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

The invention discloses a method for predicting the water-phase reaction rate constant of organic matter and a sulfuric acid free radical in a water phase through quantitative structure activity relationship. Based on a compound structure, a molecular descriptor with a structural feature is calculated, and by the adoption of a multiple linear regression method, a QSAR model of an organic compound kso4- is constructed. Establishment of the model strictly confirms to the constructing and using guideline of the QSAR model from organization for economic cooperation and development, the constructed model has a definite application field, contains 197 kinds of organic compounds of different structures, and contains compounds containing carbon-carbon double bonds, carbon-carbon triple bonds, a hydroxyl group, phenolic hydroxyl, a carbonyl group, an aldehyde group, a carboxy group, an ester group, an amide group, nitro, an amino group, a cyano-group, an ether bond, a disulfide bond, fluorine, chlorine, bromine, iodine, arsenium and other radical groups. The mode has good fitting capacity, robustness and predicting capacity, can rapidly and accurately predict the kso4- value of the organic compound, and reference is provided for application of an advanced oxidation process based on a sulfuric acid free radical.

The invention discloses an eddy current sensor machine tool online calibration device and method, which include a sensor fixing device and a measured conductor fixing device. In the sensor fixing device, the sensor bracket is fixed on the machine tool table by pressing bolts, and the probe of the eddy current sensor is fixed on the sensor bracket by the sensor probe fixing nut. The connecting column in the fixture of the tested conductor is clamped by the elastic collet in the tool handle, the spring clamp is fixed on the connecting column through nuts and bolts, and the measured conductor is clamped by the spring clamp. Use the CNC machine tool to control the measured conductor to stay away from or close to the eddy current sensor probe. The positive and negative strokes are performed once each time. Every time the machine tool moves Δx, a set of nine point data is measured; for a set of data, the quadratic bilinear interpolation method is used. The current value at the center position is obtained; the nonlinear section and the linear section of the eddy current sensor are segmented, and the relationship between the displacement and the center current value is obtained by using the least square method for the linear section. The invention has the advantages of simple structure, convenient operation and small error.

The present disclosure relates to an image processing method and device, electronic equipment, and a storage medium. The method includes: performing feature equalization processing on a sample image through an equalization sub-network of a detection network to obtain an equalized feature image of the sample image; Perform target detection processing on the balanced feature image to obtain the predicted area of ​​the target object in the balanced feature image; respectively determine the intersection ratio of each predicted area; according to the intersection ratio of each predicted area, sample multiple predicted areas to obtain the target area ; According to the target area and the marked area, train the detection network. According to the image processing method of the embodiment of the present disclosure, performing feature equalization processing on the target sample image can avoid information loss and improve training effect. Moreover, the target area can be extracted according to the intersection ratio of the prediction area, which can increase the probability of extracting the prediction area that is difficult to determine, improve the training efficiency, and improve the training effect.

The invention discloses a method for detecting ship CFAR in SAR images based on bilateral truncation statistical characteristics. The steps include: 1. Obtain a SAR image, set a local sliding window composed of a target window and a background window, and calculate the pixels in the background window The logarithmic domain mean, logarithmic domain standard deviation and variation index of the logarithmic domain mean and logarithmic domain standard deviation and variation index are obtained, and the truncation rule is obtained to remove the heterogeneous pixels; 2. The logarithmic domain mean and logarithm Domain standard deviation estimation; 3. Use the lognormal distribution to model the gray probability density of real sea clutter; 4. Establish a decision rule based on a given detection false alarm rate, and target the pixels to be tested in the target window judge. The invention can effectively remove heterogeneous pixels in the background window, thereby effectively improving the detection rate of ship targets under complex sea conditions while maintaining a low false alarm rate, and has high calculation efficiency, and has good engineering application value.

The invention discloses a wind-solar combined output modeling method based on a Copula function. The method comprises the following steps: converting the output power of a wind farm and a photovoltaicfield into a wind power output sequence and a photovoltaic output sequence, estimating the core density, and obtaining the edge distribution function of the wind power output and the photovoltaic output; substituting the edge distribution functions of wind power output and photovoltaic output into a Copula function to obtain static correlation coefficients of the Copula function, and using the static correlation coefficients of the Copula function to establish the static model of wind-solar combined output; substituting the wind power output sequence and photovoltaic output sequence into an evolution equation, and utilizing a maximum likelihood estimation method to obtain dynamic correlation coefficients of the Copula function, and using the dynamic correlation coefficients of the Copulafunction to establish a dynamic model of wind-solar combined output. The optimal model is obtained from the static model and the dynamic model of the wind-solar combined output. The invention reducesthe estimation error and makes the model fitting goodness higher.

The invention discloses a method for improving element measurement accuracy based on double pulse and space confinement, which belongs to the technical field of atomic emission spectrometry. In this method, pits or cavities are first made on the surface of the sample to be tested, and then the sample is hit by a double-pulse laser. The first pulse creates a low-pressure environment inside the pit or chamber, and the second pulse excites the sample to create a plasma. The plasma is limited by space during the expansion process, which makes the plasma temperature and electron density significantly increased, and the plasma is more uniform, which is conducive to increasing the stability of the measurement signal, improving the signal-to-noise ratio, and reducing the detection of trace elements in the sample. limit. This method has the characteristics of good calibration and high prediction accuracy.

The invention discloses a road accident frequency prediction method considering the time-varying characteristics of risk factors. The steps are as follows: 1. Collect and process historical traffic accident data and related risk factor data; 2. Establish a Logistic regression model; 3. Use Youden Calculate the classification threshold of the Logistic model by the index method; 4. Calculate the positive predictive value (Positive predictive value) and negative predictive value (Negative predictive value) of the model based on the Logistic model and historical accident data; 5. Use the calculated positive predictive value and The negative predictive value is used to predict the frequency of accidents. The invention can overcome the problem that the traditional accident frequency model cannot reflect the influence of time-varying characteristics of risk factors on accidents, and is beneficial to improving the prediction accuracy of the accident frequency prediction method.

The invention relates to a method, device and electronic equipment for determining the storage life of a resin-type adhesive. By improving the Arrhenius model, the quadratic relationship between the logarithm of the accelerated life of the resin-type adhesive and the reciprocal of the temperature is established, thereby obtaining the resin-type adhesive. The relationship between the accelerated life of the adhesive and the temperature is obtained to obtain its actual storage life at a normal storage temperature. This method has a higher degree of fitting for the life and temperature, and the storage life of the obtained epoxy resin adhesive is also more accurate; the present invention Expand the nonlinear relationship between temperature and accelerated life, and use the improved model for data fitting, which solves the problem that the traditional Arrhenius cannot reproduce the relationship between the high-order term of temperature and life, and reduces the life estimation error; The invention updates the calculation at different temperatures. Compared with the existing methods, the present invention is not only applicable to the life estimation of resin adhesives at low temperature, but also applicable to the situation of high temperature life estimation.

The invention discloses a device and method for predicting reverse osmosis membrane fouling, including: controlling the corrected water production flux J of the reverse osmosis membrane to remain unchanged, and measuring the operating pressure ΔP of the water sample at the initial moment 0 , and the operating pressure ΔP at time t, calculate the normalized pressure at time t to calculate the time change rate of the standardized pressure at different times; use the membrane fouling model to fit the normalized pressure at time t, the specific process is: if the time of normalized pressure When the rate of change increases with the increase of filtration time t, model 1 is adopted: where, t is the filtration time; ΔP is the operating pressure at time t, ΔP 0 is the operating pressure at the initial moment, and is the normalized pressure; j is the dimensionless flux; Δπ N is the osmotic pressure standardized by the initial operating pressure; α is the compensation coefficient; k is the resistance increase coefficient; b is the model parameter related to the initial resistance of filtration. The invention can accurately simulate and predict the change of the standardized pressure with the running time, and has a high fitting degree.

The invention discloses a marine target feature extraction method based on a concomitant variance correction model, which belongs to the technical field of marine remote sensing, including constructing an OceanTD deep learning model, and performing preliminary detection of marine target features based on the OceanTD deep learning model; and then constructing a concomitant variance The model is revised, the variance of the logarithmic distribution function is corrected, the standard deviation and absolute value error indicators are considered, the variance correction item is added, and its formula is derived, which improves the goodness of fit for complex sea conditions and sea clutter; finally, based on the The variance correction model uses the CFAR algorithm to extract ocean target features, and visualizes the extracted ocean target features, which improves the recall rate of ocean target detection and the accuracy of ocean target parameter extraction; the method proposed in the present invention, sea clutter fitting The superiority is significantly improved, the amount of false alarms in target detection is significantly reduced, and the accuracy of feature extraction is effectively improved.