31results about How to "Correction bias" patented technology

The invention discloses a smart factory management platform and management method based on digital twinning, and particularly relates to the technical field of digital twinning, and the method comprises the following steps: carrying out the division and preliminary analysis of a digital twinning smart factory, carrying out the feature extraction and anomaly analysis of components with hidden dangers, and carrying out the feature extraction and anomaly analysis of the components with hidden dangers; calculating a data source quality index and a prediction performance quality index, and evaluating the comprehensive operation quality by using a pre-trained machine learning model; based on the evaluation result, dividing the components into low-efficiency and high-efficiency operation components; judging an optimization mode for the low-efficiency component; performing intelligent regulation and control on the efficient component based on the initial sampling frequency; the method can comprehensively evaluate the operation state of the digital twin model assembly, evaluate the adaptation degree of each assembly and the actual factory assembly, intelligently regulate and control the high-efficiency operation assembly, optimize the low-efficiency operation assembly, significantly improve the resource utilization rate, ensure the stability and reliability of the system under various operation conditions, and improve the reliability of the system. And system faults and shutdown are prevented.

The invention relates to the field of microbial culture management, in particular to a microbial culture management system and method. The method comprises the following steps: calculating growth factors based on the number of microorganisms in a microorganism growth process and microorganism culture environment parameters which are collected in real time, and calculating a predicted value of the number of microorganisms at a future moment; detecting the growth state of the microorganisms based on the predicted value of the number of the microorganisms in the future; the initial adjustment amount of the microbial culture environment parameters is obtained through a multi-parameter nonlinear optimization algorithm, the final adjustment amount of the microbial culture environment parameters is obtained by introducing a synergistic effect mechanism, and the microbial culture environment parameters are adjusted. The problems that traditional microbial culture environment parameters depend on manual monitoring and experience adjustment, so that response is lagged, and adjustment is not accurate enough are solved; the future growth trend of microorganisms is difficult to accurately predict, so that the culture management reaction is delayed; the problems that the microorganism culture environment is imbalanced and optimal management cannot be realized due to the fact that microorganism culture environment parameters cannot be automatically optimized are solved.

The invention discloses a rock burst position early warning method for tunneling in a high-stress environment, and relates to the technical field of tunnel rock burst early warning. The method comprises the following steps that sensors are arranged on three sections A, B and C in a tunnel, the sensors are arranged on each section, the sensors are placed in a sleeve, the sleeve is placed in a drilled hole formed in a rock mass in advance, the sleeve and surrounding rock are coupled into a whole, a sensor tail fiber is fixed to an opening of the sleeve through sound insulation cotton, and a sound insulation cover is added to the outermost layer of the sleeve; the sensor collects micro-vibration information and transmits the micro-vibration information to the monitoring host, the monitoring host filters the micro-vibration information to obtain a micro-vibration event, the micro-vibration information obtained by the sensor is filtered, obtained micro-vibration event data are more real and effective, automatic verification of an early warning result is achieved through the image terminal, and the accuracy of early warning is improved. The accuracy of the early warning result is improved, and the deviation of the early warning result is corrected.

The utility model discloses an external wheel forming die's correction mechanism, including base and support column, the support column symmetry fixed junction in base both sides to set up the correction subassembly in the base upper portion, the correction subassembly is used for heating correction to the external wheel, still include the rotating subassembly of setting up between the support column, the rotating subassembly is used for driving the external wheel and carries out the clamping rotation. The utility model belongs to external wheel forming die technical field, and specifically is the correction mechanism of automatic external wheel forming die of real -time monitoring the deviation situation of external wheel and automatic control correction accuracy.

The invention provides a film nozzle for a free electron laser spectroscopy experiment and a liquid sample environment system for hard X-ray spectroscopy, and relates to the technical field of sample environments of free electron laser experiment stations. The film nozzle provided by the invention comprises a first-stage nozzle and a second-stage nozzle which is nested in the first-stage nozzle and is detachably connected with the first-stage nozzle; a reducing channel and a spraying channel which are gradually closed up but not closed up in the flowing direction of liquid flow are sequentially arranged in the second-stage nozzle, and the cross sectional area of a liquid outlet of the spraying channel is 0.05 mm2 to 0.005 mm2. Through the arrangement, the thin film nozzle can generate a flat ultrathin liquid film of which the liquid flow thickness is less than 20 microns and the influence on the time resolution is less than 66fs, so that the strict requirement of a hard X-ray spectroscopy experiment on the thickness of a liquid sample can be met, and the precision of the time-resolved hard X-ray spectroscopy experiment is greatly improved.

The invention relates to the field of audio recognition, in particular to a humming audio recognition method and device, electronic equipment and a storage medium. According to the embodiment of the invention, for people with unclear mouths and teeth, such as humming of children, the conditions of inaccurate pronunciation and unclear character biting exist. By acquiring the pronunciation data corresponding to each original keyword and correcting the original keyword sequence based on the pronunciation data, keyword deviation caused by inaccurate pronunciation can be corrected. Therefore, the accuracy of subsequently retrieving the target song is improved. And then the target keyword sequence obtained after the original keyword sequence is corrected is combined with the tune features for retrieval, so that the humming intention of the user can be more comprehensively and accurately understood, and the accuracy of finding the target song is improved. And the humming styles and characteristics of different users, including non-standard pronunciation, unclear keywords and the like, can be processed, so that the universality of the embodiment of the invention is improved.

The invention relates to the technical field of artificial intelligence and human-computer interaction, and discloses a rehabilitation training adjustment method and system based on multi-modal feature fusion and a storage medium, and the method comprises the steps: obtaining voice, face and physiological signals, constructing a multi-modal data stream, extracting features, and outputting a speech error classification identifier and a physiological signal quality index; splitting the facial action net displacement into healthy and affected sides, correcting affected side features based on healthy side features, and constructing visual representation; fusing each feature to output a comprehensive state vector; updating a dynamic baseline in a task gap, mapping a state vector, and screening to obtain an effective action set; calculating a composite reward and storing the composite reward in an experience playback pool to update the strategy network; and based on the action set, re-weighting the large model output probability, and generating a target interaction corpus. According to the method, the state sensing precision is improved by correcting the deviation of the affected side through the uninjured side, and a closed loop of self-adaptive adjustment of the rehabilitation difficulty and safe interaction text generation is realized by combining the dynamic baseline and penalty mechanism optimization reinforcement learning.

The invention belongs to the technical field of spacecraft ground test equipment, and particularly relates to a microgravity simulation device for spacecraft component ground test, which comprises a base, an air floatation support mechanism, an ultrasonic standing wave stabilization mechanism, a self-adaptive clamping mechanism, a servo drive error compensation mechanism, a control system and a test platform. The air floatation supporting mechanism provides vertical non-contact supporting force to offset gravity, the ultrasonic stabilizing mechanism maintains horizontal stability, the self-adaptive clamping mechanism is adaptive to parts of multiple sizes, and the servo compensation mechanism corrects deviation in real time. The composite suspension structure is combined with PID-fuzzy closed-loop control, the simulation error is controlled within 4.0%, the defects that existing equipment is low in precision, short in time, prone to pollution and poor in universality are overcome, long-time stable simulation can be achieved, and reliable support is provided for part ground verification.

The utility model relates to the technical field of cylinder cover surface treatment, and disclose a kind of cylinder cover surface treatment frock, including base, the top of base is fixedly provided with two fixed plates, a fixed plate is fixedly provided with controller, another fixed plate is fixedly provided with observation glass window, the top of base is fixedly provided with positioning plate, the side of positioning plate is fixedly provided with first electric push rod, the side of first electric push rod is fixedly provided with clamping plate, the top of fixed plate is fixedly provided with processing frame, camera is fixedly provided on the processing frame, electric slide rail is fixedly provided in the width direction of processing frame.The cylinder cover surface treatment frock, by the cooperation of first electric push rod and clamping plate, can be accurately positioned and stably clamped according to different sizes of cylinder cover, avoid shaking or displacement in the processing process, ensure the uniformity and consistency of surface treatment effect, improve product quality.

The invention discloses an aileron deflection multi-parameter interaction and calibration method and system in an aircraft structural strength test, and belongs to the technical field of aircraft structural strength tests. The method comprises the steps that an aileron deflection control system collects and conditions displacement and angle signals; connecting the signal to a coordinated loading control system; displacement-angle calibration is carried out in the aileron deflection control system; displacement calibration and angle calibration between the double systems are carried out respectively; and emergency signal interaction and synchronous triggering are realized. Through inter-system multi-parameter interaction and two-stage calibration, sensor acquisition errors and inter-system acquisition deviations are corrected, high-precision data synchronization and monitoring of an aileron deflection control system and a coordinated loading control system are realized, and through linkage of digital emergency signals, the safety and control precision of a structural strength test are greatly improved.

The invention discloses a gigabit Ethernet equipment fingerprint classification method based on generalized extreme value distribution compensation. The method comprises the following steps: acquiring historical data basic fingerprints of a plurality of gigabit Ethernet equipment from a historical database; the method comprises the following steps: acquiring a plurality of current data of gigabit Ethernet equipment, and processing the current data in a historical data basic fingerprint acquisition mode to obtain a current data basic fingerprint; a minority oversampling technology is synthesized by using an SMOTE algorithm, and data enhancement is carried out on current data basic fingerprints; after random value extraction undersampling is carried out on historical data basic fingerprints, current data basic fingerprints are added, and a new and old data mixed fingerprint database is obtained; based on the new and old data mixed fingerprint database after generalized extreme value distribution fitting, training a fingerprint classifier by adopting an SVM (Support Vector Machine) method to obtain a fingerprint classifier with robustness for subsequent fingerprint classification; according to the compensation method based on generalized extreme value distribution, the influence of environment change on fingerprints can be resisted.

The invention discloses an overfitting relieving method, system and device based on adversarial learning and data enhancement and a storage medium, and the method comprises the steps: S1, carrying out the pre-training of a living body model according to a first training set, and obtaining a first classification result; s2, performing attribute identification on the first training set by using an attribute discriminator, enabling a condition generation network to generate a first attribute fuzzy sample according to the attribute identification, the random noise vector and the first classification result, and adding the first attribute fuzzy sample into the first training set to obtain a second training set; s3, training the living body model according to the second training set to obtain a second classification result; and S4, enabling the conditional generation network to generate a second attribute fuzzy sample according to the attribute identifier, the random noise vector and the second classification result by using an attribute discriminator, adding the second attribute fuzzy sample into a second training set, and repeatedly training the living body model until the proportion of the attribute identifier in the second classification result accords with a preset range.

The invention provides a lane information detection method. The method comprises the following steps: acquiring a monocular lane image through a monocular camera; inputting the monocular lane image into a lane information detection model, and executing depth information detection to obtain depth information; and executing lane image detection to obtain a two-dimensional coordinate of a lane image; obtaining the depth information and the two-dimensional coordinates of the lane picture which have an association relationship; and performing combination processing on the depth information and the two-dimensional coordinates of the lane image with the incidence relation to obtain the three-dimensional coordinates of the lane information. The method comprises the following steps: acquiring a monocular lane image through a monocular camera, detecting the monocular lane image by adopting a lane information detection model to obtain depth information and a two-dimensional coordinate of a lane image, and combining the associated two-dimensional coordinates of the depth information and the lane image to obtain a three-dimensional coordinate of the lane information. Deviation occurring in lane information detection can be corrected, and the efficiency and accuracy of lane information detection are improved.

This application discloses a radar echo extrapolation method based on temporal motion decomposition and dominant fusion, relating to the field of spatiotemporal sequence prediction technology. The method proposes an innovative network fusion paradigm, specifically including: acquiring implicit long-term / short-term temporal features and explicit global / transient motion features during processing using stacked hierarchical recurrent networks; then, innovatively constructing query vectors using temporal features and key-value vectors using motion features, and fusing them using a heterogeneous cross-attention mechanism to form a fusion architecture "dominated by an implicit state modeling network and referenced by an explicit motion modeling network." Based on this, through a series of collaborative designs such as directional decoupling, reference confidence estimation, and long-term feature constraints on short-term features, noise in long-term / short-term motion features is effectively suppressed, enhancing system stability. This method can improve the problems of blurred texture details and attenuation in high-value echo regions in predicted images.

An adversarial learning and training method, system and device for eliminating overfitting, and a storage medium, the method comprising: step S1, performing initial training on a living body model according to a first training set, and obtaining a first classification result; s2, performing identity identification on the first training set by using a discriminator, enabling a generator to generate a first non-identity sample according to the identity identification and the first classification result, and adding the first non-identity sample into the first training set to obtain a second training set; s3, training the living body model according to the second training set to obtain a second classification result; and S4, using the discriminator to enable the generator to generate a second identity-free sample according to the identity label and the second classification result, adding the second identity-free sample into the second training set, and repeatedly training the living body model to obtain the living body model. And until the proportion of the identity label in the second classification result accords with a preset range.

The application relates to a fine path loss prediction method of a bimodal network correction, and belongs to the technical field of wireless communication.The method is used for solving the problem that an existing method cannot balance the complexity and accuracy of path loss prediction, the method gives a statistical prediction value by using a path loss empirical model, completes common characterization of path loss, then a bimodal network is used to learn specific propagation characteristics of a communication environment from two modes of vision and features, a correction value of the path loss empirical model is obtained, and specific description of path loss at different positions is completed.The application is applied to the field of wireless communication, combines artificial intelligence and statistical methods, and can realize fine modeling of path loss.

The invention discloses a material electromagnetic parameter testing system and method. The system comprises a product conveying device, a material testing device and a control device. The product conveying device comprises a first conveying module and a second conveying module; the first conveying module sucks a material plate from a discharging area and places the material plate at a preset detection position of the second conveying module; the second conveying module is arranged in the center of the material testing device and used for adjusting the material plate placed at the preset detection position to a detection posture and moving the material plate into the material testing device for electromagnetic testing; the material testing device comprises a plurality of pairs of testing antennas with different testing frequencies; the control device is used for controlling the product conveying device to convey the material plate and controlling the material testing device to test the material plate to obtain the electromagnetic parameter of the material plate, the full-automatic electromagnetic parameter testing can be realized, and the testing efficiency and the testing precision are improved.

The utility model discloses a kind of equipment for preparing multistage hole MFI molecular sieve membrane, belong to membrane material preparation technical field.The equipment includes feeding roller shaft, homogenizing mixer, vacuum conveying pump, slit coating machine, temperature control box, water tank, detection system and storage roller shaft in order along membrane material advancing direction.Homogenizing mixer integrates vacuum stirring and defoaming agent adding function, eliminate bubble from source;Temperature control box adjusts solvent evaporation and two-phase compatibility by accurate temperature control, reduces interface defect;Detection system identifies membrane material surface defect in real time and monitors thickness by linear array CCD camera and line scanning light source;Control circuit integrates feedback module, and process parameter is adjusted in real time according to detection signal.The utility model realizes that hybrid molecular sieve membrane is from mixing, coating, phase conversion, solidification, detection to winding Whole continuous, automation production, effectively improve the uniformity, selectivity and stability of membrane product, suitable for large-scale application.

The application discloses a method for correcting and predicting radial section sound pressure of underwater focused ultrasonic sound field, which comprises the following steps: obtaining an axial scanning signal of the underwater focused ultrasonic sound field and extracting a signal envelope; constructing a proportional correction coefficient based on the ratio of the signal envelope to a theoretical axial sound pressure; modeling the change of the proportional correction coefficient with the axial position by using Gaussian process regression to obtain the corrected axial sound pressure; predicting the radial section beam width of the underwater focused ultrasonic sound field at any axial position; decomposing the existing measured radial section into modes, extracting a spatial mode function and a mode weight coefficient, establishing a prediction model of the change of the mode weight coefficient with the axial position by using Gaussian process regression, reconstructing the preliminary radial section sound pressure distribution at any axial position, and correcting the preliminary radial section by using the corrected axial sound pressure and the predicted beam width. The application can realize the prediction of the radial section sound pressure of the underwater focused ultrasonic sound field at any axial position.

The application discloses an online oral language understanding system based on a BERT-BJNN-EMGCRF model, comprising the following steps: S1, constructing a model framework composed of a shared network, a bidirectional joint network and a conditional random field of an end-to-end shielding graph; S2, training the model in S1, updating the parameters of the entire network model, and obtaining a semantic slot filling and intention detection model; S3, building an online system composed of a user end and a server end; S4, importing the semantic slot filling and intention detection model in S2 into the online system based on S3 to form an online oral language understanding system. The application greatly utilizes the associated information between the intention and the semantic slot, sufficiently learns the compatibility implicit connection of the label pairs in the semantic slot and between the intention and the semantic slot, improves the overall performance of the two tasks, and realizes the extraction of the semantic slot and the detection of the intention in the detected dialogue text sentence in an online mode.

The invention discloses a self-supervised deep learning magnetic resonance image denoising method for deviation correction, and the method comprises the steps: obtaining a noisy magnetic resonance image and corresponding noisy magnetic resonance image square data, and constructing a training set and a test set; constructing a self-supervised denoising network; training the self-supervised denoising network based on the training set; and performing square processing on a noisy magnetic resonance image to be processed, inputting the processed noisy magnetic resonance image to the trained self-supervised denoising network, and further obtaining a background mask and a denoised magnetic resonance image. According to the method, only noisy magnetic resonance images are used as a training set, manpower resources consumed for collecting noise-free magnetic resonance images are reduced, deviation caused by non-zero mean noise can be corrected, a more accurate denoising result is provided, and the method has practical application value. During application, extra manual optimization is not needed, and the denoising effect is excellent.

The present invention discloses a smart factory management platform and management method based on digital twins, which specifically relates to the field of digital twin technology, and includes the following steps: dividing and preliminarily analyzing the digital twin smart factory, then extracting features and analyzing anomalies of components with hidden dangers, calculating a data source quality index and a predicted performance quality index, and evaluating the comprehensive operation quality using a pre-trained machine learning model; based on the evaluation results, dividing the components into inefficient and efficient operation components; determining the optimization method for the inefficient components; and performing intelligent regulation on the efficient components based on the initial sampling frequency; the present invention can comprehensively evaluate the operation status of digital twin model components, evaluate the degree of adaptation of each component to the actual factory components, perform intelligent regulation on the efficient operation components, and optimize the inefficient operation components, thereby significantly improving resource utilization, ensuring the stability and reliability of the system under various operating conditions, and preventing system failures and shutdowns.

The invention relates to the technical field of energy scheduling, and discloses a converter gas scheduling method, device and equipment, a storage medium and a product, and the method comprises the steps: predicting the supply and demand information of converter gas in a preset time range based on the obtained converter gas information and production plan data; based on converter gas holder data in the supply and demand information and the converter gas information, establishing a capacity model and a mixed heat value change model of the gas holder, and determining state information of the gas holder under different scheduling strategies according to the capacity model and the mixed heat value change model; and solving an optimal scheduling scheme of the converter gas in a preset time range through a rolling optimization strategy by taking minimization of the total cost as an optimization target and combining preset constraint conditions and state information. A full-chain collaborative optimization framework of generation, buffering and consumption of converter gas is constructed, future cabinet position and heat value changes of the gas cabinet under different scheduling strategies can be predicted, and collaborative optimization scheduling of a converter gas system is achieved.

The present invention relates to the field of aviation data calculation technology, and discloses a method and device for correcting the longitude and latitude of an aircraft based on an ILS approach mode. The method comprises: selecting a flight data file of a target aircraft based on flight information, obtaining the target runway data of the landing airport of the target aircraft in the flight data file, converting the localizer deviation parameter and the glide path deviation parameter into angle units, and calculating the localizer deviation angle and the glide path deviation angle; calculating the distance of the target aircraft from the runway entrance based on the distance from the glide path signal touchdown point to the runway entrance combined with the radio altitude of the target aircraft and the glide path deviation angle, and calculating the corrected target longitude and target latitude of the target aircraft in combination with the runway entrance longitude, runway entrance latitude, localizer deviation angle, and runway true heading angle. The corrected longitude and latitude data are more accurate, better reflect the actual position of the aircraft, and effectively improve the accuracy of the longitude and latitude data recorded by the aircraft during the flight descent and approach phases.

The present invention discloses a method and device for target classification, and relates to the field of computer technology. A specific implementation of the method includes: clustering the targets to be classified using the feature data of at least one dimension of multiple targets to be classified to form multiple categories; obtaining the average value of the feature data of the targets to be classified in each dimension in each category, and determining the average value as the attribute data of the category in the dimension; using the attribute data of each dimension of each category to determine the initial score and skewness of the category; correcting the initial score according to the skewness to obtain the final score of the category; and determining the level of each category according to the final score. This implementation can achieve reasonable and accurate target classification, and since no human participation is required in the classification process, the classification results can be dynamically adjusted in real time according to data changes.

The present invention relates to the field of microbial culture management, and in particular to a microbial culture management system and method. The contents include: based on the number of microorganisms and microbial culture environment parameters collected in real time during the microbial growth process, calculating the growth factor and calculating the predicted value of the number of microorganisms at a future time; based on the predicted value of the number of microorganisms at a future time, detecting the growth state of the microorganism; obtaining the preliminary adjustment amount of the microbial culture environment parameters through a multi-parameter nonlinear optimization algorithm, introducing a synergistic mechanism to obtain the final adjustment amount of the microbial culture environment parameters, and adjusting the microbial culture environment parameters. The problem that the traditional microbial culture environment parameters rely on manual monitoring and experience adjustment, resulting in delayed response and inaccurate adjustment; it is difficult to accurately predict the future growth trend of microorganisms, resulting in delayed culture management response; it is impossible to automatically optimize the microbial culture environment parameters, resulting in an imbalanced microbial culture environment and an inability to achieve optimal management is solved.

The invention provides a vehicle steering control method, a vehicle steering control device and a vehicle, the method is applied to the field of vehicle control, and the vehicle comprises a rear wheel steering control system. Obtaining a first yaw velocity, and predicting a target yaw velocity corresponding to the yaw velocity and the steering instruction; wherein the first yaw velocity is the actual yaw velocity, detected in the previous collection period, of the vehicle; the predicted yaw velocity is the predicted yaw velocity of the previous collection period; based on the first yaw velocity and the predicted yaw velocity, a compensation signal of a feedback compensator in the rear wheel steering control system is obtained; and controlling rear wheels of the vehicle to steer through a rear wheel steering control system based on the compensation signal and the target yaw velocity. According to the method, rear wheel steering can be compensated, and the response speed of the rear wheels during vehicle steering is increased.

The application provides a tumor cell auxiliary positioning system and method based on a composite image, relates to the technical field of medical imaging, and comprises the following steps: a data acquisition module is used for synchronously collecting radio frequency echo signals, ultrasonic images and computed tomography data of a target biological tissue, and generating original fusion data sets; a feature processing module is used for spatial registration and feature extraction on the original fusion data sets, so as to obtain a multi-dimensional feature set; and an identification imaging module is used for extracting multi-frequency band radio frequency echo signal intensity level values of a target region based on the multi-dimensional feature set, generating a tissue radio frequency characteristic diagram through logarithmic compression, and performing pixel-level fusion with a synchronous ultrasonic image to generate a composite image. Through multi-modal data fusion and accurate calibration, the application determines a tumor boundary three-dimensional coordinate in combination with anatomical structure constraints, and improves the accuracy of tumor cell positioning.

The invention provides a clamp floating roller device which comprises safety tongs and further comprises a guide wheel assembly and a clamp floating device, and the clamp floating device is connected with the safety tongs. The clamp floating device comprises a rotary block, a supporting block, a cover plate and a gasket; the supporting blocks are fixedly connected with the cargo carrying table side plates through fasteners. The supporting block is provided with a kidney-shaped hole, the rotating block is provided with a first boss, and the first boss extends into the kidney-shaped hole and exceeds the supporting block. The cover plate is provided with a second boss, the second boss is flush with the gasket, and the second boss is fixedly connected with the first boss through a connecting piece. The rotary block can rotate left and right on the supporting block; guide wheel assemblies are arranged at the upper end and the lower end of the rotary block respectively. The problem that the T-shaped guide rail is easily clamped by the safety calipers in the up-and-down movement process of the cargo carrying table is solved.

The invention provides a periodic time series data analysis and treatment method and product, and relates to the technical field of data analysis and treatment. In the embodiment of the invention, the time series data are cut according to the periodicity of the data, the data are coded according to the occurrence time of the periodic time series data, the coded time series data are clustered, and the time series data are clustered based on the clustered result obtained by clustering the coded time series data. The abnormal data cannot correspond to the time slots of the data of other periods, so that a class cluster can be independently formed, and the abnormal data can be found and eliminated based on a clustering result. For the remaining data after the abnormal data is eliminated, noise reduction can be performed on the data in the slot position at each moment, the deviation of the data corresponding to the slot position at the same moment is corrected, and the purpose of improving the data quality is achieved.