56results about How to "Fewer input parameters" patented technology

The invention claims a fault diagnosis method of a thermal management system of a hybrid electric vehicle, including such steps of selecting the water outlet temperature, pressure and flow rate of each component of thermal management system as input variables, arranging temperature, pressure and flow rate sensors in each loop of thermal management system, and collecting sample data under normal condition and fault condition respectively; using a principal component analysis (PCA) to extract the features of the sample data, and dividing the sample data into training set and test set; designingthe fault diagnosis model of RBF neural network, and optimizing the basis function center, variance and connection weights of RBF neural network by particle swarm optimization algorithm; applying thetrained RBF neural network to the fault diagnosis of HEV thermal management system, and obtaining the fault states of each actuator directly. The method of the invention utilizes the real-time parameters of each component to diagnose the state and the fault of the thermal management system, and can detect the fault of the thermal management system in time and directly determine the fault position.

The invention provides an on-line life calculation method for a battery system. The on-line life calculation method for the battery system provided by the invention comprises the following three steps: inputting the total energy of charging and discharging of the battery system during the entire life process; calculating the total energy of the finished charging and discharging of the batter system at the moment; and utilizing the two energies to obtain an on-line life of the battery system. Compared with the existing calculation method, the on-line life calculation method for the battery system provided by the invention has the following two advantages: few input parameters are needed for calculating the on-line life and the parameters can be collected by a simple device, namely, the battery system can be carried out the on-line life calculation simply by calculating the total energy of charging and discharging of the battery system; and meanwhile, the calculation process is relatively simple and can be realized without the complicated software and hardware devices.

The invention discloses an intelligent multisource assessment method and device for limit mileages of intelligent unmanned vehicles. According to the method, a slope energy consumption value, a temperature energy consumption value, a wind resistance energy consumption value and length index at road sections with different speeds are put forward by combining road environments, influence factors, for unmanned vehicle power consumption, of complicated road environments are quantified, and an unmanned vehicle power consumption prediction model is established; the power prediction of the method isunrelated to batteries, vehicle models and vehicle part performance, and training is carried out by the model in real time, so that the prediction result is not influenced by the using time and life of the batteries and the vehicles; and when the vehicle running speed is high, the wind resistance is the main part of the vehicle energy consumption. According to the method, different power consumption prediction models are selected according to the sizes of wind speeds, and different power consumption prediction models are used for high and low wind speeds, so that input parameters of the modelsare decreased, the calculation time is shortened, the prediction precision of the models can be improved, and the timeliness is stronger.

The invention discloses a hybrid square root cubature Kalman filtering method used for target tracking. The method includes the steps that according to an initialized state vector, an initialized process noise covariance square root factor, an initialized measuring noise covariance square root factor and an initialized state covariance square root factor, a state vector cubature sampling point is calculated; a measuring vector cubature sampling point and a measuring vector cubature sampling point are calculated to obtain measured vectors; a Kalman gain and a posterior state covariance square root factor are calculated; a posterior state vector is calculated. On the premise that filtering precision is unchanged, the filtering efficiency is greatly improved, and the instantaneity is very high.

The invention discloses a method for designing a manifold regularization based semi-supervised classifier for dynamic vision. A training data source of the classifier under the environment of the dynamic vision is as follows: data with class information acquired by a user, positive class sample including objects, negative class sample not including objects, and data without class information randomly acquired by a computer in a video; continuity regular terms of the classifier on a sample data aggregation is defined by utilizing a local linear reconstruction coefficient of the data so as to make the utilization rate of the classifier on the data without the class information be improved. In addition, the definitions of a function complexity and function continuity regular terms are configured to be a form of 1 norm in an optimization problem of solving the coefficients of the classifier, thus the solution of the optimization problem of the classifier is a sparse solution of the coefficients of the classifier, namely the classifier obtained by training is also sparse. Therefore, the real time property of the classifier in a dynamic vision task is improved.

The invention discloses a boiler combustion efficiency predicting method based on a support vector machine incremental algorithm. The boiler combustion efficiency predicting method based on the support vector machine incremental algorithm is characterized by including the following steps: (1) a kernel function is selected; (2) an initial data set is formed; (3) the initial data is pre-treated; (4) a training sample is taken out and tested; (5) a sensitivity coefficient Epsilon is 0.0001, a training precision is 0.00001 and the default values of a penalty coefficient C and a width coefficient sigmate Sigma are respectively 10 and 0.0001; (6) generalization is determined; (7) the optimum coefficient pair is selected; (8) an initial classifier Omega 0, a support vector set and a non-support vector set are obtained through training; (9) sample points which are not in line with a generalized karush-kuhn-tucker (KKT) condition, namely yif (xi)>1 are found out in a newly added sample set X1; (10) a new set is formed; (11) in terms of X, a classifier Omega and a support vector SV are determined; (12) a support vector machine predicting model on boiler combustion efficiency is established. Less input coefficients are input so as to facilitate measuring, a complicated calculation process is removed, training time of working conditions of boiler combustion is shortened, a requirement for online calculation of a distributed control system (DCS) is met and prediction precision is high.

The invention discloses a method for determining the optimal offset angle of a non-sun-oriented solar cell array, and the method mainly comprises the steps: obtaining a calculation input condition, such as an orbit height H and an orbit incident angle range; calculating a current offset angle and a ground shadow field angle under a current orbit incident angle; calculating the output energy of thesolar cell array under the current bias angle and the current orbit incident angle by taking the earth shadow field angle as input; and traversing all bias angles and all orbit incident angles, and evaluating the output energy of the solar cell array to obtain an optimal bias angle. The method is used for selecting the optimal bias installation angle of the satellite solar cell array to obtain the comprehensively optimal solar cell array output energy, traversing all bias angles and the solar cell array output energy under all orbit incident angles, and evaluating the calculation result of the solar cell array output energy to realize the selection of the optimal bias angle.

A constant displacement quality control method of steel rail flame welding heating and post-weld normalizing comprises the following steps: locking an upsetting oil cylinder by controlling a magnet valve and a pilot operated check valve on the stock cylinder oil circuit of the upsetting oil cylinder so that the longitudinal displacement of the steel rail keeps constant; simultaneously, detecting the retaining pressure of the upsetting oil cylinder in real time through a pressure sensor, and then calculating and obtaining the temperature of the steel rail according to the curve of the relationship between the preset welding temperature of steel rail and the longitudinal deformation resistance of steel rail; when the temperature of steel rail reaches the preset welding temperature, controlling a heating device to stop heating, carrying out upsetting pressure retaining shearing, air-cooled cooling and post-weld normalizing. The method is capable of automatically controlling the temperature of gas-pressure welding and pressure retaining and normalizing of steel rail in closed loop, thereby having few control input parameters, simple and convenient operation, accurate control for temperature, strong anti-interference, small error, high precision, good technical uniformity and repeatability, more stable and reliable welding quality, and high one-time welding success ratio.

The invention relates to a method for calculating the unfrozen water content of frozen soil based on the ion concentration gradient of a clay diffusion layer. The method comprises the steps of (1) collecting a soil sample, measuring the particle size distribution of the soil and the humidity at normal temperature, determining the mass percentages of clay, silt and sand in the soil, and taking the measured soil humidity as the initial water content w0; (2) substituting the mass percentages of clay, silt and sand in the soil into a specific surface area estimation model to obtain the specific surface area AS of the soil, wherein the specific surface area estimation model refers to AS = 0.042 + 4.23 Clay% + 1.12 Silt%-1.16 Sand%, and in the formula, Clay%, Silt% and Sand% are the mass percentages of the clay, the silt and the sand respectively; and (3) establishing an unfrozen water content model shown in the description, and calculating the unfrozen water content wu. The method is simple in model expression and accurate in calculation.

The invention discloses a maximum correlation kurtosis deconvolution method without a period. The method comprises steps of firstly, carrying out truncation and mean removal processing of an acquiredvibration signal; performing Hilbert transform of the vibration signal to obtain an analytic signal of the vibration signal; selecting a maximum value point of the part behind a first zero crossing point in the autocorrelation spectrum as an iteration period to carry out deconvolution operation; then, continuously calculating the maximum value point of the part behind the first zero crossing pointin the autocorrelation spectrum of the signal after each iterative filtering as an iterative period to update a filter so as to obtain an optimal filter coefficient; and lastly, performing envelope analysis of the filtered signal, and extracting the fault characteristic frequency from the envelope spectrum. The method is advantaged in that human participation is not needed in the extraction process of the characteristic frequency, automation of fault characteristic extraction and diagnosis monitoring can be achieved, the time is saved, and efficiency is higher.

The invention relates to the technical field of aquaculture, in particular to a method for detecting the feeding state of fishes. The detection method comprises the following steps: acquiring texturalfeatures and shape features of fish schools during ingestion; and substituting the textural features and the shape features of the fish shoal during ingestion into a preset depth forest model to obtain the ingestion state of the fish shoal to be detected. According to the method, the texture features and the shape features of the fish schools during ingestion and the corresponding ingestion states are combined to construct the deep forest model, so that the ingestion states of the fish schools can be judged directly through the texture features and the shape features of the ingestion picturesof the fish schools, the accuracy is high and can reach about 95%, and theoretical support and solution are provided for accurate feeding of fishes.

The digital display instrument reading identification method based on the inspection robot comprises the following steps: firstly, acquiring a digital display instrument picture through a visible light camera carried by an inspection robot holder and an image video management process; then intercepting an external rectangle of the instrument dial plate as a template image; thirdly, intercepting adigital display area and obtaining position information, segmenting numbers in the digital display area, and recording decimal point digits; 4, obtaining a to-be-detected picture and completing initialization; fifthly, template matching and correction are conducted on the instrument position of the picture to be detected; sixthly, intercepting a digital display area picture to perform gray processing, judging the difference between the foreground font and the background, performing adaptive binarization on the digital display area, performing single digital segmentation on the digital displayarea, and refining the outline of the digital display area; and converting a single number into a pixel for storage, generating a data format required by a classifier, and predicting and outputting aresult. According to the invention, the recognition speed and universality of the intelligent inspection robot for the reading of the digital display instrument in an application scene are improved.

The invention discloses an ozonosphere forecasting algorithm based on artificial intelligence. The ozonosphere forecasting algorithm comprises the following steps: S1, establishing an ozone concentration monitoring station; S2, collecting and acquiring historical meteorological data; S3, selecting influence factors; S4, carrying out preliminary prediction on the O3-8 h value of one day; S5, establishing a new hyperchaotic system; S6, establishing an artificial neural network; S7, establishing a chaotic artificial neural network; and S8, performing long-term and short-term forecasting by using the chaos artificial neural network. The research process of a traditional numerical weather forecasting method can be simplified, the traditional numerical weather forecasting method is often complex and high in calculation requirement, and the CANN operation adopted by the method is similar to other neural networks in that the CANN operation does not depend on the complex relation between parameters and output, but depends on continuous changes of weights, therefore, parameters are closely associated with output, and tedious mathematical modeling is avoided.

The invention discloses a method for measuring and calculating the comprehensive cost of a household broadband service. In order to solve the problem that an effective method for network cost calculation and benefit evaluation is lacked in the prior art, the method comprises the following steps: S1, inputting a key information matrix; s2, using a built-in integrated calculation module to calculate an investment, income and cost correlation matrix; s3, calculating total investment, total income and total cost; s4, the investment benefit and the investment benefit rate in the whole life cycle are calculated; and S5, single-user benefit analysis. According to the scheme, upper-layer network investment, indirect investment, indirect income and related cost are calculated, a network benefit model calculation system is complete, the calculation accuracy is high, the calculation granularity is finer, and accurate construction is facilitated.