摘要：【中文】本公开提供了一种检测设备、控制设备及自动校验系统和方法。检测设备由控制设备控制实现对待检测物品的自动检测，该检测设备包括：传送组件，用于在接收到控制设备发送的传送信号的情况下，传送待检测物品；信息采集组件，沿传送组件的传送方向设置，用于根据控制设备发送的采集信号，在待检测物品经由传送组件传送至信息采集组件的采集范围时，采集待检测物品的物品信息，其中，待检测物品的物品信息是控制设备确定待检测物品是否合格的依据。 【EN】The disclosure provides a detection device, a control device, an automatic checking system and an automatic checking method. The detection equipment is controlled by the control equipment to realize the automatic detection of the article to be detected, and the detection equipment comprises: the conveying assembly is used for conveying the article to be detected under the condition of receiving a conveying signal sent by the control equipment; the information acquisition assembly is arranged along the conveying direction of the conveying assembly and used for acquiring the article information of the article to be detected according to the acquisition signal sent by the control equipment when the article to be detected is conveyed to the acquisition range of the information acquisition assembly through the conveying assembly, wherein the article information of the article to be detected is the basis for the control equipment to determine whether the article to be detected is qualified.

摘要：【中文】本发明提出的一种深度多层网络驱动的特征聚合类别划分器，比单模态的类别划分有较大提升，通过对每个类别的错误分类实例进行分析，发现多模态的学习分类方法明显错误更少，错误原因更易查找和解决。在对统计差异性大的不同模态特征的关联学习上，深层的学习模型更具优势。而对于使用不同特征的同一模型，实验发现使用区域卷积神经网络驱动的模态聚合模型在单个类上的准确率大都优于使用文档主题和卷积神经网络驱动的模态聚合模型。另外，在性能上，由于使用区域卷积神经网络驱动的模态聚合模型具有运算性能上的优势，特征提取速度更快，从而使整个过程获得了更高的效率和准确率。 【EN】Compared with single-mode class division, the deep multi-layer network-driven feature aggregation class divider provided by the invention is greatly improved, and by analyzing the error classification example of each class, the multi-mode learning classification method is obviously less in error and the error reason is easier to find and solve. The deep learning model is more advantageous in the association learning of different modal characteristics with large statistical difference. For the same model with different characteristics, experiments show that the accuracy rate of the modal aggregation model driven by the regional convolutional neural network on a single class is superior to that of the modal aggregation model driven by the document theme and the convolutional neural network. In addition, in performance, the modal aggregation model driven by the regional convolutional neural network has the advantage of operational performance, and the feature extraction speed is higher, so that the whole process obtains higher efficiency and accuracy.

摘要：【中文】本发明是一种气动系统节能领域的新型节能技术。发明内容主要面向电解铝行业打壳气缸的气控回路，旨在检测打壳装置的压缩空气泄漏量，为现场检修人员提供气缸维修指示,减少打壳装置压缩空气泄露量，实现电解铝打壳节能。本发明用在电解铝打壳控制回路上，增设具有流量测量和管理功能的旁路装置，采用分时区别供气方式，再实现系统泄漏量指示，同时通过切换气路匹配设备不同工艺时段用气需求，大幅缩减压缩空气泄露，实现打壳系统大幅节能。 【EN】The invention relates to a novel energy-saving technology in the field of pneumatic system energy conservation. The invention mainly aims at a pneumatic control loop of a crust breaking cylinder in the electrolytic aluminum industry, and aims to detect the compressed air leakage amount of a crust breaking device, provide a cylinder maintenance instruction for field maintainers, reduce the compressed air leakage amount of the crust breaking device and realize energy conservation of electrolytic aluminum crust breaking. The invention is used on the electrolytic aluminum crust breaking control loop, is additionally provided with the bypass device with flow measurement and management functions, adopts a time-sharing difference air supply mode, realizes system leakage indication, and greatly reduces compressed air leakage by switching air circuit matching equipment to meet air use requirements in different process periods, thereby realizing the great energy saving of the crust breaking system.

摘要：【中文】本说明书实施例公开了一种敏感资讯检测方法、装置、设备及系统。所述方法包括获取资讯信息；判断所述资讯信息中是否存在预设敏感词库包括的敏感词；确定不存在时，将所述资讯信息输入预先构建的资讯检测模型中，所述资讯检测模型基于不符合信息规范的资讯文本训练获得；获取所述资讯检测模型输出的资讯检测结果。利用本说明书实施例可以减少人工审核的成本与压力，有效提高对资讯的检测效率与检测质量，最大程度减少敏感资讯而产生的不良影响。 【EN】The embodiment of the specification discloses a method, a device, equipment and a system for detecting sensitive information. The method comprises the steps of obtaining information; judging whether the information has sensitive words included in a preset sensitive word bank or not; when the information does not exist, inputting the information into a pre-constructed information detection model, wherein the information detection model is obtained based on information text training which does not accord with the information specification; and acquiring an information detection result output by the information detection model. By using the embodiment of the specification, the cost and pressure of manual examination can be reduced, the detection efficiency and the detection quality of the information can be effectively improved, and the adverse effect caused by sensitive information can be reduced to the maximum extent.

摘要：【中文】本发明属于海洋领域，具体涉及一种波浪参数的概率统计分析方法。基于观测点的波浪要素长期的观测数据计算转换高斯函数值，将时历数据转换为高斯过程，再利用高斯过程的统计分析模型以及转换高斯函数值计算得到观测数据的统计分析结果，并可以此改进高斯模型进行波浪极值的预测。本发明提出的改进高斯模型能够精确的对波浪参数进行统计分析，并能在海况较高的情况下，有效规避高斯模型对波峰与波谷的对称估计，满足stokes的二阶波浪理论。不论在低海况还是高海况的情况下，该模型计算得到波浪的极值及周期的概率分布结果与观测数据结果吻合度高，能作为波浪参数概率分析模型进行波浪极值的预测。 【EN】The invention belongs to the field of oceans, and particularly relates to a probability statistical analysis method for wave parameters. Calculating a conversion Gaussian function value based on long-term observation data of the wave elements of the observation points, converting the time-history data into a Gaussian process, calculating by using a statistical analysis model of the Gaussian process and the conversion Gaussian function value to obtain a statistical analysis result of the observation data, and predicting a wave extreme value by improving the Gaussian model. The improved Gaussian model provided by the invention can accurately perform statistical analysis on wave parameters, effectively avoids the symmetric estimation of the Gaussian model on the wave crest and the wave trough under the condition of high sea condition, and meets the second-order wave theory of stokes. Under the conditions of low sea condition and high sea condition, the probability distribution result of the extreme value and the period of the wave obtained by the model calculation has high coincidence degree with the observation data result, and can be used as a wave parameter probability analysis model to predict the extreme value of the wave.

摘要：【中文】一种防爆灯及其灯罩安装方法。其特征在于：灯壳一端设有内环壁、外环壁，内环壁与外环壁之间形成环形灌胶槽，外环壁上端内壁设有若干周向间隔布置的第一限位凸台；灯罩包括开口端，开口端外侧壁设有若干周向间隔布置的第二限位凸台，开口端设置在环形灌胶槽内，且第二限位凸台与第一限位凸台轴向限位配合；环形灌胶槽内灌注有密封胶，开口端与内环壁、外环壁、灯壳端面之间均通过密封胶密封粘接。其优点在于，通过密封胶实现灯罩与灯壳的的连接，结构简单，安装方便，成本低，同时配合第一限位凸台与第二限位凸台的限位，使得灯罩与灯壳之间连接牢固可靠，寿命高，且开口端与内环壁、外环壁、灯壳端面之间均通过密封胶密封粘接，密封防爆性能好。 【EN】An explosion-proof lamp and a method for mounting a lampshade thereof. The method is characterized in that: an inner ring wall and an outer ring wall are arranged at one end of the lamp shell, an annular glue pouring groove is formed between the inner ring wall and the outer ring wall, and a plurality of first limiting bosses which are circumferentially arranged at intervals are arranged on the inner wall at the upper end of the outer ring wall; the lampshade comprises an opening end, a plurality of second limiting bosses which are circumferentially arranged at intervals are arranged on the outer side wall of the opening end, the opening end is arranged in the annular glue pouring groove, and the second limiting bosses are axially limited and matched with the first limiting bosses; the annular glue pouring groove is filled with sealant, and the opening end is hermetically bonded with the inner annular wall, the outer annular wall and the end face of the lamp shell through the sealant. The lamp shade has the advantages that the connection of the lamp shade and the lamp shell is realized through sealant, the structure is simple, the installation is convenient, the cost is low, the limit of the first limit boss and the second limit boss is matched simultaneously, the connection between the lamp shade and the lamp shell is firm and reliable, the service life is long, the opening end and the inner annular wall, the outer annular wall and the end face of the lamp shell are sealed and bonded through the sealant, and the sealing and explosion-proof performance is good.

摘要：【中文】本发明公开了一种资讯分类模型的训练方法及装置，采集训练资讯，得到训练资讯集；对训练资讯集中的训练资讯进行多类别标注；对多类别标注后的训练资讯集进行切词和过滤，得到训练词集；将训练词集作为输入，根据改进朴素贝叶斯模型进行模型训练，得到资讯分类模型。可见，本发明的资讯分类模型通过多类别标注后的训练资讯集进行模型训练，最终资讯分类模型得到的输出结果是输入资讯属于不同资讯类别的概率，而不是像现有技术一样直接将资讯标定为某一资讯分类，不考虑资讯内容包含的其他分类内容，本发明的资讯分类更加清楚准确。 【EN】The invention discloses a training method and a device of an information classification model, which are used for acquiring training information to obtain a training information set; marking the training information in the training information set in multiple categories; performing word segmentation and filtering on the training information sets labeled by the multiple classes to obtain training word sets; and taking the training word set as input, and carrying out model training according to the improved naive Bayesian model to obtain an information classification model. Therefore, the information classification model of the invention carries out model training through the training information set labeled by multiple classes, and the output result obtained by the information classification model is the probability that the input information belongs to different information classes, rather than directly labeling the information as a certain information classification like the prior art, and other classification contents contained in the information content are not considered, so that the information classification of the invention is more clear and accurate.