6results about How to "Strong self-learning ability" patented technology

The invention relates to the technical field of semiconductor test equipment, in particular to a full-automatic high-precision probe station, which utilizes multi-physics field coupling sensing and compensation to remarkably improve the positioning precision: by comprehensively collecting multi-physics field data such as a temperature field and a stress field and combining a coupling algorithm to optimize a compensation strategy, multi-factor interference is effectively counteracted, and the accuracy of the probe station is improved. The positioning repetition precision can reach + / -0.05 [mu] m, and the test requirement of the advanced semiconductor technology is met; a dynamic three-dimensional modeling technology and dynamic adaptability enhancement are achieved, a dynamic three-dimensional modeling system with the size affected by multi-physical field data can update the form of a test target in real time, test deviation caused by target deviation or surface fluctuation is avoided, and the adaptability of equipment to a complex test scene is improved. The automation degree is high, the testing efficiency is improved, the whole process of data acquisition, modeling, compensation and testing can be completed without manual intervention, the calibration process is simplified, the testing period is shortened, and the method is suitable for batch detection scenes.

The application discloses a kind of based on space alignment and timing correction multi-intersection adaptive signal control method, belong to traffic signal control field, including: through the traffic detection sensor deployed in intersection, real-time traffic state data are collected and sent to signal lamp controller;According to road network topology, send cooperative data to downstream intersection;Traffic flow and candidate phase set are constructed;The traffic data associated with candidate phase is aggregated as demand value and executes ordering transformation to obtain combination state;To combination state, execute spatial feature alignment processing and map as state feature vector;State feature vector is input into near-end strategy optimization controller, and output target phase and basic phase duration;The traffic flow information of last phase period on upstream intersection is combined to carry out timing correction to basic phase duration;Drive signal lamp hardware to execute and generate reward feedback and update near-end strategy optimization controller parameter.The application realizes the cooperative optimization of multi-intersection traffic signal, improves the traffic efficiency and adaptive capacity.

This invention discloses a method for determining the elemental content of iron ore using variable importance-artificial neural network-assisted laser-induced breakdown spectroscopy. The method includes the following steps: S1: Collect LIBS spectral data and elemental content data from at least 13 batches of iron ore in at least four categories. Use radiometric analysis (RF) to measure the importance of LIBS spectral features. A variable importance threshold is used to optimize the input variables of the BP-ANN model. The variable importance threshold and the number of neurons in the BP-ANN model are optimized using the determination coefficients and root mean square error of five-fold cross-validation to establish a VI-BP-ANN model. S2: Using the VI-BP-ANN model from S1, input the LIBS spectral data. The model ranks the features according to variable importance and calculates the elemental content in the iron ore. This invention's detection method can rapidly detect the total iron content, calcium content, magnesium content, aluminum content, and silicon content in iron ore.

A temperature noise correction method for CMOS space cameras based on attention mechanisms and LSTM is presented. This method relates to the field of optical remote sensing technology, specifically to the field of temperature noise correction for CMOS space cameras. The core of the method is the use of a multi-level Long Short-Term Memory (LSTM) network with an attention mechanism to explore how temperature changes affect the noise performance of CMOS space cameras under two different operating conditions: dark and bright fields. This deep learning model possesses powerful autonomous learning capabilities, capable of mining and understanding complex noise patterns hidden within massive amounts of data, thereby improving the accuracy of noise identification and calibration. Its key advantage lies in its end-to-end learning approach, automatically learning and establishing a deep, nonlinear relationship between temperature and noise directly from the raw data, without requiring manual intervention for feature extraction or designing complex correction algorithms. This significantly enhances the model's generalization ability and adaptability to unseen data scenarios.

The invention relates to the technical field of automatic control and image recognition, and particularly discloses a hollow rotating platform positioning system based on intelligent image recognition. The system comprises a hollow rotating platform body, a high-precision encoder, a visual reference module with a specific reference pattern, a high-speed image acquisition module, a real-time error resolving module and a dynamic compensation controller. According to the method, a visual reference and image recognition technology is introduced, and an absolute space reference system independent of a platform mechanical shaft system is constructed. The system can directly measure and quantify real-time geometric errors generated by vibration and deformation of a platform during operation by resolving sub-pixel-level offset of an image acquisition module on a rotating part relative to a fixed visual reference in real time, and changes maintenance of positioning precision from dependence on periodic off-line calibration to online real-time sensing and compensation. The problem of accumulative errors caused by calibration parameter failure in a traditional method is solved.

The application discloses a high-speed gas suspension control method based on multi-source air pressure information fusion and neural network, and steps are as follows: 1) collecting real-time data for pretreatment; 2) constructing a dynamic state space model of a fluidized bed reactor; 3) using a discrete state space model to describe a system dynamic process, performing linear dynamic estimation, and obtaining linear state estimation results; 4) using a first BP neural network to perform non-linear residual error compensation on state estimation errors, and obtaining fused fluidized state estimation results; 5) constructing a neural network adaptive double closed-loop control structure based on performance index optimization, calculating control errors, and realizing discrete adaptive PID control; and 6) constructing a system performance index function, realizing online adaptive setting of PID parameters, and adjusting the fluidized state through an actuator. The application belongs to the technical field of chemical catalytic reaction control, and improves the stability and control precision of gas-solid two-phase flow under complex working conditions.