40results about How to "Promote self-learning" patented technology

The invention relates to a transmission line audible noise prediction method based on BP neural network optimization. The method includes acquiring and preprocessing data parameters, establishing a BP neural network prediction model of the data parameters, performing optimization training on the BP neural network by the ant colony algorithm, and utilizing the network to predict the transmission line audible noise. The method has high prediction accuracy, fine convergence and high stability, the problem that the BP neural network falls into local minima is avoided, and great guidance significance is provided for transmission line audible noise prediction and audible noise reducing development.

The invention relates to a self-adaption wavelet neural network abnormity detection and fault diagnosis classification system and method, which can be applied to the fields, such as economic management abnormity detection, image recognition analysis, video retrieval, audio retrieval, signal abnormity detection, safety detection, and the like. The system comprises the following seven parts: an acquisition device, a transmitter device, an A/D (Analog/Digital) conversion device, a self-adaption wavelet neural network abnormity detection and fault diagnosis classification processor, a display interaction device, an abnormity alarm device and an abnormity processing device. The abnormity detection and fault diagnosis classification object of the self-adaption wavelet neural network abnormity detection and fault diagnosis classification system is acquired from samples for which a self-adaption mechanism is automatically established by the self-adaption wavelet neural network of a system to be detected, the characteristic information of a signal can be effectively extracted through wavelet transform multi-scale analysis, and a more accurate abnormity detection and fault diagnosis locating result can be obtained. The device adopting the method has the advantages of generalization, high accuracy in the application field, capability of real-time monitoring and low cost.

The invention relates to a method for predicting radio disturbance of an electric transmission line by using an improved BP (back propagation) neural network. The method comprises the following steps of acquiring and preprocessing data parameters; establishing a BP neural network predicting model of the data parameters; performing optimal training on the BP neural network by using a genetic algorithm and a simulated annealing algorithm; and predicting radio disturbance of the electric transmission line by using the network. The method is high in prediction precision, high in convergence and high in stability; the problem that the BP neural network is caught in a local minimum point is solved; and the method has a great guiding significance on prediction of radio disturbance of the electric transmission line and research on reduction of radio disturbance.

The invention discloses a distributed optical fiber sensing signal mode identifying method and system. The distributed optical fiber sensing signal mode identifying method includes steps of acquiring samples, and acquiring several groups of samples for excitation signals of each behavior, so as to set up a sample library; roughly screening samples, screening the acquired sample library to remove samples high in discreteness; finely screening the samples, training to generate a classifier used for signal mode identification by means of the screened sample library and storing; identifying signal modes, calculating characteristic vectors for a newly triggered and read section of signal data and subjecting the characteristic vectors to calculation in the stored classifier to obtain mode identification results. The distributed optical fiber sensing signal mode identifying method and system can effectively identify invasion or damage signals in case that signals are carried with noise or distortion, and in the meantime, mistaken reports are reduced.

The invention discloses an intelligent gear defect analysis method based on fractional wavelet transform and a BP neutral network. The intelligent gear defect analysis method includes: taking transform order as a variable to perform fractional Fourier transform on gear vibration signals to determine optimal order, and performing fractional wavelet transform on the gear vibration signals under the optimal order for denoising to realize separation of useful component and background noise of the gear vibration signals; calculating feature parameters of the signals after being denoised to form a group of feature vectors which are used for representing features of gear vibration after denoising; averagely dividing the feature vectors into two groups which serve as a training sample and a testing sample respectively, and inputting the feature vectors into the BP neutral network for learning and classifying. By the intelligent gear defect analysis method, background noise mixed in the gear meshing vibration signals is inhibited well, useful signal component related to defects is retained, and gear defect features can be extracted effectively; self learning and classifying capability of the BP neutral network is utilized, so that defect mode of gears can be quickly recognized qualitatively with high accuracy.

The invention provides a prediction method for the short-term output of a distributed photovoltaic power generation system. The method comprises the steps: S1, collecting the historical output data, atmospheric temperature data and solar irradiation data of the distributed photovoltaic power generation system, and carrying out the normalization preprocessing of the data; S2, establishing a first prediction model based on an extreme learning machine (ELM) by using a multi-model univariate prediction method, and carrying out preliminary prediction on a to-be-predicted day; and S3, establishing asecond prediction model based on a long-term and short-term memory (LSTM) deep learning network by using a single-model multivariable prediction method, and correcting the result of the first prediction model in combination with the historical data to obtain a final photovoltaic output prediction result. With implementation of the method, the prediction precision and the adaptability can be improved.

The invention relates to an uncalibrated hand-eye coordination fussy control method based on support vector regression (SVR) learning. The method comprises the steps of recording the moving conditions of a paw of a robot and an object by a camera, projecting the moving conditions to an image plane, marking the projection positions of the paw and the object in the image plane by adopting an image processing and object identification method, inputting the position difference between the paw and the object to an SVR-based fussy controller designed in the invention as an input quantity, and outputting the change quantity of an arm joint of the robot by the controller so as to control the paw to move to the tracked object. By adopting the method, the quantity of samples is greatly reduced, a few steps are adopted and better effects are achieved in the control, meanwhile, the method also has the characteristics of good self-learning and generalizing capability and language interpretability. The design of the fussy controller based on a fuzzy basic function not only provides convenience for analyzing the similarity of functions, but also provides an effective path for combining expert knowledge and fussy rules generated by data information.

The invention discloses an air conditioner cold load dynamic prediction method based on the combination of a PSO-BP and a Markov chain. The air conditioner cold load dynamic prediction method comprises the following steps of 1, classifying the air conditioner energy consumption data; 2, carrying out the cold load correlation analysis on ten input variables and output variables at the moment T, such as the outdoor temperature of an air conditioner cooling load at the moment T, the outdoor temperature at the T-1 moment, solar radiation quantity at T moment, the solar radiation quantity at T-1 moment, solar radiation quantity at T-2 moment, relative humidity at T moment, outdoor wind speed ag T moment, cold load at T-1 moment, cold load at T-2 moment and cold load at T-4 moment, etc.; 3, carrying out load prediction by using a PSO-BP neural network; 4, dividing an error interval by utilizing a prediction result of the PSO-BP neural network, and constructing a Markov probability transfer matrix; and 5, carrying out Markov chain error correction to obtain a final prediction value. According to the method, the energy consumption conditions in the week and at the end of the week are distinguished, the variables related to the cold load are subjected to correlation analysis, the variables with the high correlation are selected as the input variables of the model, the error correction is performed on the combined model, the redundancy and the complexity of the feature model are reduced, and the operation efficiency of the algorithm is improved.

The invention discloses an intelligent assembly process design method based on morpheme division and an artificial neural network. The method comprises two parts of assembly process knowledge expression and assembly process knowledge reasoning, and mainly comprises the following steps of: 1) dividing assembly process knowledge morphemes; 2) constructing, storing and expressing an assembly structure tree; 3) constructing an assembly process knowledge base; 4) calculating and comparing assembly structure tree structure similarity; 5) carrying out multi-constraint matching and retrieval on the assembly process knowledge; and 6) matching the intelligent weight adjustment with the auxiliary knowledge of the artificial neural network training model. According to the method, repeated manual laborof technicians can be reduced; the error-tolerant rate in the process design engineering is improved; the time cost in an assembly process design project is saved, the production efficiency is improved, and meanwhile, in order to establish an assembly process knowledge base, in combination with the reasoning ability of people and support an automatic assembly process design method, more comprehensive and effective guidance can be provided for assembly process design, and the assembly process design efficiency and accuracy are improved.

A piano practice system is provided, which is applied in an electronic device. The electronic device comprises a microphone and a display screen. A projection device is arranged at the bottom of the display screen. The display screen is fixed by a bracket, and the bracket is connected with a piano. The system comprises a storage module, which is used for importing and storing music electronic files for piano practice. An analysis module, configured to analyze the music score electronic file according to notes, timbre and rhythm; The grade selection module is used for selecting the grade mode corresponding to the practitioner before the piano playing practice, including three grades of entry level, primary level and advanced level; An audio recognizing module for recognizing pitch, beat, tone color and volume of sound played by an exerciser when playing a piano repertoire; The calculation and evaluation module is configured to calculate and compare the analysis result of the analysis module and the recognition result of the audio recognition module, and generate corresponding exercise evaluation data according to the comparison result.

The invention discloses a campus platform based on professional education informatization. The platform comprises a student information system, a teaching affair management system, an education platform resource system, a teaching evaluation system, a public information system, a database system and a network security module. With the help of the big data processing, network communication, computer programming, artificial intelligence and other technologies, the information sharing inside and outside the campus and between enterprises can be realized; the self-learning and evaluation of students, the teaching effect feedback of teachers and the butt joint of enterprise demands and student skill improvement are facilitated, the development of personalized teaching and guidance is facilitated, the self-improvement of the students is rapidly realized, the comprehensive reception and management of information are facilitated, and the efficiency and quality of campus management are greatlyimproved.

The invention discloses a group building synchronous construction risk management system based on a BP neural network, and the system comprises a safety risk evaluation database module which is used for storing collected and processed safety risk information data; a project safety risk information input module used for inputting safety risk information data of a to-be-evaluated project; a risk simulation prediction module used for predicting possible risk events of the to-be-evaluated project; a risk deviation correction module used for correcting the deviation of a simulation prediction result; and a risk assessment report output module used for outputting the corrected risk assessment result and the corresponding risk management and control suggestion measures. According to the risk management system, a database is imported into a BP neural network for training learning and simulation early warning, and a prediction result is corrected in combination with feedback information acquired by various sensors on site, so that a set of complete safety risk management system with group building synchronous construction characteristics is constructed; and the reliability of field safety risk management is improved.

The invention discloses an on-line examination management system and method. The system includes a management system used for managing an on-line examination and personnel related with the examination; an information management module for a system administrator to manage information of the personnel related with the examination; a statement management module used for inquiring examination information; and an examination management module used for managing the on-line examination. Compared with a paper examination, the on-line examination management system provided by the invention saves both time and labor, thereby greatly reducing training cost, the system can enable staff to have certain self-cognition for knowledge reserve of posts of themselves, thereby promoting self-learning of staffto improve abilities of themselves, an examination paper is automatically generated according to content of a large number of test questions in a question bank, thereby reducing the workload of examination affair, and improving work efficiency, and the problems of examination cheating and examination paper revealing do not exist, thereby ensuring fairness of the examination.

The invention relates to an electric power engineering fund prediction method and device, belongs to the technical field of electric power engineering, and solves the problem of inaccurate prediction result in the prior art. The electric power engineering fund prediction method comprises the steps: comprehensively analyzing a plurality of historical projects to determine influence factors of electric power engineering fund budget, wherein the influence factors comprise project periods, the number of participants, material usage and equipment usage; collecting influence factors of historical projects and corresponding electric power engineering funds; performing normalization processing on the influence factors, and then dividing the normalized influence factors and the corresponding electric power engineering funds into two sets including a training set and a verification set; constructing a neural network model based on Bayesian formula optimization, and training and verifying the neural network model through the training set and the verification set to obtain a prediction model; and inputting the influence factors of the to-be-predicted project into the prediction model to obtain a fund prediction value. The accuracy of prediction results is improved.

This invention relates to an interactive phonetic-stroke code input method. Wherein, with tone-form input method as the main and the spelling/ stroke input method as the supplement, applying freely on one input state without any switch; if knowing nothing about the Chinese tone-form code, inputting the word with the spelling or stoke input method to learn every code information of the word on selection window, including the right spelling, split result or right stroke order, tone-form code and the source of every code; if does not the spelling, inputting the word with stroke method to see the information about the word. This invention has no coincident code for high-speed touch typing.

The invention provides a method and a system for constructing an efficient quality inspection model of a self-learning and updating training sample, wherein two SqueezeNet Pro supervised learning models are adopted, fast k. The means algorithm acts as a preselected tag; The loss function of each model is optimized by comparing the two data enhancement algorithms. Conducting self-learning and training models; And update the weights of the parent model. The invention effectively improves the identification efficiency or the identification accuracy, improves the speed of the algorithm calculation, does not increase the time complexity of the operation, and can effectively prevent the occurrence of the local optimal situation.

The invention relates to the technical field of man-machine interaction, and discloses a method for evaluating the interaction performance of an interaction system. The method for evaluating the interaction performance of the interaction system comprises the steps that the interaction state of each log in interaction logs of the interaction system is obtained, and each log is marked according to the interaction state; and according to the labeling result of each log, determining a self-learning index of the interaction system so as to evaluate the interaction performance of the interaction system. By adopting the method for evaluating the interaction performance of the interaction system, the dynamic process of improving the performance of the interaction system through self-learning can be evaluated. The invention further discloses a device for evaluating the interaction performance of the interaction system.

The invention discloses a teacher-student architecture noise-label-containing image classification method based on feature space reforming, and belongs to the field of noise label learning and image classification. The success of deep learning mainly depends on a large-scale tagged data set. Most of data sets are manually labeled, so that the labeling accuracy is difficult to guarantee. Therefore, noise tags are inevitable in a real-world data set. Due to the strong learning ability of the deep neural network, the noise label is very easy to over-fit, and the model performance is reduced. In order to solve the problem, the invention provides a teacher-student structure based on feature space reformation, and the teacher-student structure is used for learning noise labels. A teacher model is trained to correct noise labels for student models. A feature space reforming mechanism is applied to guide a student model to carry out self-learning, so that the student model learns more meaningful information. We assess our methods on the three reference data sets. Experimental results verify the effectiveness of the proposed architecture.

The invention discloses a curve envelope fitting method based on a VGG16 network, and the method comprises the following steps: training a neural network by using a data set sample which is provided with a label and is acquired and established by a CCD, and applying the neural network algorithm to an acquired data set to verify the accuracy and calculate the microscopic morphological characteristics of the surface of an optical fiber; creating a read data set through a tenserflow framework, recording a gray value change sequence of the group of images at a certain pixel point (a, b) as X(a, b)(t), supplementing the sequence X(a, b)(t) into a one-dimensional sequence X2(a, b)(t) with the size of 224 * 224 by adopting a cubic Hermite interpolation method, and then converting the sequence X2(a, b)(t) into a two-dimensional image matrix X2(a, b)(m, n); processing and outputting the predicted actual height of the pixel point through a specially designed convolutional neural network, and comparing the actual height of the pixel point with a sample label to enable an error to be within a set threshold range. The application of the neural network enables the algorithm to have better self-learning, self-organizing and fault-tolerant capabilities and excellent nonlinear approximation capability, can improve the accuracy and fault-tolerant capability of the envelope algorithm, and has certain reference significance.

The present invention discloses a method for predicting nitrogen of soil and particularly relates to a method for predicting nitrogen distribution of surface soil. The present invention provides a method which combines wavelet analysis with a neural network model to establish a wavelet neural network coupling model on the basis of remote sensing data and a ground auxiliary variable so as to predict nitrogen distribution of the surface soil. According to the prediction method provided by the present invention, the effect of carrying out spatial interpolation by a small number of sampling points to obtain nitrogen content of soil in a sampling area is realized, so that the sampling points, sampling time and sampling cost are saved, cost of acquiring soil property spatial data is effectively reduced, spatial interpolation accuracy is improved and spatial distribution of nitrogen of the surface soil can be predicted under high accuracy.