241 results about "Probabilistic neural network" patented technology

A method of predicting failures of data storage devices in a data storage network. In predicting failures, birth and/or health records of data storage devices in the data storage network are requested. The records are then scaled and thresholded and processed using a probabilistic neural network to classify the data storage devices. Based on the classification, an action is taken to improve the reliability and availability of the data storage network.

The invention is a automatic identification method and system based on the power quality disturbances of the information fusion, characterized by: collecting the transient and steady-state measurement datum associated with the power quality disturbances from the Power Quality Monitoring System and other automation systems, disposing of noise such as pretreatments; Using the method combining Fourier analysis, small wave multi-resolution decomposition and analysis of the correlation functions, distilling the from the disturbance datum, establishing the disturbance eigenvector, and as a the input characteristic vector of three probabilistic neural networks, realizing the mapping from a feature space to the disturbance space; the output of three probabilistic neural networks regarded as the evidence body of independent of each other, realizing Information Fusion by the use of D-S evidence theory, obtaining recognition results. This invention through the correct selection and extraction of disturbance eigenvectors, can input neural network parallel in classification and reflect the disturbance situation from the various aspects, thus effectively enhancing the correct identification rate of disturbance, a first step in Intelligent Recognition of the power quality disturbances.

The invention relates to a device and a method based on machine vision for detecting the surface defects of solder balls in the field of automatic optical detection. An optical imaging system collects chip images, and an image collection system receives the collected chip images, intercepts a single-frame gray image from video stream signals output by the optical imaging system, and stores the single-frame gray image in the form of a two-dimensional integer matrix; an image segmentation module segments the whole two-dimensional integer matrix corresponding to the whole image into subsidiary matrices containing the solder balls, an image characteristic extraction module receives and processes the two-dimensional integer subsidiary matrices output by the mage segmentation module and outputs a one-dimensional floating point vector, and a probability neural network module receives, trains, and tests the one-dimensional floating point vector output by the image characteristic extraction module and divides the solder balls into two categories, namely good solder balls and defective solder balls. The device of the invention has simple structure, the graduation of a characteristic extraction method and a probability neural network is good, and the accuracy is high so as to realize the NDE (Non-Destructive Examination) of the surface defects of the solder balls.

The invention discloses an infrared nondestructive testing method based on thermal image time sequence characteristics. The new method with pixel as a unit is used for the infrared nondestructive test of flaws and is based on thermal image time sequence characteristics and PNN (probabilistic neural network). The infrared nondestructive testing method comprises the following steps: firstly, obtaining an infrared time sequence thermal image of a heating test piece in a cooling process by an infrared thermal imager; secondly, extracting the gray values of normal and abnormal areas in the time sequence thermal image, establishing the relationship between the gray values of different areas and time and then obtaining the corresponding initial time sequence characteristics; thirdly, reducing the dimensions of the initial characteristics by adopting main constituent analysis so as to obtain the time sequence characteristics; and finally establishing the probabilistic neural network with the time sequence characteristics as training samples.

The invention discloses a hyper-spectral image technology-based detection method for distinguishing rice growing areas. The method comprises the following steps: (1) correcting black and white boards; (2) a training process: firstly, collecting a train sample image, correcting the sample image, extracting an interesting area, and extracting characteristics, wherein the characteristics comprise the percentage of chalky rice of rice samples of each growing area, the length-width ratio of each rice grain, and 15 texture eigenvalues, and building a prediction model by using a probabilistic neural network PNN method; and (3) testing to-be-detected rice samples. The detection method disclosed by the invention is fast and stable, and meanwhile, the classification accuracy is effectively improved.

The invention discloses a cutter damage adaptive alarm method based on a wavelet packet and a probability neural network. The method comprises the following steps of: fixing an acoustic emission sensor on a cutter bar, acquiring acoustic emission signals, performing three-layer wavelet packet analysis, selecting characteristic frequency bands and taking root mean square values thereof, normalizing the root mean square values to obtain smoothing factors and prior probability, establishing a cutter damage state probability model by using a probability neural network, determining an alarm value of the cutter abrasion state according to the model and the Pauta criterion, forming a dynamic alarm line, and performing adaptive alarm monitoring of the cutter operating state according to the dynamic alarm line. By the method, the probability distribution curve of the root mean square value related with the cutter abrasion can be found, the alarm value is determined by using a mathematical statistic method, the dynamic alarm line is formed together with the cutter abrasion state change, and missing alarm and error alarm are not caused.

The invention discloses a wheel service state safety margin estimation and fault diagnosis method. The wheel service state safety margin estimation and fault diagnosis method comprises the following steps of firstly carrying out feature extraction for a steel rail vibration signal, decomposing the steel rail vibration signal by adopting an EMS (Empirical Mode Decomposition) method, calculating related characteristic indexes of each IMF (Intrinsic Mode Function) to serve as a characteristic vector of a wheel service state; secondly, classifying normal and fault states by utilizing LSSVM based on state characteristic vectors of steel rail vibration signals of a normal wheel and a faulted wheel, obtaining a safety margin of a train wheel and estimating the train wheel service state; and finally, carrying out fault mode identification for a normal wheel, a flat damage wheel and an out-of-round wheel by adopting a PNN (Probabilistic Neural Network), and providing reference basis for a vehicle maintenance department. The wheel service state safety margin estimation and fault diagnosis method has the advantages of high reliability and good engineering feasibility.

The invention relates to an excitation inrush current identification method based on wavelet transformation and a probabilistic neural network (PNN), in particular to a novel method for identifying excitation inrush current by combining wavelet analysis and the PNN. Data of the excitation inrush current and fault current are acquired by simulation by a model constructed in Matlab/Simulink; and energy characteristic values of fault signals on different frequency bands are extracted by using the wavelet transformation, and are used as input parameters of the PNN, so that the excitation inrush current and short circuit current can be identified. The method is high in identification capacity; and confused judgment on a real short circuit fault and the excitation inrush current is avoided.

The invention relates to the field of pattern recognition, in particular to a pattern recognition method based on a self-adaptation correction neural network. The method comprises the steps of classifying input training samples through a probabilistic neural network model so as to obtain samples accurate in classification and samples inaccurate in classification; adding an input layer, a central layer and an excitation layer on the basis of the probabilistic neural network model structure so as to construct a self-adaptation correction neural network model structure; for the samples inaccurate in classification in the probabilistic neural network model, using themself as central points, calculating the allowance radius between the the samples and samples of other classifications, clustering error samples of same category so as to realize batch correction of classification patterns and replanning of a judging interface and build the self-adaptation correction neural network; finally, conducting pattern recognition on input testing samples based on the self-adaptation correction neural network model. The pattern recognition method has the advantages of being high in accuracy in mode classification, strong in mode generalization ability, good in classification real-time performance, wide in application prospect, and the like.

The invention discloses a probabilistic neural network-based tolerance-circuit fault diagnosis method, which comprises the following steps of: selecting a pulse signal source as the energization of a fault circuit to be detected; carrying out Monte Carlo analysis on the fault circuit so as to obtain an amplitude-frequency response signal of the fault circuit to be detected; carrying out three-layer wavelet packet decomposition on the amplitude-frequency response signal of the fault circuit so as to obtain low and high frequency coefficients of the amplitude-frequency response signal, carryingout threshold quantification on the wavelet packet decomposition coefficients, then carrying out wavelet packet reconstruction according to the lowest-layer low frequency wavelet packet decompositioncoefficients and the high frequency wavelet packet decomposition coefficients subjected to threshold quantification so as to complete the de-noising processing of the wavelet packet; calculating the band-gap energy of the response signal according to the low and high coefficients obtained after wavelet packet reconstruction, and constituting a fault characteristic vector by using the band-gap energy; and inputting the fault characteristic vector in a fault grader of the probabilistic neural network to realize circuit fault diagnosis. The method has the advantages of high right fault diagnosisrate, simple structure, short training time, high fault tolerance and strong extrapolation ability.

The invention relates to an online safety detection prewarning device of an elevator. The device comprises a car prewarning device and a machine room prewarning device, wherein the car prewarning device consists of a first sensing unit, a first acquiring unit, a first processing unit and a first near field communication unit; the machine room prewarning device consists of a second sensing unit, a second acquiring unit, a second processing unit and a second near field communication unit; the first near field communication unit and the second near field communication unit are in wireless communication; and the second processing unit is remotely communicated with a remote platform through a remote communication gateway. The invention further discloses an online safety detection prewarning method of the elevator. The device adopts a probability neural network algorithm; and the algorithm enables measured characteristic parameter thresholds to continuously self-learn, self-adapt and self-adjust along with such conditions as elevator ageing degree, maintenance condition and load magnitude, so that the wrong judgment or judgment missing probability is largely reduced, and the long-time use reliability and accuracy of a prewarning function of the device are improved.

The invention discloses a probabilistic neural network algorithm-based cable state evaluation method. The method includes the following steps that: a multi-source information database is constructed according to many types of information sources such as basic information, operation information, operation inspection information and family defects of cable equipment; and the database under an incomplete information condition is pre-processed by using a factor analysis method, and a probabilistic neural network-based multi-source information algorithm is put forward, and the overall state of a cable under an incomplete information condition can be comprehensively evaluated. With the method of the invention adopted, the fault of a cable can be detected timely and judged accurately, and the operating state of the cable can comprehensively reflected.

The invention discloses an intelligent diagnosis method for the failure of a wind generating set. The diagnosis method includes the steps that first, according to historical failure knowledge information of the wind generating set, a failure tree model of the wind generating set is established; then according to the structure of the failure tree model, a probability neural network structure modelis established, and historical failure sample data of the wind generating set are mapped into failure mode space to form a failure diagnosis network model with high fault tolerance and adaptive ability; finally, the failure data is input into the established failure diagnosis network model to obtain a diagnosis result, a failure mode is output and then matched with a corresponding failure tree branch, leaf nodes governed by the failure tree branch are positioned, and namely reasons or reason combinations leading to the failure are found out. Based on a failure tree and a probability neural network, intelligent diagnosis is performed on the multi-form failure of the wind generating set, multi-failure diagnosis analysis is performed on complex failures of the wind generating set under incomplete information, and failure reasons are accurately positioned.

The present invention discloses a smart mobile terminal authentication method based on multi-finger touch behavior characteristics. The method comprises a first step of obtaining touch screen operating information of a user through multi-finger touch screen operation, the multi-finger touch screen operation being touch screen sliding of multiple fingers at a together-put state; a second step of extracting a physiological feature and a behavior feather of a touch screen from the user touch screen operation information, and combining the above feathers as one feature vector; a third step of training a probabilistic neural network classifier and a support vector machine classifier by using the above combined feather vector, to generate a training model of valid users; and a fourth step of collecting touch screen operation information of test users, processing the information according to the second step to obtain combined feature vectors of the touch screen operation of the test users, and putting the combined feature vectors into the two classifiers to compare with the training model of the valid users, thus to obtain classifying results if the test users are valid users.

The present disclosure proposes a method of moving object detection in variable bit-rate video steams based on probabilistic neural networks, and the method features a background generation module and a moving object detection module. The background generation module produces a model of background images which express properties of variable bit-rate video streams. The moving object detection module distinguishes a moving object in both low and high bit-rate video steams in an efficient manner. The detection result is generated by calculating the output value of the probabilistic neural networks.

The invention discloses a power transformer fault diagnosis method based on an improved firefly algorithm (PFA) optimized probabilistic neural network (PNN). The power transformer fault diagnosis method comprises the following steps: firstly, collecting fault characteristic gas by using a gas chromatographic analysis method and carrying out pretreatment by using a fused DGA algorithm; initializinga PNN neural network, a firefly algorithm and a two-dimensional particle swarm; taking the PNN smoothing factor as a firefly individual, and calculating the position and brightness of the firefly; feeding the solving result of each firefly algorithm back to the particle swarm algorithm, carrying out fitness evaluation on each particle, and updating the positions and speeds of the particles; carrying out loop iteration, substituting the obtained optimal smoothing factor into the PNN to carry out fault prediction, and training a PNN model after PFA optimization; inputting a test sample, and outputting a fault type result, thereby achieving the fault diagnosis of the power transformer. The method is high in search speed, high in diagnosis precision, small in error, and obvious in classification effect.

The invention discloses a visible light channel joint balance method based on orthogonal mapping and a probabilistic neural network. The method comprises a transmitting end and a receiving end. Signals are transmitted from the transmitting end to the receiving end through visible light MIMO (Multiple Input Multiple Output) channels. The visible light MIMO channels are multiple input multiple output channels. According to joint balance, preposition balance and postposition balance are combined. According to the method, a joint balance scheme combing a preposition balance technology and a postposition balance technology is employed, namely, the visible light multiple input multiple output channel joint balance method based on the orthogonal mapping and the probabilistic neural network is employed. The interference between the channels of a visible light MIMO communication system can be effectively suppressed and the data transmission reliability is improved.

The invention relates to a photovoltaic array fault diagnosis method and device based on spatiotemporal distribution characteristics. A photovoltaic output time component and a spatial component are calculated and obtained by a time and spatial distribution function for the photovoltaic array output starting from the historical operation data and the historical environment data of a photovoltaic power plant, the components are used as a training data set, a probabilistic neural network is trained by the training data set, a photovoltaic array fault diagnosis model is obtained, a time functionand a spatial function for photovoltaic branch output are established to effectively describe the distribution law of string current under different fault conditions, the photovoltaic power station DCside array fault diagnosis method established based on the probabilistic neural network is high in diagnosis precision, and finally, branch current data are inputted to the photovoltaic array fault diagnosis model to realize diagnosis of multiple types of fault, and the existing data of the actual project can be fully utilized effectively.

The invention discloses a rapid unit failure diagnosis method based on full state information. The method includes the steps that a probabilistic neural network is used for conducting unit failure recognition; 14 types of typical failure data are used for obtaining the characteristic values of all the 14 types of failure through holographic spectral analysis or time domain statistic analysis, and all the characteristic values constitute characteristic vectors; the 14 types of sample characteristic vectors are used as the weight vectors Wj of 14 types of mode units respectively; the data to be diagnosed are selected to conduct the holographic spectral analysis so that the characteristic vectors can be obtained to be used as the input vectors Xj of a neural network input layer, and input sample data and training sample data adopt the same parameter; scalar product calculation is carried out on the input vector Xj and the weight vector Wj of each mode unit; summation of the outputs g(Zj) of the mode units corresponding to the same failure mode is conducted so that the probability density of the failure can be estimated; the outputs fR(X) of 14 accumulation layers corresponding to the 14 types of failure modes are used as inputs so that the failure modes can be judged through the Bayes judgment strategy. By means of the rapid unit failure diagnosis method based on the full state information, the expertise in the field can be fully utilized, and the requirement for the experience of a user himself or herself is lowered.

The method uses self-learning mode to construct equipment state probability model with probability nervous network based on dynamic data of equipment operation. The model adjusts its probability distribution by following to the equipment operation to describe variation rule of the equipment state dynamically. Relying on the model, the dynamic evaluation rule of the equipment state is studied and simultaneously the threshold for different state is set up to form adaptive alarming limit for the equipment operation. The equipment state can be graded in three classes of normal, fault and quick deterioration by utilizing this adaptive alarming method.