73results about How to "Improve fault recognition rate" patented technology

The invention discloses a wind turbine gear box fault recognition method. The method comprises the following steps: historical wind turbine gear box operation data in a certain time range are acquired; autocorrelation analysis is adopted for carrying out wavelet de-noising processing on the historical data; through fast Fourier transform, time domain and frequency domain characteristic parameters in the historical data after de-noising are extracted; a kernel principal component analysis method is adopted to carry out dimensionality reduction on the characteristic parameters, and several nonlinear principal elements with the maximum variance cumulative contribution rate are extracted; the nonlinear principal elements extracted by the historical gear box normal operation data are used for building a normal model, a support vector machine is used for training to guide the nonlinear principal elements extracted by later gear box operation historical data to the model after training, and thus, the gear box fault is recognized. The vibration signal processing ability is improved, and an important role is played in gear box fault recognition.

A rolling bearing fault on-line detection and state assessment method is disclosed. The method comprises the following steps: twelve dimensional dimensionless parameters are extracted; the twelve dimensional dimensionless parameters comprise six dimensional time domain statistical parameters, three dimensional frequency domain statistical parameters and three dimensional dimensionless parameters in a small wave envelope spectrum; standardized reconstruction characteristic vectors can be obtained; whether a rolling bearing malfunctions is determined, and a state of the rolling bearing is assessed. Via the rolling bearing fault on-line detection and state assessment method, the twelve dimensional dimensionless parameters which can be used for effectively representing the state of the rolling bearing can be automatically extracted, the twelve dimensional dimensionless parameters are subjected to decorrelation and standardization operation, standardized reconstruction characteristic vectors that are distributed to form a hypersphere with an original point being a sphere center, and fault detection and state assessment of the rolling bearing can be realized via 2-norms of the standardized reconstruction characteristic vectors; difficult problems of long on line training time, low efficiency, and hard-to-obtain fault samples and the like of a rolling bearing state assessing model can be solved.

The invention provides a remote real-time monitoring and fault diagnosis system of a mine fan based on an LabVIEW platform. The system comprises hardware of an industrial control computer, a terminal board, a PCI data acquiring card, signal conditioning equipment and a vibration speed sensor, and comprises software of a data acquiring module, a data memory and management module, a data display module, a history inquiring module, a fault diagnosis module and a network distribution module. The system realizes real-time acquisition of fan vibration signals, dynamic monitoring of equipment operation states, the setting, display and warning prompt of equipment operation normal values, warning values and stop range values and the storage and inquiry of such related data as various fault data and diagnosis results, and adopts an analytic hierarchy process algorithm to obtain various fault generation probabilities. A wire communication network is built; and the remote real-time monitoring and diagnosis is realized by using a share variable technology. The system can improve the big data processing capacity, and realizes the local and remote monitoring and fuzzy fault diagnosis of the fan operation states.

The invention discloses a system and method for monitoring damaged on-off quantity of a surge protection device and ground resistance online in the real time. A measuring system comprises a server, one or a plurality of remote monitoring terminals and one or a plurality of data collecting systems. The server is respectively connected with various remote monitoring terminals and all data collecting systems. Each data collecting system comprises a damaged on-off quantity data collecting module and a ground resistance collecting module. Each damaged on-off quantity data collecting module comprises m row-connecting lines and n line-connecting lines. Each ground resistance collecting module comprises a micro-processor module, a constant current source injection module, a sampling voltage reading module and a ground resistance measuring software module. The micro-processor modules are connected with the constant current injection modules, the sampling voltage reading modules and the ground resistance measuring software modules respectively. The system and method can guarantee effective judgment of damaged on-off quantity of all surge protection devices and simultaneously can measure the ground resistance in the real time precisely. The system and method is especially suitable for railway signal lightning protection.

The invention provides a planetary gear case intelligent fault diagnosis method based on a built-in encoder signal. Firstly the built-in encoder signal of the output shaft position of a planetary gearcase is read by using an encoder data acquisition card to obtain the angular position signal of a test shaft, then the instantaneous angular acceleration signal of the planetary gear case is acquiredby using a polynomial fitting method, and instantaneous angular acceleration signal is segmented so as to obtain a training set, a verification set and a test set; then a deep convolutional neural network model is constructed; then the deep convolutional neural network model is trained by using the training set, and the parameter of each layer of network is adjusted through the verification set and a batch random gradient descent method so that the final convolutional neural network model is obtained; and finally the test data are inputted to the convolutional neural network model and the fault is identified and classified. The data acquisition program is simplified and the test expense is reduced, and the signal includes lots of state of health information so that extraction of the faultfeatures and intelligent and automatic monitoring and diagnosis can be facilitated by the convolutional neural network.

The present invention provides a method and device for extracting a vibration signal characteristic frequency band. The method comprises a step of extracting the marginal spectrum of a vibration signal, a step of dividing the marginal spectrum into a plurality of window marginal spectrums by using a sliding window and carrying out clustering analysis on the window marginal spectrum sets under the same frequency band in different fault states so as to calculate and generate the clustering effect evaluation index of each frequency band window marginal spectrum set, and a step of extracting a fault sensitive characteristic frequency band according to the clustering effect evaluation index. The vibration signal characteristic frequency band extracted by the method and device of the invention has the advantages of high fault recognition rate and good noise immunity ability quality.

The invention relates to the technical field of fuel cell control, in particular to a fuel cell system control method and device. The device comprises a memory, a processor and a computer program which is stored in the memory and executable on the processor. The processor obtains control signals when executing the program so as to form an original data sequence. The corresponding prediction valueis calculated according to a multi-variable grey prediction model. When the prediction value exceeds the set threshold range, a control strategy for reversely adjusting the fuel cell is controlled. According to the method, the multi-variable gray-scale prediction method is introduced into the fuel cell control method so that that the cross correlation problem of the control signals, which is the fuel cell multi-system variable, can be eliminated, multiple control signals can be predicted simultaneously, reverse control can be performed in time when the prediction value is abnormal, the prediction control of the fuel cell system is realized and the inertia and delay problems of the fuel cell system can be solved.

The invention provides a rolling bearing prediction method and system based on a local binary pattern and a deep belief network. The rolling bearing prediction method comprises the steps of firstly extracting a vibration signal of a one-dimensional rolling bearing, and then converting the vibration signal into a two-dimensional time domain grayscale image; blocking the two-dimensional time domaingrayscale image; respectively extracting local textural features of block images by using LBP<P, R><u2>, and then combining statistical histograms of the block images based on the LBP<P, R><u2> to serve as the input of a DBN (Dynamic Bayesian Network), wherein the DBN performs automatic extraction on deep features of the statistical histograms, and model parameters of the whole DBN are adjusted through DBN forward self-learning and gradient descent based backward propagation so as to obtain the trained DBN; and finally enabling the statistical histogram features, which are obtained through theLBP<P, R><u2>, of the two-dimensional grayscale image of the vibration signal of the rolling bearing with the state being unknown to serve as the input, extracting high-level features capable of reflecting intrinsic information layer by layer by using the trained DBN network, and then inputting a layer-by-layer feature self-extraction into top classification so as to realize fault recognition forthe rolling bearing under the condition of multiple loads and strong noise.

The invention discloses an OLTC fault diagnosis method based on combination of a wavelet packet and a neural network. The method comprises the following steps: (1) applying a vibration detection probeon a top end of a tank wall of an on-load tap-changer, respectively vibration signals produced in the action process under a normal state, a contact loose state, a contact abrasion state and a contact burning state of the on-load tap-changer, and respectively collecting multiple groups of vibration signals under each state; 2) decomposing the OLTC vibration signals into different frequency bandsutilizing a wavelet decomposition principle; (3) computing an energy spectrum entropy of each frequency band, thereby constructing a wavelet packet energy spectrum entropy vector as the input vector of the neural network; and (4) performing fault mode recognition by utilizing the method of combining the confidence coefficient and the neural network. A working state of a transformer on-load tap-changer can be monitored in real time, a real-time fault diagnosis requirement of the on-load tap-changer can be satisfied, the data support and theoretical evidence are provided for the purposeful overhauling, and the manpower, the material resource and time waste are avoided.

A bearing fault detection method based on active learning is disclosed. A vibration method is used to carry out limited time data acquisition on a bearing operation condition in each second. A peak factor, a pulse factor, a kurtosis factor, a pulse factor and a waveform factor of a bearing vibration signal are extracted in a time domain and are taken as characteristic factors for representing a bearing operation state so as to generate a data table. Each bearing is taken as a point and a maximum distance M between the two points is calculated, each one is taken as a center, 0.1M is a radius, and the point in a circle is taken as a density point, and a density rho is determined. The density is compared through a traversing method, and the distance with the closest point which is greater than the density is delta. According to a main-dominant relation acquired from the delta, a spanning tree is drawn. The first N points with the largest rho*delta value are selected as classification center points so as to carry out classification, and a training set is used to represent a classification label. Label verification is performed, and after the data corresponds to the label, a fault typecan be determined. Classification is accurate, and a bearing fault can be effectively detected and diagnosed.

The invention provides a driving motor fault diagnosis model construction method based on intra-class feature transfer learning and multi-source information fusion, and the method comprises the steps:firstly proposing an improved hierarchical transfer learning method MSTL, which considers the neighbor relation between intra-class samples, maintains the local manifold structure of intra-class data, also can improve the separability of the domain data subjected to transfer learning to different categories, so that the adaptability of the fault diagnosis model to different distribution domain samples is improved; meanwhile, the feature set dimension can be reduced, and the fault diagnosis performance of the fault diagnosis model under variable working conditions is improved. Besides, aimingat the problem that a certain uncertain factor exists in a signal acquired by a single sensor, the D-S evidence theory is adopted to carry out driving motor multi-source information decision-making layer fusion, and secondary D-S evidence fusion is carried out on diagnosis results of vibration and current signals on a model. According to the feature transfer learning method MSTL and the multi-source information fusion diagnosis model provided by the invention, the fault diagnosis accuracy can be improved, and the method has a certain practical value.

The invention provides a server alarm identification system based on a machine vision technology. According to the invention, the alarm automatic identification function of a server, a memory, a network switch and other devices in a data center is realized by using the robot. Mainly by means of the characteristics of high speed and low time delay of a 5G mobile network, equipment information, such as a server, a memory and a network switch, acquired by a robot is analyzed and processed through a cloud control technology, and the information is fed back to the robot to execute corresponding operation, so that automatic identification and alarm of the equipment are realized. Various equipment, power and environment subsystems of the whole machine room are integrated on a comprehensive monitoring and management platform, and maintenance personnel can monitor any equipment of the machine room at any time and any place through a simple and easy-to-use graphical user interface of the monitoring platform to acquire required real-time and historical information and perform efficient global event management. Due to the application of the inspection robot, the fault recognition rate, timeliness and high efficiency of equipment such as a server are greatly improved.

A method for identifying faults of distribution network based on improved multi-class support vector machine includes such steps as collecting three-phase current and bus zero-sequence voltage simulation waveform data of main transformer low-voltage side one cycle aft fault as input signal, 2, performing wavelet decomposition processing on that input signal, and reconstructing a low-frequency component to obtain a reconstructed signal; 3, extracting the characteristic vector of the reconstructed signal by the method of obtaining the root mean square distance and the Euclidean distance; 4, constructing a multi-level support vector model, and searching the optimal parameters based on the radial basis function to obtain the trained multi-level support vector machine model; 5: inputting the feature vector of the reconstructed signal to the trained multi-level support vector machine model to obtain the fault classification. The invention is based on the improved multi-classification supportvector machine, and greatly improves the accuracy rate and the fault recognition rate of the fault classification.

The invention discloses a GHMM's NPC photovoltaic inverter fault diagnosis method based on improved hidden Markov model, which belongs to the field of power electronic application and fault diagnosis technology. This method combines the hidden Markov model (HMM) with genetic algorithm before being introduced to the fault diagnosis of NPC photovoltaic inverter. When the HMM is applied for fault diagnosis, the selection of the initial value of the model could reach local optimum and affect the diagnostic accuracy. Aimed at solving this defect, the genetic algorithm is introduced, which means the HMM is combined with genetic algorithm (GHMM) for inverter fault diagnosis. Compared with the existing photovoltaic inverter fault diagnosis methods, the number of the iteration steps of the training model in the invention is far less than conventional method, and the training time is short; the recognition speed is fast. Compared with the only use of the HMM, the GHMM further improves the recognition accuracy. In the dynamic process of system operation, the GHMM monitors and diagnoses, and the model achieves the global optimum, greatly raising the fault recognition rate.

The invention discloses a rolling bearing fault diagnosis method based on multivariate time-shifting multi-scale permutation entropy, which belongs to the technical field of fault diagnosis. The method comprises the steps of firstly collecting an original fault vibration signal of a to-be-diagnosed object, extracting a multivariate time-shifting multi-scale permutation entropy value of the original fault vibration signal, conducting dimensionality reduction on the multivariate time-shifting multi-scale permutation entropy by adopting a Laplacian score method, dividing the dimension-reduced fault feature samples into a plurality of training samples and test samples, training a multi-fault feature classifier of the support vector machine based on bat algorithm optimization by adopting the plurality of training samples, classifying the test samples by adopting the trained multi-fault feature classifier, and finally identifying the fault type and the degree of the object according to the classification result. The method has relatively high innovativeness in processing multi-channel signals of signals acquired by the sensor, and has relatively high recognition degree in a fault recognition process.

The invention discloses an angle similarity stage division and monitoring method in a microbial pharmacy process. In order to better process multi-stage characteristics in the penicillin fermentationprocess, an effective fault monitoring model based on a multi-stage division method is established. The method comprises two stages of off-line modeling and on-line monitoring. The off-line modeling comprises the following steps: firstly, expanding three-dimensional data of a fermentation process along a time axis; dividing the data into C0 sub-periods; and then establishing respective KECA modelsby using the sub-period data, finally calculating T2 and SPE statistics of the data, and determining the control limit of the statistics in each period. The on-line monitoring comprises the steps ofprocessing newly collected data according to a model, dividing the data into sub-periods, calculating the statistics of the data, and comparing the statistics with a control limit to judge whether theproduction process is faulty or not. According to the method, the multi-stage characteristics of the intermittent process are fully considered, and the fault monitoring accuracy is satisfactory.

The invention discloses a system and a method for switching value data collection of a lightning protection device. The system comprises a data collection unit and a data processing unit which are arranged in matrix mode. The lightning protection device comprises m rows of and n lines of switching values which are arranged in the matrix mode, and two ends of the switching values are respectively connected with corresponding row connection lines and line connection lines. The data processing unit comprises a lightning protection device failure scanning module, and potentials of the m row connecting lines and the n line connecting lines are arranged to be high potentials. The potential of a line connecting line in the n line connecting lines is sequentially arranged to be the low potential. If the low potential occurs in the m row connecting lines, and lightning protection device failures occur in a cross area between the ith row connecting line Ki with the potential to be low and present jth line connecting lines Ij. The system and the method can reduce connection cables of the switching values through matrix wiring. The system and the method can achieve line self detection and lightning protection device detection, well confirm whether the failures are line self failures or the lightning protection device failures.

The invention discloses a GA-BP maglev train fault detection method based on threshold determination. Characteristic parameters of data obtained by a sensor are normalized and then used as BP input, avariance threshold and a change rate determination fault are used as output, and an optimal network weight and a threshold are optimized by adopting a genetic algorithm, so that the detection precision of the BP neural network is improved. It can be accurately diagnosed whether a maglev train has a vibration fault or not, and the problem of false alarm is avoided.

The invention discloses an OLTC fault diagnosis method based on fuzzy clustering. The method comprises the steps of (1) attaching a vibration detection probe to the top end of a box wall of an on-loadtap changer, and collecting multiple groups of vibration signals generated in an action process by the on-load tap changer in each of a normal state, a contact loose state, a contact wear state and acontact burnout state; (2) denoising the vibration signals by using a wavelet packet threshold method; (3) extracting characteristic quantities for the denoised vibration signals; and (4) performingfault identification by the fuzzy clustering. According to the method, a working state of the on-load tap changer of a transformer can be monitored in real time, and the requirement for real-time fault diagnosis of the on-load tap changer of the transformer is met. A data support and a theoretical basis are provided for purposeful maintenance, so that the waste of manpower, material resources andtime is avoided.

The invention provides an urban rail transit train safety key system identification method. The method includes the steps that the fault number and harm types of all sub systems of an urban rail transit train are collected, the equivalent weight fault number of all the sub systems of the urban rail transit train is computed, the accumulative frequency is computed, an equivalent weight fault accumulative frequency curve is drawn, the catastrophe point of the curve is solved, and thus a safety key system of the urban rail transit train is identified.

The invention discloses a visualized warning device which comprises a signal acquisition module, a signal processing module and an alarm display module. The signal acquisition module is used for acquiring monitoring signals, the signal processing module is used for processing the monitoring signals to obtain corresponding diagnosis signals, and the alarm display module is used for displaying corresponding alarm signals according to the diagnosis signals. The invention further discloses a visualized warning system and a hydraulic station visualized system. The visualized warning device, the visualized warning system and the hydraulic station visualized system have the technical advantages that the acquired monitoring signals can be transmitted to the signal processing module by the signal acquisition module mounted at positions where data monitoring is required, the diagnosis signals can be formed after the monitoring signals are processed by the signal processing module, the corresponding alarm signals can be displayed by the alarm display module according to the diagnosis signals, the positions with faults can be judged by staffs according to the different alarm signals, and accordingly the fault recognition rate of the staffs can be obviously increased.

The invention discloses an OLTC (On-Load Tap Changer) fault diagnosis method based on LCD (Local Characteristic-scale Decomposition) and a permutation entropy. The OLTC fault diagnosis method comprises the steps of: (1) sticking a vibration detection probe at the top end of a box wall of an OLTC, respectively acquiring vibration signals generated in the actuating process of the OLTC in a normal state, a contact loosened state, a contact worn state and a contact burn-down state, and denoising multiple groups of vibration signals in each state, which are respectively acquired; (2) carrying out adaptive decomposition on the denoised vibration signals by utilizing an LCD method, and selecting a plurality of ISCs (Instrinsic scale components) in the front as principal components; (3) calculating permutation entropies of the selected ISCs, and forming a characteristic vector; and (4) inputting the characteristic vector into an established SVM classifier to carry out working mode identification. According to the method, the working state of the OLTC of a transformer can be monitored in real time, and requirements for real-time fault diagnosis on the OLTC are met. Data support is providedfor purposive overhaul, and waste of labor, material resources and time is avoided.