12results about How to "Avoid misleading" patented technology

This disclosure proposes a method for synthesizing multimodal data for decision-making attribution explanation in the mining industry, comprising the following steps: First, extracting causal logic from industry texts to construct a structured rule base; second, collecting on-site multimodal data and filtering out data pairs strongly correlated with rules through scenario-based weight modeling and cross-modal association mechanisms; third, using a multimodal generative adversarial network with rule constraint mechanisms and combined with rule mutation technology to synthesize multimodal data with complete logical chains and covering various complex working conditions; finally, ensuring data quality and achieving self-evolution of the knowledge base through a three-level verification and rule feedback mechanism, and storing qualified data in a hybrid storage database. This invention solves the problems of complex causal relationships and scarce high-quality training samples in the mining field, and can automatically and in batches generate logically rigorous and scenario-diverse multimodal data, improving the training efficiency and generalization ability of attribution decision models.

A kind of high reliability anti-reverse connection circuit suitable for redundant dual controller, increase high side MOS gate auxiliary circuit and sensor low side MOS control circuit.When power is connected, auxiliary circuit does not work;When power is reversed, triode T1, T2 is turned on, and then the gate voltage of MOS Q1, Q2 is pulled to the GND of reverse connection, so that the gate-source voltage of Q1, Q2 is lower than threshold value, to ensure the reliable and stable turn-off of MOS on power bus, at the same time, the gate voltage of MOS Q3, Q4 is also pulled to the GND of reverse connection, so that the gate-source voltage of Q3, Q4 is lower than threshold value, to ensure the reliable and stable turn-off of MOS on sensor ground wire.Diode D1 and D2 realize that this part of circuit does not conduct when power is connected, and only works when power is reversed, resistance RH and RL provide base bias voltage of triode and play the role of current limiting.This application realizes the high reliability anti-reverse protection of redundant dual controller system and external common ground sensor through the distributed configuration of high and low side MOS on the input end of automobile main load power supply and the backflow end of sensor ground wire and the design of targeted auxiliary circuit, and the anti-reverse circuit design scheme suitable for redundant dual controller can effectively realize the high reliability anti-reverse protection.

This invention relates to the field of natural language processing, specifically to a method and system for evaluating and selecting training value of corpus for incremental pre-training. The method involves: acquiring the forward inference results of a base model on a general corpus; segmenting the acquired candidate incremental corpus into text segments; performing forward inference on each text segment using the base model to obtain a first predicted probability distribution; performing forward inference on the lightly augmented text segments using the base model to obtain at least one second predicted probability distribution; calculating four evaluation signals for each text segment based on the first and second predicted probability distributions; weighting and fusing the four evaluation signals to obtain a comprehensive training value score for each text segment; and selecting the target corpus from the candidate incremental corpus based on the comprehensive training value scores of all text segments. This invention solves the problem of a single dimension in corpus value evaluation, effectively controls noisy samples, and significantly improves the resource utilization efficiency of incremental pre-training.

The application discloses a kind of anti-degradation multi-sensor fusion SLAM method and system, belong to robot synchronous positioning and mapping technical field.The method includes obtaining original data from camera, laser radar, inertial measurement unit and global navigation satellite system, based on VIG tight coupling optimization unit, original data is preprocessed and initial estimate in front end, obtain global pose, based on LIG tight coupling optimization unit, motion distortion correction and feature extraction processing are carried out by original data and global pose, based on two-stage factor graph optimizer, laser feature and GNSS observation information are fused in two stages, global optimization is executed, and high-precision, globally consistent pose and map are output.The application overcomes the limitation of single sensor, suppresses cumulative error, realizes high-precision, high robustness and global consistency positioning and mapping in large-scale complex scene outdoors.

This invention relates to a method for real-time online detection of wastewater indicators. Before detection, a low-concentration water body is injected into the detection chamber to perform micro-dose detection. By observing whether there is a nonlinear shift in the initial response behavior, it is determined whether there is residual response on the inner wall of the detection chamber. If so, a local delayed detection and pre-rinsing operation is performed to suppress interference responses caused by the chemical memory of the detection chamber. After completing the pre-response suppression, a slow-varying physical stimulus process is applied to the wastewater to be tested, stimulating the release of masked pollutants in the wastewater. The pollutant response signals are continuously tracked, and fluctuating behavior is identified. Based on the changing trend of the response curve, a stable response interval is identified as the sampling point for final detection. For cases with superimposed responses from multiple pollutants, the rise rate, fall rate, and tailing characteristics of the response profiles of each indicator are analyzed to identify the physical differences of multiple response signals and perform physical-level decoupling, improving the accuracy of identifying multiple pollutants.

This application relates to the field of slope monitoring technology and discloses a calibration method and apparatus for highway slope monitoring signals based on deformation characteristics. The method includes: collecting raw monitoring data through a network of monitoring sensors installed on the surface of the highway slope; cleaning the raw monitoring data to remove outliers and noise interference, obtaining a standardized deformation data sequence; extracting deformation features based on the standardized deformation data sequence using a deformation trend recognition algorithm to generate a deformation feature set; converting the dimensions of multiple deformation features in the deformation feature set into calibration signal control dimensions; wherein the calibration signal control dimensions include control amplitude dimension, control frequency dimension, and control phase dimension, and each dimension of the deformation feature is converted into a corresponding calibration signal control dimension; generating a calibration monitoring signal based on the multiple dimensions of the calibration signal control dimensions and outputting it. This achieves effective calibration of slope deformation monitoring signals, providing reliable support for slope stability monitoring.

This application relates to the field of semiconductor technology. Embodiments of this application provide a surge negative charge driving circuit and a switching power converter for suppressing crosstalk, comprising: a driving module and a surge negative charge module; a first terminal of the driving module receives an input signal, and a second terminal of the driving module is connected to the gate node of a silicon carbide MOSFET, used to control the turn-on and turn-off of the silicon carbide MOSFET according to the input signal; a first terminal of the surge negative charge module receives the input signal, a second terminal of the surge negative charge module is connected to a power supply terminal, a third terminal of the surge negative charge module is connected to a ground terminal, and a fourth terminal of the surge negative charge module is connected to the gate node of the silicon carbide MOSFET, used to release negative charge to the gate node during the turn-off process of the silicon carbide MOSFET when the drain voltage of the silicon carbide MOSFET rises, causing Miller current to flow back to the gate node, in order to neutralize the Miller current. Highly efficient crosstalk suppression is achieved through the physical method of charge neutralization, without relying on a continuous negative voltage power supply, simplifying the system architecture and reducing costs.

This application relates to a circuit breaker, electronic equipment, and a method for driving the circuit breaker. The driving circuit includes a first suppression circuit and a second suppression circuit; the first suppression circuit is connected to the control electrode of a first switching transistor and a controller external to the circuit breaker, respectively; the second suppression circuit is connected to the control electrode of the second switching transistor and the controller, respectively; the first suppression circuit is used to transmit the control signal from the controller to the second switching transistor and suppress the crosstalk current generated during the turn-on and turn-off of the second switching transistor; the second suppression circuit is used to transmit the control signal from the controller to the first switching transistor and suppress the crosstalk current generated during the turn-on and turn-off of the first switching transistor. Using this application, bidirectional suppression of both positive and negative crosstalk can be achieved, enhancing the reliability of the circuit.

The application relates to a robot reinforcement learning control method, system, device and storage medium. The method comprises the following steps: obtaining real interaction data of a robot control task, constructing an environment dynamics model based on a probability neural network set by using the real interaction data, and capturing the uncertainty of a real environment and generating virtual data by using the environment dynamics model; combining the virtual data with the real interaction data to generate mixed training samples, training a parameterized policy network and a value network by using the mixed training samples; performing model-based reinforcement learning on the policy network and the value network, introducing a dynamic relative entropy regularization mechanism in the policy optimization process of the reinforcement learning, and generating an optimal policy of the robot control task. By introducing the dynamic relative entropy regularization constraint mechanism in the model-based policy optimization framework, the cumulative influence of the environment model prediction error is inhibited, and the training instability problem caused by the model bias is effectively alleviated.

This invention provides a retrieval enhancement generation method and system based on adaptive information bottleneck. The system includes a multi-strategy retrieval module, a joint embedding encoding module, a compressed-sensory reordering module, an adaptive information bottleneck compression module, and a large language model generation module. The method includes the following sub-steps: S1: Obtain candidate text blocks; S2: Perform compressed-sensory reordering on the initial candidate text blocks to filter highly relevant text blocks; S3: Perform vector encoding on the context text and query statement q; S4: Perform adaptive information bottleneck compression; S5: Generate the question-and-answer result for query statement q. This method dynamically compresses the retrieval results through adaptive information bottleneck, reducing the context occupation of the large language model, lowering inference costs, shortening response latency, and improving system operating efficiency; it filters highly relevant text blocks through compressed-sensory reordering, filtering irrelevant information, avoiding noisy content from misleading the model, and improving the accuracy of the answer.

This application discloses a digital management method and system for hospital nursing quality indicators, relating to the field of medical and nursing quality information technology. The method includes the following steps: setting assessment rules for evaluating nursing process quality indicators; acquiring patient nursing records and constructing a patient-centered event time series; based on the assessment rules, filtering nursing events related to the corresponding nursing process from the event time series; dynamically adjusting the strictness of the assessment rules based on the identity information of nursing staff and the clinical status information of patients; verifying the event time series according to the adjusted assessment rules and obtaining verification results; determining the compliance of the nursing process based on the verification results, and generating a compliance conclusion and audit report for the nursing process. This application can intelligently distinguish between genuine nursing quality defects and discrepancies in record appearances caused by differences in personnel and context, solving the problems of fragmented processes, isolated data, and delayed assessment in traditional nursing quality management.

This invention relates to the field of artificial intelligence technology, specifically a method, system, device, and storage medium for meta-information-driven synthesis training of large-scale Chinese painting models. The method is executed by a computer device and includes the following steps: Step S1: Data acquisition and parsing; Step S2: Meta-information-driven structured data synthesis; Step S3: Multi-level quality filtering; Step S4: Progressive supervised fine-tuning; Step S5: Reinforcement learning based on a generative reward model. The main objective of this invention is to provide an integrated method and system for meta-information-driven synthesis training of large-scale Chinese painting models, capable of generating high-quality professional data at low cost, effectively suppressing hallucinations, and possessing honest expression capabilities.