39results about How to "Improve goodness of fit" patented technology

The invention discloses a wind generator set engine room transfer function subregional fitting method. The method comprises the steps of 1, determining a target set, and collecting running data of a wind measurement mast and the target set within the same time; 2, determining the hub height H of the target set and calculating average wind speed data of the wind measurement mast at the height H; 3, utilizing IEC standards and the average wind speed data calculated in the step 2 as the basis, and removing abnormal running data in the step 1; 4, calculating a transfer function of an engine room of the target set, and correcting the transfer function, wherein the step 4 comprises the substep 4-1 of conducting fitting on a low wind speed area and a high speed wind speed area which are obtained by using the rated wind speed value as the limit and the substep 4-2 of calculating a fitting function order of the low wind speed area and a fitting function order of the high wind speed area respectively. By correcting the wind speed of the engine room, the wind speed of free flow at the center of a wind generator set wind wheel is obtained, and a power curve of the performance of the set can be more objectively reflected.

The invention relates to an image processing method and device, electronic equipment and a storage medium, and the method comprises the steps: carrying out the feature equalization processing of a sample image through an equalization sub-network of a detection network, and obtaining an equalization feature image of the sample image; Performing target detection processing on the equalization feature image through a detection sub-network to obtain a prediction area of a target object in the equalization feature image; Determining the cross-to-parallel ratio of each prediction area; Sampling a plurality of prediction areas according to the cross-to-parallel ratio of each prediction area to obtain a target area; And training the detection network according to the target area and the labeling area. According to the image processing method disclosed by the embodiment of the invention, feature equalization processing is carried out on the target sample image, information loss can be avoided,and the training effect is improved. Moreover, the target area can be extracted according to the cross-to-parallel ratio of the prediction area, the probability of extracting the prediction area difficult in the determination process can be improved, the training efficiency is improved, and the training effect is improved.

The invention discloses a traffic accident severity prediction method applied to a regional road network. The method comprises the following steps: 1, acquiring and preprocessing regional road networktraffic accident data; 2, based on the regional road network traffic accident data, establishing a potential category analysis model; 3, respectively establishing a CART decision tree model for eachsub-category according to a potential category analysis result; and 4, respectively establishing an accident severity model (considering independent variables and interaction items) based on binary logistic regression for each subcategory, and taking intersection points of sensitivity and specificity curves as model prediction classification thresholds. According to the method, the adverse effectof accident data heterogeneity on an analysis result can be reduced, the problems that a traditional traffic accident severity prediction model ignores interaction items and the comprehensive prediction effect of unbalanced data is poor are solved, and the prediction precision and goodness of fit of the accident severity model are improved.

The invention discloses a space filter value-based logistic regressive landslide hazard analysis method. Aiming at analytical investigations of landslide hazards, a space filter value thought is imported into a common logistic regression model so as to design a regressive landslide analysis algorithm which comprises the steps of non-landslide site selection, disaster-inducing factor obtaining andgrading, adjacent matrix construction, feature value and feature vector calculation, stepwise regression feature vector selection and regressive modeling. According to the method, the problem that themodel precision is not high as logistic regression models are influenced by spatial autocorrelation between variables can be solved. By adding filter value operators which are constructed by utilizing selected feature vectors into logistic regression models, autocorrelation influences of residuals can be effectively filtered, the fitting goodness and prediction correctness of the regression models can be improved and correct simulation and prediction for landslide hazards can be realized.

The invention discloses an eddy current sensor machine tool online calibration device and method. The eddy current sensor machine tool online calibration device comprises a sensor fixing device and ameasured conductor fixing device. A sensor bracket in the sensor fixing device is fixed to a machine tool workbench through a holddown bolt, and a probe of an eddy current sensor is fixed to the sensor bracket through a fixing nut. A connection column of the measured conductor fixing device is clamped by an elastic collet in a cutter handle, a spring clip is fixed to the connection column througha nut and a bolt, and a measured conductor is clamped by the spring clip. The a measured conductor is controlled by a numerical control machine tool to get away from or close to the probe of the eddycurrent sensor, one time of the forward stroke and one time of the backward stoke are conducted, and when the machine tool is moved by deltax every time, a group composed of nine point data are measured; for a group of data, a current value of the center position is obtained through a secondary bilinear interpolation method; and for the non-linear section and the linear section of the eddy currentsensor, the relation between displacement and the center current value of the linear section are obtained through a least square method. The eddy current sensor machine tool online calibration deviceis simple in structure, convenient to operate and small in error.

The invention discloses a metro accident delay time prediction method based on a maximum likelihood regression tree. Since there may be interaction among a plurality of variables causing subway delay,the present invention utilizes the constructed bivariate splitting method to establish the maximum likelihood regression tree (MLRT) model to describe and analyze subway accident delay. The MLRT model based on Hong Kong Metro accident data from 2005 to 2012 consists of 13 leaf nodes, each leaf node is assigned a logarithmic logistic distribution accelerated failure model (AFT). The results show that the two-factor split maximum likelihood regression model is better than the traditional AFT model and the single-factor split maximum likelihood regression tree model. On this basis, the method ofthe invention can accurately predict the subway delay accident and the probability that the subway accident delay exceeds the maximum bearing range, as the important basic information to remind passengers to replan the journey.

The invention discloses a device and a method for predicting sewage blockage of a reverse osmosis membrane. The method comprises the following steps: the corrected water production flux J of a reverse osmosis membrane is controlled to be unchanged, the operation pressure delta P0 of a water sample at an initial moment and the operation pressure delta P at a moment t are measured, the standardized pressure delta P/delta P0 at the moment t is calculated, and the time change rate of the standardized pressure at different moments is calculated; and the standardized pressure delta P/delta P0 at the moment t is fitted through a membrane sewage blockage model, and the specific process is that if the time change rate of the standardized pressure delta P/delta P0 is increased along with the increase of the filtering time t, a first model (shown in the specification) is adopted, wherein t is the filtering time; delta P is the operation pressure at the moment t, delta P0 is the operation pressure at the initial moment, delta P/delta P0 is the standardized pressure; j is dimensionless flux; delta pi N is osmotic pressure standardized by the initial operation pressure; alpha is a compensation coefficient; k is a resistance increasing coefficient; and b is a model parameter related to the filtering initial resistance. The method can accurately simulate and predict the change of the standardized pressure along with the operation time, and is high in goodness of fit.

The invention discloses an oil paper insulation performance analysis method and device considering a non-uniform medium. An existing oiled paper insulation equivalent model cannot accurately reflect the polarization condition of a non-uniform medium interlayer. The invention discloses an oil paper polarization equivalent model considering interlayer polarization. The polarization response processof actual oil-paper insulation can be better reflected, a parameter identification method of an equivalent model is provided based on frequency domain dielectric response actual measurement data of the oil-paper insulation, numerical approximation of an original spectrogram is achieved, and goodness of fit is introduced to serve as an index for evaluating the accuracy of the parameter identification method.

The invention discloses an SAR image ship CFAR detection method based on bilateral truncation statistical characteristics. The method comprises the steps: 1, obtaining an SAR image, setting a local sliding window composed of a target window and a background window, calculating a logarithm domain mean value, a logarithm domain standard deviation and a variation index of pixels in the background window, obtaining a truncation rule, removing the heterogeneous pixels therein; 2, performing logarithm domain mean value and logarithm domain standard deviation estimation on the reserved real sea clutter by adopting a maximum likelihood estimation method; 3, modeling the real sea clutter gray scale probability density by using logarithmic normal distribution; and 4, establishing a judgment rule according to a given detection false alarm rate, and performing target judgment on the to-be-detected pixel in the target window. According to the method, the heterogeneous pixels in the background window can be effectively removed, so that the detection rate of the ship target under the complex sea condition can be effectively improved on the premise of keeping a relatively low false alarm rate, the calculation efficiency is high, and the method has a relatively good engineering application value.

The invention relates to a method for confirming the storage service life of a resin adhesive and a device and electronic equipment. The method comprises the following steps: by improving an Arrheniusmodel, establishing a secondary relationship of a logarithm of an acceleration service life of a resin adhesive and a temperature reciprocal so as to obtain a relationship of the acceleration servicelife and the temperature of the resin adhesive; calculating an actual storage service life of the resin adhesive at a normal storage temperature. By adopting the method, relatively high fit goodnessof service lives and temperatures can be achieved, and the storage service life of an epoxy resin adhesive is relatively accurate; a nonlinear relationship of temperatures and acceleration service lives can be widened, data fitting can be carried out through an improved model, the problem that the relationship of high secondary items of temperatures and service lives cannot be reproduced by usinga conventional Arrhenius method is solved, and service life estimation errors can be reduced; calculation can be updated at different temperatures, and compared with a conventional method, the methodis not only applicable to service life pre-estimation on the resin adhesive at a low temperature, and is also applicable to situations of high-temperature service life estimation.

The invention belongs to the technical field of solar cells, and discloses a dynamic programming-based InGaN/GaN multi-quantum well solar cell model parameter extraction method, which comprises the following steps of S1, constructing an InGaN/GaN multi-quantum well solar cell model; s2, constructing an I-V characteristic curve expression of the model of the InGaN/GaN multi-quantum well solar cell; s3, constructing an equation set about the short-circuit current Isc, the peak current Im, the open-circuit voltage Voc and the peak voltage Vm; s4, defining a range of an ideal factor n of a series resistor Rs and a diode in the model, and grouping the series resistor Rs and the diode; and S5, on the basis of the step S4, solving values of five parameters of Rs, n, Iph, I0 and Rsh in the model. According to the method, parameter extraction can be carried out on the InGaN/GaN multi-quantum well solar cells with different In components, the filling factor error and the photoelectric conversion efficiency error are small, the curve fitting goodness is high, and the parameter extraction precision is high.

The invention discloses a method for rapidly measuring the polymerization degree of refined cotton, which comprises the following steps of: taking a proper amount of refined cotton sample, treating the sample by adopting a high-temperature drying method, and weighing the absolutely dry sample to measure the polymerization degree of the refined cotton. The method for rapidly measuring the polymerization degree of the refined cotton has the advantages of being rapid, simple, convenient and feasible, capable of saving cost, accurate in measured data, high in operability and the like.

The invention discloses a Weibull function-based pulse wave fitting method, which comprises the following steps: truncating data, namely cutting and dividing acquired human body radial artery pulse wave data by taking a pulse cycle as a unit to obtain a data set of single-cycle pulse waves; and fitting the pulse wave data by using a two-parameter or three-parameter Weibull function mode. The invention has the beneficial effects that the fitting function, namely the Weibull function, suitable for each wavelet of the pulse wave is selected according to the shape of the pulse wave; according to the method, the goodness of fit can be remarkably improved, the number of data which cannot be fitted is remarkably reduced, and meanwhile, the method has universality.