62results about How to "Improve anomaly detection accuracy" patented technology

The invention discloses a distributed migration network learning-based intrusion detection system and a method thereof, and mainly solves the problems that the prior method has low efficiency in detection of some attack types and is difficult to search data again. The whole system comprises a network behavior record preprocessing module, an abnormality detection module and an abnormal behavior analyzing module. The network behavior record preprocessing module completes the quantification and normalization processing of a network behavior record; the abnormality detection module uses an abnormality detection learning machine to completes the classification and identification for an input record, determines whether the record is a normal behavior, and completes the detection if the record is a normal behavior or transmits the record to the abnormal behavior analyzing module if the record is an abnormal behavior; and the abnormal behavior analyzing module uses an abnormal behavior analyzing learning machine to carry out the classification and identification of the input records and outputs the attach type of the record. The system and the method have the advantages of using other existing resources to improve the detection rate for the prior attach types with low detection rate and avoiding searching the data again and can be used for network intrusion detection.

The invention discloses a rail fastener anomaly detection system based on binocular vision and laser speckle, which is composed of a laser speckle projector, a camera, a laser displacement sensor, a wheel encoder, an RFID detector and an industrial computer. The speckle projector and the camera constitute a binocular stereo vision speckle imaging system, which is used for three-dimensional measurement of the fastener; the laser displacement sensor is used to measure the height of the fastener installation support, which is used to control the camera imaging, wheel encoder and The RFID detector is used for mileage counting, the industrial computer completes the image acquisition and processing, and the method based on binocular vision stereo matching is used to detect whether the fastener is abnormal. The device uses a laser speckle light source to artificially increase the mottled texture on the surface of the fastener, which can improve the matching accuracy of image blocks and accurately reconstruct the three-dimensional shape of the fastener, thereby effectively detecting various abnormalities of different types of fasteners.

The present invention provides an elevator door control device that improves the abnormal detection accuracy of the door opening and closing action by properly detecting the abnormality of the torque command, and reduces the false detection of the abnormal door opening and closing action. It is equipped with a speed command unit that outputs a speed command; The speed control unit of the torque command based on the deviation between the speed command and the feedback speed; store a plurality of torque command graphics formed by sampling a plurality of torque commands according to the door opening and closing action, and store the torque command graphics based on the multiple torque command graphics The reference torque instruction graph storage unit of the reference torque instruction graph obtained by each common sampled torque instruction; the abnormal detection torque graph generation unit that generates the abnormal detection torque graph from the reference torque instruction graph; when the torque instruction exceeds An abnormality avoidance unit that outputs an abnormality avoidance command to a speed command unit when an abnormality detection torque pattern is detected.

The invention discloses an abnormal access data detection method and device, and relates to the technical field of computers. A specific embodiment of the method comprises the steps of training an abnormal access data detection model according to a pre-established initial training set; after the training is finished, determining an abnormal probability of access data in a pre-established initial verification set by utilizing an abnormal access data detection model; labeling a plurality of pieces of access data of which the abnormal probability and accuracy meet a preset discrimination condition in the initial verification set as abnormal access data, and adding the abnormal access data into an initial training set; training the abnormal access data detection model according to the currenttraining set; and inputting to-be-detected access data into the trained abnormal access data detection model, and judging whether the to-be-detected access data is abnormal access data or not according to an output result. According to the embodiment, the abnormal access data detection model can be continuously optimized by utilizing the dynamically adjusted training set, so that the abnormal detection accuracy of the model is improved.

The invention provides an anomaly detection method and device based on a user peer-to-peer group, and the method comprises the steps: carrying out the clustering of a preset number of times of a userset comprising a target user according to a preset clustering algorithm, obtaining a preset number of clustering clusters to which the target user belongs, wherein each clustering corresponds to one clustering cluster; according to a preset number of clustering clusters to which the target user belongs, obtaining a peer-to-peer group probability of the target user and any user in the preset numberof clustering clusters, and constructing a probability clustering cluster based on the peer-to-peer group probability; selecting a preset threshold number of behavior characteristics from a preset behavior characteristic pool, and for each selected behavior characteristic, based on the probability clustering cluster and the selected behavior characteristic, constructing an isolated forest tree containing the target user; and detecting whether the target user is an abnormal user or not according to the obtained isolated forest tree and a preset anomaly detection algorithm. The anomaly detection precision can be effectively improved.

The invention discloses a mobile armrest inspection device of a passenger conveyer. An image is restricted by a detection body (5), so that three sets of joints obliquely arranged with respect to the elongation direction of the inner steel wires buried in the mobile armrest (2) are detected.When the abnormity is continuously detected by an image processor (8) and the variation amount delta P of each light amount is within a predetermined range of a time interval delta t, a joint is determined by a determination mechanism.When the variation amount delta P of each light amount is not within thepredetermined range of the time interval delta t, a joint is not determined by the determination mechanism.When the joint is determined, no abnormality on the joint is judged by an abnormality determination mechanism.When the joint is not determined, an abnormality on the joint is judged by the abnormality determination mechanism.Therefore, whether the abnormality of a tension element is existed or not can be quantitatively and fully detected by the inspection device, including the abnormality of the joint of the inner steel wires buried in the mobile armrest (2).

The invention belongs to the technical field of software engineering and cloud computing, and particularly relates to a distributed system call chain and log fusion anomaly detection method. The method comprises the following steps: constructing a call chain event relation graph according to a call chain and log data on the basis of the call chain and log data during operation of a distributed system, and learning a call chain event relation graph mode during normal operation of the system by utilizing a graph neural network and a single-classification deep learning method; detecting a newly generated call chain event relation graph in real time during online use, and identifying a call chain generating an abnormal behavior; the method specifically comprises the steps of log event analysis, call chain event analysis, event vectorization, call chain event relation graph construction, graph neural network model training and online anomaly detection. According to the invention, operation and maintenance personnel and developers can be helped to quickly discover system exceptions, corresponding alarm information is generated, the speed of fault positioning and online problem solving is accelerated, and the labor cost is reduced.

The invention discloses an effluent BOD sensor anomaly detection method based on a modular neural network, relates to the field of artificial intelligence, and is directly applied to the field of sewage treatment. In order to solve the problems that in the current sewage treatment process, an effluent BOD sensor drifts and abrupt abnormity cannot be detected in real time under complex working conditions, samples are automatically classified and input according to the working conditions by adopting a density-based clustering algorithm; extracting an effluent BOD auxiliary variable as an input variable of each sub-network in the modular network by using a mutual information-based method; designing a self-organizing RBF neural network based on error correction as a sub-network, and training the network through an improved Levenberg-Marquardt (LM) algorithm to improve the training speed. The result shows that the abnormity detection method is compact in structure, abnormity of the effluent BOD sensor in the sewage treatment process can be rapidly and accurately detected, and technical guarantee is provided for safe and stable operation of sewage treatment.

The invention relates to an equipment state anomaly detection method and device, computer equipment and a storage medium. The method comprises the following steps: acquiring a feature set of initial fault-free operation data of equipment as a historical initial feature set; searching the historical initial feature set according to a distance correlation coefficient between features in the historical initial feature set to obtain a historical correlation feature subset, and screening redundant features in the correlation feature subset according to a feature representative index of the historical correlation feature subset to obtain a historical feature set; obtaining a historical data matrix according to the historical feature set, and taking the historical data matrix as a training sample to obtain a trained convolutional noise reduction network; and inputting the real-time data matrix into the trained convolutional noise reduction network to obtain an abnormal level value, and judging whether the equipment is abnormal or not according to a size relationship between the abnormal level value and an abnormal threshold value. By adopting the method, higher anomaly detection precision can be obtained under a lower false alarm rate.

The invention discloses an injection molding machine energy consumption abnormity detection method and system based on a Gaussian mixture model, and the method comprises the steps: carrying out the real-time collection of the energy consumption data of a first injection molding machine, and obtaining the first real-time energy consumption data; performing data preprocessing on the first real-time energy consumption data to obtain second real-time energy consumption data; inputting the second real-time energy consumption data into a Gaussian mixture model for clustering feature learning to obtain a first clustering data set and generate a first mark training data set; performing model training according to the first mark training data set to obtain a first anomaly detection model; and inputting a first test training data set of the first injection molding machine into the first anomaly detection model to obtain first output information. The technical problem that in the prior art, when the energy consumption abnormity of the industrial injection molding machine is detected, due to the fact that data features are not comprehensive and perfect enough, multi-dimensional data classification is not accurate enough, and the false alarm rate is high, the detection precision is not high is solved.

The present disclosure relates to a device for detecting a liquid crystal module and a method for detecting the quantity of liquid crystal. The method includes a cover for fixedly accommodating the liquid crystal module. An accommodation space for accommodating collision units is formed between the liquid crystal module and the cover, the collision units being connected to the top wall of the cover through elastic members. The collision units are spaced from the liquid crystal module in a certain distance in a static status, and collide with a panel of the liquid crystal module through vibration of the cover in the vertical direction during detection. The device for detecting a liquid crystal module improves the detection efficiency.