120results about How to "Accuracy is not affected" patented technology

The invention discloses a laser cutting machine with a semi-hollow doll type lathe bed, which comprises side plates and a semi-hollow lathe bed integrally formed by connecting the two side plates through a cross beam arranged between the two side plates, guide rail seats are mounted at the tops of the side plates, a moving mechanism for driving a laser cutting head to move is mounted on the guide rail seats, and a workbench is sleeved in the lathe bed. The plane of the working table is parallel to the lathe bed, the working table comprises a mounting frame, a plurality of transverse frames arranged in parallel are connected to the mounting frame, each transverse frame is provided with a smoke removal assembly used for extracting smoke dust in a short distance, and the smoke removal assemblies are covered with protection assemblies used for protecting and assisting the smoke removal assemblies in smoke dust extraction; the semi-hollow lathe bed is adopted, on the premise that the strength is reliable, the weight of the lathe bed is reduced, and the material cost and the transportation cost are saved; the bed body can be effectively protected through the sleeve doll type structure, the smoke removing assembly is arranged on each transverse frame, smoke generated in the cutting process can be effectively extracted in a short distance, and environmental pollution is reduced.

A positioning method based on a linear landmark effectively solves the problem of inaccurate positioning of the linear landmark, and comprises the following steps: in an application scene, taking a landmark detected by a robot at each sampling moment as an observer of the robot, and obtaining measurement data of a plurality of groups of landmarks; performing aggregation calculation on the measurement data of the plurality of groups of landmarks to obtain a plurality of aggregation results; tile numbers are added to the multiple aggregation results respectively, and multiple target data points are obtained; and according to an improved cross-ray method, matching the target data point with the contour of the landmark to obtain an updated association table of the landmark and the data file.

The utility model discloses mine waste rock recovery pretreatment equipment which comprises a crushing device, a screening device and a sorting device, and a discharging port A of the crushing device corresponds to a feeding port B of the screening device. A first-level screen, a second-level screen and a third-level screen which can screen micro-fine particles, fine particles and coarse particles are sequentially arranged on a drum screen in the screening device from left to right, corresponding discharging ports B of all levels are formed below the screens of all levels, and a first-level sorting bin, a second-level sorting bin and a third-level sorting bin corresponding to the discharging ports B of all levels are arranged in the sorting device. The waste rock crushing and screening device has the functions that waste rock can be continuously and circularly crushed, screened and sorted through the crushing device, the screening device and the sorting device which are communicated with one another, the crushing and screening steps are simplified, the screening precision is not influenced, materials containing metal and the waste rock are distinguished, manpower and material resources can be saved, and the production efficiency is improved. And the production efficiency is improved, and the method is suitable for industrial production.

The invention provides a transformer fault determination method, device and system. According to the method, iterative clustering is carried out on a gas concentration data set by adopting a K-means clustering algorithm based on a genetic algorithm, an optimal clustering center is obtained, a diagnosis sub-model is obtained, the minimum Euclidean distance is determined, and the fault type of the transformer is determined at least according to the size of the minimum Euclidean distance. The purpose of combining a traditional transformer fault diagnosis method and an intelligent transformer fault diagnosis method is achieved, the operation difficulty is reduced on the premise that it is guaranteed that the precision is not affected, and therefore the problems that operation is complex and precision is poor when transformer faults are detected through an existing scheme are solved.

The invention discloses a depth model lightweight deployment method based on algorithm-system cross-layer adaptive optimization, and particularly relates to the field of model lightweight methods.The method comprises the steps that multiple compression operators of a structural layer of a depth model and multiple acceleration operators in the depth model deployment process are determined, taking a plurality of compression operators and a plurality of acceleration operators as a cross-layer search space; determining a target function and a constraint condition according to a target performance index of the depth model; and taking the cross-layer search space, the objective function and the constraint condition as input, and combining a compression operator and an acceleration operator by using a double-layer coevolution optimization algorithm to obtain an optimal technical combination. The optimization requirements of the model and the system level are considered at the same time, the overall optimization effect of the model can be improved, and the model precision is not affected.

This invention provides an active heating method and system for a two-dimensional turntable servo mechanism, belonging to the field of laser communication technology. The method includes acquiring the feedback temperature and motor torque of the azimuth or pitch axis; when the motor torque is zero and the difference between the feedback temperature and the preset temperature is outside the preset temperature range, heating is performed by determining a first heating power based on the first d-axis current and the electrical control angle; when the motor torque is not zero and the difference between the feedback temperature and the preset temperature is outside the preset temperature range, heating is performed by determining a second heating power based on the second d-axis current and the q-axis current, until the temperature difference returns to the preset range and heating stops. This invention utilizes the heat generated by the motor's own current to replace external heating devices, achieving uniform heating from within, effectively reducing deformation and frictional torque fluctuations in key moving parts caused by temperature gradients, and improving the turning table's pointing accuracy and thermal stability under extreme temperature changes.

The embodiment of the invention discloses an airport flight plan strategy delay prediction method based on delay disassembly and assembly, relates to the technical field of air control and planning, and can timely and effectively predict the delay of arrival and departure of an airport flight plan. The airport flight plan strategy delay prediction method based on delay disassembly and assembly is used for obtaining an airport capacity probability graph by using an airport flight plan and historical flight operation data, and obtaining an airport flight arrival and departure delay prediction value according to the airport flight plan, airport operation mode modeling and the airport capacity probability graph. The method is suitable for evaluating the rationality and effectiveness of airport flight time / planning at the strategic stage.

This utility model discloses an aluminum profile production mold with a buffer structure, relating to the field of aluminum profile production technology. The utility model includes a worktable, on which a buffer mechanism and a demolding mechanism are provided. The buffer mechanism includes a connecting plate disposed at the bottom of the worktable. Two fixing blocks are fixedly connected to the left side of the connecting plate, and a sliding rod is fixedly connected between the two fixing blocks. Two sliders are slidably connected to the sliding rod, and spring sheets are fixedly connected to the sides of the two sliders that are furthest from each other. This utility model, by setting up a buffer mechanism, solves the problem that existing aluminum profile production molds are not convenient for buffering the force when merging two molds. The large impact force during merging can damage the mold and its cavity, thus affecting the mold's accuracy and causing deviations in the shape and size of the formed aluminum profile.

The present invention discloses a low-complexity multi-user communication signal recovery and perception positioning estimation method, which generates and transmits an integrated waveform of superimposed pilot and data at the transmitting end, so that the communication signal and the perception signal are linearly superimposed in the symbol domain using a power allocation method in the same time-frequency resource block, and the receiving end uses a designed low-complexity algorithm to process the received signal to realize the function of simultaneous communication and positioning of multiple users; including: using SCMA technology to design an encoder for superimposed symbols at the transmitting end and a decoder for detection and demodulation at the receiving end, and using the UAMP-SBL algorithm at the receiving end to process the received signal to separate the positioning coordinates of the communication signals of multiple users. The present invention is based on the original RIS-assisted communication system, and can use a low-complexity algorithm to simultaneously obtain the communication data of multiple users and accurate user positioning coordinates without adding additional hardware equipment.

The present application discloses a federated learning method, client, server and system based on an attention mechanism. The method of the present application performs local training based on a global model on a local client, and encrypts the trained gradient using the CKKS homomorphic encryption algorithm and uploads it to an aggregation server. The aggregation server forwards the ciphertext gradient to the collaboration server. The collaboration server calculates the weight value of the ciphertext gradient uploaded by the client based on the zoom-click attention mechanism and returns it to the aggregation server, which completes the weighted aggregation of the ciphertext. This can effectively resist poisoning attacks from malicious clients, train high-precision, high-robustness, and high-availability global models in a malicious environment, and ensure that poisoning attacks from malicious clients can be effectively resisted in a malicious environment without affecting the accuracy of the model. At the same time, it provides enhanced privacy protection for client data, and realizes an efficient and secure federated learning process.

The invention belongs to the technical field of fuel cells, and relates to an electric vehicle cluster adjustable capacity evaluation method considering user response willingness, which comprises the following steps: 1, establishing a unified energy storage model of electric vehicle EV individuals; 2, establishing a generalized energy storage (GES) resource general aggregation model of the electric vehicle EV cluster; 3, predicting an electric vehicle EV cluster aggregation model parameter set based on a bidirectional quantum long short-term memory neural network to obtain a theoretical demand response potential; 4, quantifying user side demand response willingness based on a TSK fuzzy system to obtain a demand response willingness coefficient; 5, the actual demand response potential is obtained by combining the data and the model driving result; according to the method, unified energy storage modeling is carried out on individuals by considering heterogeneity of the electric vehicles, and clusters of the electric vehicles are packaged into a generalized energy storage resource general aggregation model, so that the theoretical demand response potential of the electric vehicle cluster is accurately represented by a group of less parameters under the condition that the model precision is not influenced; and the actual demand response potential is obtained by considering the user response intention.

The utility model relates to the field of medical instruments, in particular to a medical positioning frame and medical equipment. The medical positioning frame comprises a mounting bracket and a fixing mechanism; the fixing mechanism comprises a first fixing assembly, a second fixing assembly and a third fixing assembly; the first fixing assembly is movably mounted on the mounting bracket so as to mount and position a bronchoscope; the second fixing assembly is movably mounted on the mounting bracket so as to mount and position the magnetic navigation working channel; the third fixing assembly is movably installed on the installation support so as to install and position an ablation needle. The medical positioning frame can be applied to a bronchoscope ablation operation, can fix a bronchoscope, a magnetic navigation working channel and an ablation needle at the same time, can reduce the labor cost, is higher in stability compared with a manual fixing mode, and does not affect the operation flexibility and the operation accuracy.

The utility model discloses a workbench positioning and locking mechanism which is applied to an exchange workbench of a turning plate machining center, a wedge-shaped guide rail is installed on the upper portion of a working trolley, electric permanent magnetic chucks are symmetrically installed in the middle of the working trolley, a V-shaped eccentric wheel is installed on the lower portion of the working trolley, the workbench is installed on the working trolley in a buckled mode, and a displacement sensor is installed between the workbench and the working trolley. The utility model relates to the technical field of turning plate machining center exchange workbenches, the wedge-shaped guide rail is adopted for positioning, the positioning of an inclined plane and a vertical plane is completed by using gravity, and then the electric permanent magnetic chuck is used for locking, so that the work efficiency is improved, and the work efficiency is improved. The positioning process is stable and reliable, the workbench is free of displacement in the locking process, the locking process is stable, and the positioning precision and the repeated precision are high.

A closed-loop current sensor includes: a pair of openable iron cores, a secondary feedback coil wound outside the iron core, a magnetoresistance sensing unit and a signal processing circuit; the iron core is coated with a multiferroic material film, and the magnetoresistance sensing unit is arranged at the gap of the magnetic circuit formed by the iron core; the signal processing unit includes a signal acquisition circuit, a signal amplification circuit, a current feedback circuit, a voltage follower circuit and a voltage feedback circuit; the signal acquisition circuit is connected to the output end of the magnetoresistance sensing unit, and transmits the voltage signal to the signal amplification circuit, the signal amplification circuit has two outputs, one output to the current feedback circuit, and the other output to the voltage follower circuit; the output end of the current feedback current is connected to one end of the secondary feedback coil, and the other end of the secondary feedback coil is connected to the secondary resistor; the output end of the voltage follower circuit is connected to the voltage feedback circuit, and the output end of the voltage feedback circuit is connected to the multiferroic material film. The sensor of the present invention can achieve low power consumption.

The present invention discloses a high-frequency signal acquisition and processing system for a pole-mounted circuit breaker, which relates to the field of relay protection technology and traveling wave ranging of a power system. The system comprises: a voltage acquisition module: used for acquiring high-frequency traveling wave signals of a distribution line based on a signal acquisition component built into the pole-mounted circuit breaker; a current acquisition module: used for acquiring a secondary-side induced electromotive force signal corresponding to a primary current based on the signal acquisition component, and obtaining a secondary-side induced electromotive force output waveform based on the secondary-side induced electromotive force signal; a signal processing module: used for processing the high-frequency traveling wave signal based on the signal processing component to obtain a secondary output waveform of the high-frequency traveling wave signal, and locating the fault point of the distribution line based on the secondary output waveform and the secondary-side induced electromotive force output waveform. The system can solve the problem that the pole-mounted circuit breaker cannot acquire high-frequency traveling wave signals and that the traditional high-frequency traveling wave signal acquisition device is usually designed as a complete machine, and the line access is cumbersome and costly.

The utility model discloses a working platform for assembly in the field of drilling machines, which structurally comprises an upper platform table, a support structure frame, a lifting working table, a hydraulic pump and an auxiliary fixing mechanism, the upper platform table is fixedly connected to the upper end of the support structure frame, the lifting working table is mounted on the right side of the support structure frame, and the hydraulic pump is mounted below the support structure frame. The auxiliary fixing mechanism comprises an L-shaped side plate, a locking nut, a limiting sleeve and a fixing clamping base, the L-shaped side plate is installed on the left side of the limiting sleeve, the locking nut is installed on the upper surface of the limiting sleeve, the limiting sleeve is installed on the upper surface of the lifting workbench, the fixing clamping base is installed on the lower surface of the limiting sleeve, and the fixing clamping base and the limiting sleeve are matched to fix the L-shaped side plate. An additional auxiliary fixing mechanism can be added at any time according to requirements, an original fixing clamp does not need to be changed, the time cost of assembly is reduced, accessories can be added according to requirements under the condition that the precision of the original fixing clamp is not affected, and the flexibility of assembly work is greatly improved.

The utility model relates to a self-cleaning flow chip with low power consumption. The light emitting diode comprises a substrate; the supporting structure layer is arranged on the substrate; the platinum thermal resistance wire layer is arranged on the supporting structure layer; the first insulating layer covers the platinum thermal resistance wire layer; the upstream temperature measuring element thermopiles, the central heat source and the downstream temperature measuring element thermopiles are respectively arranged on the first insulating layer, and the multiple groups of upstream temperature measuring element thermopiles and the multiple groups of downstream temperature measuring element thermopiles are respectively distributed side by side along the longitudinal direction of the chip; and front release isolation through holes are formed between thermopiles of the adjacent upstream temperature measurement elements and between thermopiles of the adjacent downstream temperature measurement elements, and the front release isolation through holes extend to the back release cavity. According to the utility model, the sensitive area of the temperature measuring element can be automatically cleaned, and the power consumption can be reduced.

The invention discloses an injection mold for automobile part machining, and relates to the field of automobile part machining, the injection mold comprises injection molding equipment, an ejection mechanism facilitating demolding of a formed part is arranged in a mold base, the ejection mechanism comprises an ejection rod installed on the inner wall of a movable mold cavity in a sliding mode, and an ejection block and a push block are arranged at the end of the ejection rod; a pneumatic ejection assembly is arranged in the movable mold, and a cooling mechanism used for cooling the movable mold is arranged in the movable mold. According to the injection mold for automobile part machining, the ejection mechanism is arranged in the movable mold, uniform ejection force can be applied to a formed part during mold opening, the problem that demolding is difficult due to the fact that the formed part adheres to the inner wall of a mold cavity is effectively solved, and part damage and the defective rate caused by forced demolding are reduced; the cooling mechanism is arranged in the movable mold, the movable mold can be continuously and efficiently cooled, the plastic melt cooling and curing period is shortened, the injection molding production efficiency is improved, and meanwhile it is guaranteed that the internal stress of parts is uniform, and the dimensional precision is stable.

The application relates to the technical field of intelligent manufacturing and computer vision, and discloses a robot welding action generation method and system and electronic equipment. The application realizes adaptive elimination of visual redundancy in the time sequence dimension through global time sequence merging, guarantees complete reservation of key visual detail features in a complex welding scene through diversity perception pruning, realizes collaborative simplification of text and visual features through attention weight screening pruning, and completes dynamic adaptive optimization of bimodal features through semantic perception pruning. The application solves the defect that the prior art cannot adaptively process visual features, removes redundant features layer by layer, reduces the calculation load, greatly improves the real-time performance of welding work, accurately reserves the core visual and text semantic features required for welding action generation, fully guarantees the control accuracy of the welding action, and significantly improves the adaptive perception capability and work reliability of the robot welding system in a complex scene.

The invention provides an Internet of Things clustering optimization method based on a fragmented block chain, and aims to solve the problems of expandability and performance bottleneck of a block chain system in an Internet of Things environment. The K-Shard system optimizes and reconfigures the nodes by combining the genetic algorithm and the cuckoo rule, so that the reconfiguration delay is reduced, and the flexibility and the response speed of the system are improved. In addition, by introducing a reputation system based on a reward and punishment mechanism, the K-Shard can effectively identify and isolate malicious nodes, and the security of the network is ensured. The system adopts a double deep Q learning (double DQN) technology, optimizes the number of fragments and consensus parameters, and further improves the throughput and efficiency of transaction processing. A simulation result shows that the K-Shard can process large-scale Internet of Things equipment and data streams while keeping low delay, and shows performance superior to that in the prior art.

The invention discloses an intelligent unmanned aerial vehicle irrigation nozzle wind direction adjusting device, and particularly relates to the technical field of agricultural irrigation, the intelligent unmanned aerial vehicle irrigation nozzle wind direction adjusting device comprises an inner fixing frame arranged outside a nozzle, an outer fixing frame is arranged outside the inner fixing frame, and one side of the outer surface wall of the outer fixing frame is fixedly connected with a connecting clamping seat. By arranging the swing structure, in the irrigation process, the spray head can periodically swing up and down, the longitudinal range of irrigation can be effectively expanded, a specific angle area right below the spray head to the upper portion can be covered, the irrigation area of the height dimension can be effectively extended, and therefore the irrigation requirements of crops of different heights can be accurately met; the spray head swings up and down to powerfully counteract the adverse effect of wind on the spraying direction, when the wind makes the spray deflect to one side, the up-and-down movement of the spray head can promote the spray to be distributed in the vertical direction, the spray deflection degree is reduced, it is guaranteed that the spray accurately covers a target area, and the stability and accuracy of irrigation operation are greatly improved.

The invention discloses an ion implantation method for an alloy steel oil separation cover of an aviation hydraulic pump. The ion implantation method is characterized by comprising the steps of blank cleaning, blow-drying, charging, vacuumizing, cleaning parameter setting, injection parameter setting, a cleaning stage, an injection stage, program operation ending, discharging and the like. The modified layer is obtained through the ion implantation process on the friction surface of the aviation hydraulic pump oil separation cover, the hardness of the oil separation cover is greatly improved, the wear resistance of the product is relatively improved, and the service life is prolonged; the technological process is stable and reliable, and it is guaranteed that the product flatness and other precision are not affected; the macroscopic morphology of the injection layer is golden yellow, the color is uniform, no piebald exists, the hardness improvement proportion is higher than 10%, and the coating thickness is excellent.

The invention discloses a submarine cable state evaluation method and system, relates to the technical field of electric power, and aims to remove redundant data by adopting a multi-level screening mode so as to ensure that key data can be screened out preferentially and focus points can be adjusted in the current submarine cable state evaluation process, and the evaluation efficiency of the submarine cable state can be improved. The evaluation key point is ensured to always meet the current actual demand, so that the current evaluation precision is improved, the current evaluation reliability is enhanced, the scientificity and accuracy of the evaluation result are effectively ensured, and a powerful guarantee is provided for long-term safe operation of the submarine cable.

The utility model discloses a weightlessness scale food and medicine grade weighing platform structure comprising a closed housing, a weighing sensor is arranged in the closed housing, the two ends of the weighing sensor are deformation transmission ends, the first deformation transmission end is connected with an upper plane, the second deformation transmission end is connected with a support part, and the second deformation transmission end is connected with an upper plane. The support part is provided with supporting legs penetrating out of the lower end face of the closed housing; the supporting legs freely penetrate through the lower end face of the closed housing. The food and medicine grade weighing platform structure of the weightlessness scale is provided with the totally-enclosed closed housing, the weighing sensor is perfectly wrapped in the internal accommodating chamber, so that the weighing platform structure has better dustproof performance, and no four corners are designed, so that materials cannot enter the closed housing, the materials cannot go bad in the weighing platform, and the weighing platform structure is convenient to use. Therefore, pollution to the production environment can be avoided, experiments and verification show that the structure meets the GMP design requirement of the food and pharmacy industry, and the precision of the sensor is not affected.

The invention provides a method for predicting wall rock caving in a roadway tunneling process, and relates to the technical field of roadway caving disaster prediction. According to the method for predicting the wall caving of the wall surrounding rock in the roadway tunneling process, the wall caving possibility and the wall caving depth can be accurately predicted through on-site sampling, indoor testing, theoretical calculation and combination of stress and energy criteria in the tunneling process, all parts of the wall surrounding rock of the roadway wall are predicted, and the prediction range is wide; the prediction method is simple in process implementation and high in prediction analysis efficiency, real-time prediction of roadway side caving can be basically realized, and factors such as geological condition change do not influence prediction precision and efficiency; by means of the prediction method, the rib caving area can be prevented and treated, and the safety and continuity of equipment such as a digging and anchoring all-in-one machine in coal mine roadway digging and supporting operation are effectively guaranteed.

This utility model discloses an oil circuit lubrication mechanism for a flat knitting machine needle punch, relating to the field of oil circuit lubrication technology. It includes a main body with a moving mechanism at its rear end, comprising a rack and a half-gear. The rack is fixedly connected to the front end of the main body, and the half-gear is meshed with the upper end of the rack. When the main body lubricates the parts of the flat knitting machine needle punch, the motor is started, and the rotating disk drives the abutment to rotate. Since the abutment is within a rectangular groove of a long rod, and the long rod is connected to a fixed shaft, the long rod drives the half-gear to oscillate. The oscillation of the half-gear drives the rack to reciprocate, which in turn drives the main body to reciprocate. During this reciprocating motion, the main body can cover the two ends of the needle bed, the edges of the needle movement trajectory, or the corners of the triangular system with oil, preventing dry or semi-dry friction of the parts and thus avoiding a reduction in the service life of the parts, ensuring that the knitting accuracy and product quality are not affected.

The present invention provides a method for predicting the erosion rate of a dilute gas-solid two-phase flow, characterized in that the method comprises: step one, collecting parameters that affect the erosion rate of equipment, devices or pipe fittings, etc.; step two, preparing for a nozzle erosion test; step three, testing the gas flow rate; step four, simulating the gas flow rate CFD; step five, comparing the simulation results with the test results to optimize the turbulence model and the near-wall treatment method; step six, testing the particle velocity; step seven, simulating the particle velocity CFD; step eight, comparing the simulation results with the test results to construct a particle motion model; step nine, testing the erosion rate; step ten, simulating the erosion CFD; step eleven, comparing the simulation results with the test results to correct the erosion rate model; step twelve, forming a method for predicting the erosion rate of a dilute gas-solid two-phase flow. The method can be used for any dilute gas-solid two-phase flow condition, and has the advantages of high accuracy and strong versatility.

The invention discloses a voice noise reduction recognition method and device in a machine room environment, and the method comprises the steps: obtaining noise data in a plurality of types of machine rooms, and carrying out the first preprocessing, and obtaining a noise power spectrum; collecting audio data in the machine room and carrying out second preprocessing to obtain an audio power spectrum; based on the noise power spectrum, performing noise reduction processing on the audio power spectrum to obtain voice data; and based on a pre-constructed speech acceleration recognition model, performing recognition processing on the speech data, and giving text information. According to the invention, the voice command can be quickly and accurately recognized, so that a corresponding response is made.

The utility model discloses a micrometer calibration clamping and fixing device which comprises a fixing seat, an adjusting groove is arranged in the fixing seat, clamping seats are symmetrically arranged in the adjusting groove through a sliding assembly, a positioning mechanism matched with the sliding assembly is arranged on the side wall of the fixing seat, and clamping mechanisms are arranged on the clamping seats through fixing assemblies. And a lifting assembly is arranged at one end of the fixed seat. The micrometer calibration device has the beneficial effects that through the arrangement of the fixing assembly and the clamping mechanism, the micrometer can be conveniently clamped and fixed, the stability in the calibration process is ensured, rigid contact can be avoided, the micrometer is protected, meanwhile, an elastic cushion can be conveniently and regularly disassembled and replaced, and the work efficiency is improved. And therefore, the clamping effect and the protection effect are prevented from being influenced by deformation or damage after use.

The invention discloses an electric power unmanned aerial vehicle inspection side equipment defect identification method and system based on an auto-encoder, and the method comprises the steps: an edge device loads an electric power unmanned aerial vehicle inspection image, and carries out the preprocessing of the image; constructing an electric power unmanned aerial vehicle inspection model which comprises a training stage model and a reasoning stage model, and reasoning the pictures; and carrying out result post-processing, judging whether the data is processed or not after post-processing, if so, ending identification, and otherwise, returning. According to the method, the idea of an auto-encoder is introduced, an original backbone network is regarded as an encoder, and a decoder is added, so that additional auxiliary training information can be added for training of the model, and the backbone part of the model can learn more sufficient and compact information. The added decoder and the network part are relatively independent and can be removed in reasoning, so that the reasoning speed of the model is increased, and the model precision is not influenced.