN】The invention belongs to the technical field of vehicle design and discloses a drive shaft bounce checking method, wherein a suspension DMU model is built through CATIA software; setting nodal patterns of an inner node and an outer node, and determining a theoretical inner node coordinate and a first set point coordinate; building an outer node into a DMU model, jumping the DMU model to obtain a theoretical outer node coordinate and a coordinate of a second set point; solving to obtain actual inner node coordinates under a plurality of different jumping working conditions; solving according to the corresponding coordinates to obtain the slippage of the inner node, the included angle of the inner node and the included angle of the outer node under a plurality of different jumping working conditions; and adjusting the length of the shaft rod to ensure that the included angle of the outer joint is within the allowable value of the included angle of the outer joint, and the sliding-swing angle curve is positioned in the middle of the safe sliding-swing angle boundary line. In the invention, the drive shaft is checked for jumping by adopting CATIA software, only a suspension model is required to be built once, and the drive shaft length is not required to be simulated one by one, so that the drive shaft can be checked quickly and accurately.

摘要：【中文】本发明公开了一种用于卷积神经网络硬件加速器的时序弹性电路，涉及数字集成电路领域，适用于卷积神经网络硬件加速器的时序错误检测和校正。时序弹性电路包括：基于数据跳变检测的时序错误检测单元、在线校正单元和时钟控制单元，其中时序错误检测单元由13个晶体管构成，检测窗口长度可根据不同的工艺、电压、温度、老化程度条件进行调节；时序错误校正单元采用功耗较小的传统的锁存器结构，由10个晶体管构成；时钟控制单元生成时钟反向信号和检测窗口时钟信号，不检测时钟上升沿附近的数据延时，提高了电路的错误容忍度。结合卷积神经网络本身的容错性，本发明能够节省传统电路保留的过多时序裕度，且在保证数据精度的同时，降低电路的功耗。 【EN】The invention discloses a time sequence elastic circuit for a convolutional neural network hardware accelerator, relates to the field of digital integrated circuits, and is suitable for detecting and correcting time sequence errors of the convolutional neural network hardware accelerator. The timing elastic circuit includes: the system comprises a time sequence error detection unit, an online correction unit and a clock control unit, wherein the time sequence error detection unit is based on data jump detection and comprises 13 transistors, and the length of a detection window can be adjusted according to different conditions of process, voltage, temperature and aging degree; the time sequence error correction unit adopts a traditional latch structure with lower power consumption and is composed of 10 transistors; the clock control unit generates a clock reverse signal and a detection window clock signal, does not detect data delay near the rising edge of the clock, and improves the error tolerance of the circuit. By combining the fault tolerance of the convolutional neural network, the invention can save excessive time sequence margin reserved by the traditional circuit, and reduce the power consumption of the circuit while ensuring the data precision.

摘要：【中文】本发明公开了一种面向宽电压的在线时序检错纠错电路，涉及集成电路计算、推算、计数的技术领域，包括：在线时序检错单元、在线时序纠错单元、控制信号发生单元。在线时序检错单元仅由9个CMOS晶体管构成，面积开销小，在超低电压(0.4~0.6V)下保持稳定的检错性能；在线时序纠错单元由基于传输门的数据选通触发器构成，可同时实现检测数据采集和在线时序错误纠正，结构简单，功耗开销低；控制信号发生单元仅由反相器控制系统全局时钟信号，产生检测控制信号。本发明结构简单、性能稳定，在宽电压下应用于神经网络硬件加速器，可在线实现同一周期多次时序检错和纠错，极大地提高加速器吞吐量，降低系统能耗。 【EN】The invention discloses an on-line time sequence error detection and correction circuit facing to wide voltage, which relates to the technical field of integrated circuit calculation, calculation and counting, and comprises the following steps: the device comprises an online time sequence error detection unit, an online time sequence error correction unit and a control signal generation unit. The online timing sequence error detection unit is only composed of 9 CMOS transistors, the area overhead is small, and the stable error detection performance is kept under the ultra-low voltage (0.4-0.6V); the online time sequence error correction unit is composed of a data gating trigger based on a transmission gate, can simultaneously realize detection data acquisition and online time sequence error correction, and has simple structure and low power consumption overhead; the control signal generating unit controls the system global clock signal only by the inverter to generate the detection control signal. The invention has simple structure and stable performance, is applied to the neural network hardware accelerator under wide voltage, can realize multiple time sequence error detection and correction in the same period on line, greatly improves the throughput of the accelerator and reduces the energy consumption of the system.

摘要：【中文】本发明提供了一种测定高炉含铁炉料间高温交互反应性的方法，所述方法用于含铁炉料交互反应性的测定及评价。传统上依据单一炉料的熔滴性能来推断综合炉料在高炉高温区的行为并不全面、准确，烧结矿与块矿/球团矿在高温状态下发生交互反应，这种高温交互反应特性对综合炉料软熔滴落性能产生影响，高温交互反应评价指标的提出，可以为高炉优化配矿提供理论支撑。 【EN】The invention provides a method for measuring high-temperature interactive reactivity among iron-containing furnace materials of a blast furnace, which is used for measuring and evaluating the interactive reactivity of the iron-containing furnace materials. Traditionally, the behavior of the comprehensive furnace burden in a high-temperature area of the blast furnace is inferred to be incomplete and accurate according to the molten drop performance of a single furnace burden, the sintered ore and the lump ore/pellet ore are subjected to interactive reaction in a high-temperature state, the soft molten drop performance of the comprehensive furnace burden is influenced by the characteristic of the high-temperature interactive reaction, and the theoretical support can be provided for the optimized ore matching of the blast furnace through the proposal of the evaluation index of the high-temperature interactive reaction.

摘要：【中文】本发明涉及一种聚乙烯耐老化母粒防助剂析出的方法，包括以下重量份数的组分：聚乙烯100份、相容剂1‑5份、增容剂10‑80份、增塑剂1‑10份、抗氧剂0.1‑3份以及光稳定剂1‑20份；本发明公开了一种聚乙烯耐老化母粒防助剂析出的方法，由于聚乙烯属于非极性材料，防老化助剂一般具有一定极性，极性差导致载体材料和防老化助剂相容性差；上述配方中的相容剂、增容剂、增塑剂等有效解决了基体材料和防老化助剂的相容性问题，使防老化助剂分散均匀，助剂不易析出。 【EN】The invention relates to a method for preventing an auxiliary agent of a polyethylene anti-aging master batch from being separated out, which comprises the following components in parts by weight: 100 parts of polyethylene, 1-5 parts of compatilizer, 10-80 parts of compatibilizer, 1-10 parts of plasticizer, 0.1-3 parts of antioxidant and 1-20 parts of light stabilizer; the invention discloses a method for preventing an additive of a polyethylene anti-aging master batch from being separated out, wherein the polyethylene belongs to a nonpolar material, the anti-aging additive generally has a certain polarity, and the poor polarity causes poor compatibility of a carrier material and the anti-aging additive; the compatilizer, the compatibilizer, the plasticizer and the like in the formula effectively solve the problem of compatibility of the matrix material and the anti-aging auxiliary agent, so that the anti-aging auxiliary agent is uniformly dispersed and is not easy to separate out.

摘要：【中文】本发明公开了一种基于级联回归与孪生网络的目标跟踪方法，在第一帧，将指定的目标图像及以目标图像为中心构建的搜索区域分别输入以孪生网络构成的特征提取网络，分别得到目标图像与搜索区域的特征图。将得到的不同层次的特征图按设计的Attention机制融合，并一同输入到后续的级联区域推荐网络之中。经过级联区域推荐网络之后得到各个锚框的分类及位置预测信息，并得到锚框的预测得分图。通过非极大值抑制以及添加余弦窗等操作筛选出得分最高的锚框，并根据回归的得到的位置信息得到最终的预测框。本发明与传统算法相比上述算法具有精度高，鲁棒性强以及运行效率高的优点。 【EN】The invention discloses a target tracking method based on cascade regression and twin networks. And fusing the obtained characteristic diagrams of different layers according to a designed Attention mechanism, and inputting the characteristic diagrams into a subsequent cascade regional recommendation network together. And obtaining classification and position prediction information of each anchor frame after the cascade region recommendation network, and obtaining a prediction score map of the anchor frame. And screening the anchor frame with the highest score through operations such as non-maximum value inhibition, cosine window addition and the like, and obtaining a final prediction frame according to the position information obtained through regression. Compared with the traditional algorithm, the algorithm has the advantages of high precision, strong robustness and high operation efficiency.

摘要：【中文】本发明实施例公开了一种外后视镜转向调节方法、系统、设备及存储介质。当确定驾驶员正在驾驶当前车辆时，采集驾驶员的眼点位置，获取当前车辆的当前转向信息，根据当前转向信息和眼点位置确定当前车辆是否为主动转向状态，如果当前车辆为主动转向状态，根据眼点位置和当前转向信息计算外后视镜的目标位置，将外后视镜的位置调节至目标位置，解决了现有技术中在转弯和变道过程中，由于外后视镜的镜片位置保持不变，导致的给驾驶员造成视觉盲区的问题，达到了根据眼点位置和当前转向信息自动计算外后视镜的目标位置，并自动将外后视镜的位置调节至目标位置的目的，实现在车辆转向时增加驾驶员视野观察范围，以提高驾驶安全性的效果。 【EN】The embodiment of the invention discloses a method, a system, equipment and a storage medium for adjusting the steering of an exterior rearview mirror. When the driver is determined to be driving the current vehicle, the eyepoint position of the driver is collected, the current steering information of the current vehicle is obtained, whether the current vehicle is in an active steering state or not is determined according to the current steering information and the eyepoint position, if the current vehicle is in an active steering state, the target position of the outer rearview mirror is calculated according to the eyepoint position and the current steering information, the position of the outer rearview mirror is adjusted to the target position, the problems that in the prior art, in the turning and lane changing process, the problem of visual blind area caused to the driver due to the fact that the position of the lens of the outer rearview mirror is kept unchanged is solved, the target position of the outer rearview mirror is automatically calculated according to the eyepoint position and the current steering information, and the position of the outer rearview mirror is automatically adjusted to the target position, so that the visual field observation range of a driver is increased when the vehicle turns, and the driving safety is improved.