139 results about "Nonlinear coupling" patented technology

A cross-component superconducting gravity gradiometer sensitive to off-diagonal components of the gradient tensor includes, for each gradient axis, a pair of closely matched angular accelerometers coupled by superconducting circuitry, including sensing circuits designed to minimize the sensitivity of the instrument to angular acceleration of the platform at which the angular accelerometers are mounted; and a mode-splitting circuitry designed to reduce a nonlinear coupling of angular acceleration to the output of the gravity gradiometer and to attain the operability of the instrument in a broader range in the frequency domain.

A cross-component superconducting gravity gradiometer sensitive to off-diagonal components of the gradient tensor includes, for each gradient axis, a pair of closely matched angular accelerometers coupled by superconducting circuitry, including sensing circuits designed to minimize the sensitivity of the instrument to angular acceleration of the platform at which the angular accelerometers are mounted; and a mode-splitting circuitry designed to reduce a nonlinear coupling of angular acceleration to the output of the gravity gradiometer and to attain the operability of the instrument in a broader range in the frequency domain.

The present invention relates to an apparatus and methods for achieving phase-matching between various waves and / or modes by operation of a waveguide. Phase-matching between interacting waves is achieved by total internal reflection and transverse Bragg reflection waveguides. Using second harmonic generation in GaAs / AlGaAs as an example, properties are investigated and quantified such as nonlinear coupling efficiency, bandwidth, tunability, and limitations due to dispersion. The technique is advantageous when compared to alternate technologies, where it is particularly attractive as a material independent means to obtain ultra-low-loss nonlinear optical elements for monolithic integration with coherent light source and other active devices.

The invention provides a nonlinear coupling control method for a bridge crane based on generalized movement of a load by aiming at an under-actuated crane system. The nonlinear coupling control method has good trolley positioning and load hunting elimination performances. Compared with an existing crane control method, the method is applicable to not only stabilization control but also trajectory tracking control. The method includes: designing a signal similar to horizontal displacement of the load, and constructing a novel error signal on the basis, so that the coupling relation between a trolley and the load is enhanced. According to the novel nonlinear coupling control method, the crane system is converted into an interconnected system consisting of a load hunting sub-system and an error signal sub-system. The control performance of a closed-loop system is proved by means of inputting the state stability theory and the LaSalle's invariant principle. Experimental results show that good control effects can be obtained by the control method, and the method has good practical application values.

The invention relaters to the weakly light signal detection field, especially to a kind of high-effect infrared single photon detection method including the infrared single photon signal source, the pumping source, the nonlinear crystal, the single photon signal channel and pumping light channel. It is characterized in that the pumping source is composed by the optical fiber laser; the infrared single photon signal inputted by the optical fiber transmission and the pumping light inputted by the optical fiber laser are coupled by the optical fiber polarized controller separately, which are inputted into the wavelength division multiplexer. The output pumping and signal light are outputted by one optical fiber collimator, which can save the complex system accesspries and adjusting process. It has the compact structure, good portable. The laser power is outputted stably, also it has good mode and easy for nonlinear coupling, so it can reach the good room mode matching of two beams in nonlinear crystal. The whole detector can be made by all optical fiber and has low energy cost. It is also convenience to connect other optical fiber devices and reaches the alter use of many systems.

The invention discloses a method for controlling the stability of a vehicle based on vertical load distribution of a tire. The control method comprises the following steps of: 10, giving steering input out by a driver, estimating motion state information and ideal state information of the vehicle by using a finished vehicle model and a finished vehicle reference model, and judging whether the vehicle is positioned in a stable state or not by the comparison of vehicle information; 20, inputting tracking errors of the vehicle into a proportion integration differentiation (PID) controller, and acquiring an active force feedback of a suspension frame by a corresponding PID control algorithm; and 30, after feedback control is received by the vehicle, redistributing a vertical load of the tire of the vehicle to generate calibration yawing moment so as to restrain the sideslip of the vehicle and improve the stability of the vehicle. In the method, the vertical load of the tire is distributed by the active force of the suspension frame; the calibration yawing moment is generated due to a nonlinear coupling relationship between the vertical load of the tire and cornering force; therefore, yawing motion of the vehicle is controlled, the sideslip of the vehicle is restrained, and the lateral stability of the vehicle is improved.

The invention provides a relevant parameter fault classification method based on big data fusion clustering analysis. By use of the fault classification method, fault data is picked up according to an interpretation rule from mass data which operates in equipment, supervised machine independent clustering is carried out, an automatic classification result of relevant parameter faults is formed, the problem that existing equipment fault diagnosis excessively depends on an expert knowledge base but ignores an incidence relation among deep nonlinear coupling parameters of each subsystem is solved, and the problem that mass effective data can not be favorably mined and utilized in practical equipment type operation is solved. Meanwhile, since the implementation of the fault classification method does not need to depend on the accurate physical modeling of object equipment, a difficulty that a traditional complex system is difficult in modeling is avoided, the intelligent classification of faults and relevant parameter analysis on the basis of mass data mining can be realized, and the method has a fault classification capability with controllable accuracy.

The invention discloses a coordinated turning control method for a tailless air vehicle on the basis of fusion of multiple controllers and belongs to the technical field of automatic control over air vehicles. According to the coordinated turning control method for the tailless air vehicle on the basis of the fusion of the multiple controllers, the controllers are divided into inner ring controllers and outer ring controllers, wherein robust self-adaptive nonlinear inverse controllers are used as main controllers by the inner ring controllers, H<8> times optimum controllers are used as compensators by the inner ring controllers, and nonlinear inverse controllers are adopted by the outer ring controllers. According to the kinematics characteristics of the tailless air vehicle in the turning process, multiple control methods are fused, and therefore the problems, such as nonlinear coupling, heading static instability and uncertain disturbance, existing in the turning process of the tailless air vehicle are solved. The tailless air vehicle can complete coordinated turning stably, accurately and rapidly by adopting the control mode that the multiple controllers are coordinated and combined.

The invention discloses a totally-blind digital speech authentication method, which comprises the following steps: (1) performing framing treatment on a speech signal to be judged; (2) dividing each frame signal into K sections, and performing mean value processing on various sections of values; (3) performing FFT transformation on various sections of data of various frames, and solving double spectral values of the various frames of data; (4) making a double spectral contour diagram through the double spectral estimate values of the various frames of data, which are obtained by smoothing; (5) judging whether nonlinear coupling exists through the double spectral contour diagram of the frame speech signals; (6) judging whether the falsified frame speech signal which is determined for the first time needs the secondary determination of more precise falsified position; and (7) performing more fine framing on the falsified frame speech signal requiring secondary determination, repeating the steps from (2) to (6), and determining more accurate falsified position of the frame speech signal. The method can effectively judge whether the digital speech signal is falsified and determine the falsified position of the digital speech signal under the condition of the deficiency of an authentication message of digital watermarking or digital signature, and does not need to embed any data into the speech signal in advance in the authentication process.

The embodiment of the invention provides a method and a device for constructing a physical model of a thermodynamic system. The method comprises the following steps: thermal resistance, thermomotive force and / or an energy source are introduced to construct an equivalent energy flow model of each independent component of the thermodynamic system; the equivalent energy flow model of each independentcomponent is connected through the corresponding isothermal point, and the whole equivalent energy flow model of the thermodynamic system is constructed; Using Kirchhoff's law, the control equationsof energy flow model are established; the dynamic drag equilibrium equations of the working fluid flow process in the thermodynamic system are established; constraint equations between pressure and temperature of working fluid are established; the physical model of the thermodynamic system is constructed by combining the above equations. The method and the device for constructing the physical model of the thermodynamic system provided by the embodiment of the invention effectively combine the visual angle of the energy flow and the working fluid flow analysis, not only reducing the number of the control equations of the model, but also effectively solving the nonlinear coupling relationship among the system variables, and providing convenience for the subsequent model solution.

The embodiment of the invention discloses an underactuated mechanical arm layering sliding mode control method based on fuzzy optimization. The method comprises the steps that a plane two-freedom-degree active-passive (AP) underactuated mechanical arm dynamical model is set up, a nonlinear coupling dynamical model is simplified into an affine nonlinear system, and two joint angles are adopted as control targets; then, layering sliding mode controllers are designed, a subsystem is formed by the angle and the angular speed of each joint, the equivalent input of the two subsystems is solved, a sliding mode main switching surface is constructed through a Lyapunov feedback function, and the control rate is obtained; and finally, a fuzzy rule is designed to dynamically optimizing the switching robust item in the sliding mode control rate, and the buffeting problem existing in a previous control system is solved. Compared with other plane two-freedom-degree AP mechanical arm control methods,by means of the technical scheme, the steady-state time of the control process can be reduced, and the control precision can be improved.

The invention discloses a longitudinal / lateral / vertical personified cooperative control method for an off-road electric drive unmanned vehicle, and relates to the vehicle control. The method comprises the steps: taking a driver as a research object, and analyzing the steering / acceleration and deceleration operation behavior features of a driver under the off-road condition; constructing a personified control system based on multi-performance target optimization according to the parameter uncertainty, nonlinear coupling and time lag of an unmanned vehicle in a non-structure road environment; building a longitudinal / lateral / vertical coupling dynamic personified control system for the off-road electric drive unmanned vehicle, and achieving the dynamic coupling control of the off-road electric drive unmanned vehicle through the transmission, conversion and evolution of the energy and information of dynamic coordination of the coupling dynamic control system, thereby effectively improving the autonomous running comprehensive energy of the electric drive unmanned vehicle, and remarkably improving the performances of the control system.

The invention discloses a hydro-thermal generator set combination optimization scheduling method based on mixed integer programming. The method comprises a step of obtaining the basic data of a single hydropower station, wherein the basic data comprises a reservoir water amount, a limit value of the reservoir water amount and a limit value of power generation flow, a step of carrying out piecewise linearization on a nonlinear constraint condition in the constraint conditions of a single hydropower station optimization scheduling model, a step of solving the single hydropower station optimization scheduling model which allows a target function F to be minimal according to a constraint condition which is subjected to piecewise linearization, a linear constraint condition of the single hydropower station optimization scheduling model and the basic data of the single hydropower station, and a step of determining a hydro-thermal generator set combination optimization scheduling model according to the single hydropower station optimization scheduling model which allows the target function F to be minimal. According to the method, the direct solution of a non-convex nonlinear problem of the nonlinear coupling of variables in time and space is avoided, and the operation efficiency of a power system is improved.

The invention discloses a method for simulating aircraft rock movement with an RANS / LES (Reynolds average numerical simulation) / LES (large eddy simulation) mixing technique, and aims to solve the problem about self-excitation rock movement under the condition that wide-range separated flow of an aircraft cannot be accurately simulated in the prior art. The method mainly comprises steps as follows: a length scale of a turbulence model is appropriately corrected on the basis of a conventional unsteady RANS method, and DES (detached eddy simulation) methods are constructed; tightly coupled solution of a flow field governing equation and a rigid body motion equation is realized in the dual-time-step pseudo time iteration process with adoption of a prediction-correction algorithm, and the changing course of aerodynamic parameters and motion parameters of the aircraft with time is obtained. On the premise that certain computational efficiency is guaranteed, self-excitation rock motion of the aircraft under the complicated flowing condition is more accurately simulated, and an effective means is provided for nonlinear coupling property evaluation of aerodynamic motion / movement of the aircraft.

The invention discloses a redundant drive vehicle dynamics control distribution method, and aims at solving the problem on the control distribution of constrained redundant drive vehicle dynamics. According to the redundant drive vehicle dynamics control distribution method disclosed by the invention, an expected generalized force / torque of a vehicle is determined based on a sliding mode variable structure control, an optimal distribution relationship between the expected generalized force / torque decided by an upper controller and a wheel slip ratio as well as a sideslip angle is established based on an improved fixed-point square control distribution method, priority levels of control tasks are designed based on driving and operating conditions of a vehicle, and a wheel slip ratio and side slip angle joint tracking controller is designed based on an integral sliding mode control, to coordinate and control a steering / driving / braking system of the automobile and ensure each wheel to move according to an expected slip ratio and slip angle. The invention mainly combines a control distribution method and a redundant drive vehicle dynamics system, a nonlinear coupling constraint to tire side longitudinal side force is taken into consideration, stability domain of a vehicle system is expanded, and system safety and reliability are improved.

The invention discloses a method for detecting an early fault of a bearing through a through secondary phase coupling and improved bi-spectrum algorithm, and the method comprises the following steps:obtaining a resampling signal sequence; calculating the resampling signal sequence through an improved bi-spectrum model of secondary phase coupling detection; and carrying out the detection analysisof the early faint defect feature component of the bearing. The method starts from the coupling effect between a fault behavior and a dynamic behavior of a system, combines with the advanced signal processing theory and method for the exploration of a detection method for the early fault features of a rolling bearing, and aims at recognizing a bearing fault at an early stage through a vibration signal, the nonlinear coupling effect with consideration to the fault impact, and improved higher-order statistical analysis. The method has the stable and reliable detection capability for the detection of the early faint fault of the bearing, enables a bearing fault to be timely found, and avoids a bearing accident. The method is wide in application range, and can be used for the detection and tracking of early faults of bearings in various types of transmission systems.

The invention discloses a finite element prediction method, under the action of force-electricity-heat multi-field coupling, of electrode-containing ferroelectric single crystal based on phase-field method analysis. The method comprises the following steps of establishing a new phase-field model with the multi-field coupling based on a Landau theory and an expansion interface model; obtaining the model establishment and the grid division of the electrode-containing ferroelectric single crystal by using ANSYS finite element software, and extracting node information and unit information; importing the obtained node information, the obtained unit information and considered different boundary condition information, and realizing a solution to the force-electricity-heat nonlinear coupling problem of the electrode-containing ferroelectric single crystal by using an FEAP fast solver; importing field variables which are obtained through FEAP finite element solution into a postprocessing program which is written by Fortran, and obtaining visual field variable results by using TECPLOT software. According to the method provided by the invention, not only is the process simple, but also the adaptability and the stability are good; a commercial finite element program can be developed to flexibly adapt to problem changes.

The invention belongs to the technical field of spintronics, and particularly relates to a spintronics device based on a metamaterial. The spintronics device disclosed by the invention is a composite metamaterial system formed by combining a magnetic metal material and a non-magnetic metal material; the upper layer of the spintronics device is of a metal structure consisting of two similar split-ring resonator structures back to back, metal structural units are periodically attached onto an insulating substrate, and the dimensions of the upper layer are in the sub-wavelength magnitude; and the lower layer of the spintronics device consists of periodical metal strips and strips with gradually changing widths, and a magnetic film is arranged at the narrowest place of each metal strip. The structural resonance of the non-magnetic metal material can concentrate electromagnetic waves inside the structure at maximum efficiency, namely a perfect absorption mode is realized; and moreover, the electromagnetic waves are absorbed by using a spinning rectification effect generated by electron spin motion and nonlinear coupling of charge current in a ferromagnetic material, so that a direct current voltage signal is generated. Therefore, the magnetic metal material is placed at the position with strongest electromagnetic field and current in the system, the electromagnetic waves generated by structural resonance are absorbed, and available direct current voltage is generated.

The invention discloses a non-control bullet firing angle determining method. The method is implemented through a mastering computer (1), a data bus (5) and detecting equipment (6); a standard ballistic trajectory firing table manufacturing module (2) manufactures a standard ballistic trajectory firing table, a firing table querying module (3) figures out the firing angle corresponding to the target position coordinate on the standard ballistic trajectory condition, and a firing angle accurate correcting module (4) corrects the firing angle output by the firing table querying module. According to the method, the firing angle on the standard ballistic trajectory condition can be acquired in a manner of firing table querying, the result is figured out according to the ballistic trajectory on the non-standard ballistic trajectory condition, the firing angle of the standard ballistic trajectory condition is corrected, the table querying interpolation error of the firing tablet querying manner is overcome, and the ballistic trajectory deviations caused by nonlinear coupling of non-standard ballistic trajectory parameters are corrected accurately.

The invention relates to a vibration control method based on internal resonance, which is oriented to the longitudinal torsional vibration of an engine, and belongs to the technical field of vibration control of a vehicle engine. The vibration control method comprises the following steps of: establishing a six-degree-of-freedom vibration equation of the engine; constructing a control model of a vibration absorber; establishing a seven-degree-of-freedom vibration equation of the engine with the vibration absorber; decoupling the seven-degree-of-freedom vibration equation, and adjusting the frequency of the vibration absorber to be half of the frequency of a controlled mode; and extracting a controlled mode and vibration absorber mode vibration equation, and solving the controlled mode and vibration absorber mode vibration equation by a multi-scale method. According to the vibration control method disclosed by the invention, through the construction of a nonlinear coupling term, the internal resonance is formed, the vibration energy of the engine is transferred to the vibration absorber, and the vibration energy of the engine is dissipated by the damping of the vibration absorber. The vibration control method disclosed by the invention is obvious in vibration reduction effect, simple in structure and low in energy consumption.

The present invention relates to an apparatus and methods for achieving phase-matching between various waves and / or modes by operation of a waveguide. Phase-matching between interacting waves is achieved by total internal reflection and transverse Bragg reflection waveguides. Using second harmonic generation in GaAs / AlGaAs as an example, properties are investigated and quantified such as nonlinear coupling efficiency, bandwidth, tunability, and limitations due to dispersion. The technique is advantageous when compared to alternate technologies, where it is particularly attractive as a material independent means to obtain ultra-low-loss nonlinear optical elements for monolithic integration with coherent light source and other active devices.

The invention discloses a feedforward control stochastic resonance-based bearing early-stage fault diagnosis method. The method includes the following steps that: (1) a sensor collects original bearing signals; (2) the original bearing signals are processed, so that square wave-like signals can be obtained and are adopted as feedforward control signals; (3) the original bearing signals and the feedforward control signals are jointly subjected to scale change so as to be converted into small-frequency signals with a feedforward control function; (4) two bistable systems are adopted to construct a coupled bistable system through a nonlinear coupling mode; (5) the small-frequency signals obtained in the step 3 are adopted as input signals, and a genetic algorithm is utilized to obtain the optimal parameters of the coupled bistable system; and (6) the small-frequency signals obtained in the step 3 are inputted into the optimal coupled bistable system, and diagnosis on a bearing early-stage fault can be realized through scale restoration. With the feedforward control stochastic resonance-based bearing early-stage fault diagnosis method of the invention adopted, the optimal system parameters of the coupled bistable system are searched through the genetic algorithm, and the feedforward control of stochastic resonance is combined with an improved potential function, so that an enhanced stochastic resonance effect can be generated.

The invention provides a vehicle four-wheel steer-by-wire system, and a control method thereof. The vehicle four-wheel steer-by-wire system is characterized by comprising a sensor module, an electronic control module and an execution module; the sensor module comprises a steering sensor, a deviation rate sensor, an acceleration sensor and a wire displacement sensor; the electronic control module comprises a state monitoring unit, a driving state optimization unit and a state control unit; the execution module comprises a steering mechanism and a driving mechanism; and in addition, according tothe control method provided by the invention, the modeling process of a complex vehicle system and massive calculation of nonlinear coupling variables are omitted, and the response time is saved. Thevehicle four-wheel steer-by-wire system provided by the invention applies ''decomposition-synthesis'' strategies; four tire systems are ''decomposed'' into a plurality of sub-models, and then the sub-models are weighted and combined together, so the whole system runs stably, and is higher in robustness. The control effect is better.

The invention discloses an intelligent filtering and driving device which comprises a filtering reducer assembly, a servo motor and an acceleration / speed / temperature sensor. The intelligent filtering and driving device integrates the filtering reducer assembly, the servo motor and the sensor, thus effectively simplifying the structure, reducing the volume and mass of the device, and improving the system precision. Through a controllable interface, a filtering gear can generate elastic deformation so as to effectively prevent nonlinear coupling vibration generated by influence of extreme working conditions and special environment, avoid the problems of reliability such as non-smooth work, even blockage and the like, guarantee a driving piece to work within the designed range of driving precision and bearing capacity, control the fluctuation generated by the motor due to the change from a high-speed small torque to a low-speed large torque, and reduce vibration noise and fatigue wear. The sensor and the driving part are integrated, so that the running state of the device can be detected in real time, failure signals can be extracted, and the failure can be estimated and predicted so as to guarantee the reliability of the device.

The invention relates to a gearbox fault diagnosis method based on multi-element modal decomposition-transfer entropy, and belongs to the technical field of gear fault analysis. According to the technical scheme, a multi-element modal decomposition noise reduction algorithm is adopted to extract system main characteristics firstly, then transfer entropy is used for describing system complexity, and is used for gear fault diagnosis. According to the gearbox fault diagnosis method, noise-assisted multi-element empirical modal decomposition-transfer entropy is used to analyze gearbox fault vibration information transmission characteristics, nonlinear coupling and information transmission characteristics between output shaft end and input shaft end signal frequency bands under a gearbox faultcondition are quantitatively described so that a transmission path of gearbox fault vibration signals can be explored, a gearbox state evaluation index based on the multi-element empirical modal decomposition-transfer entropy is established, and a new effective means is provided for fault diagnosis, performance degradation state identification and trend prediction of a rotary machine.

The invention provides a multiple-TORA (translational oscillator with rotational actuator) system non-linear coupling adaptive control method and belongs to the technical field of automatic control of underactuated mechanical systems. The method is characterized in that, with unknown parameters and uncertainty factors existing in a system being taken into consideration, a coupling term is introduced to construct a novel energy storage function, and the nonlinear coupling adaptive control method is designed, thereby enhancing transient performance of the system; the method realizes stabilization control of a multiple-TORA (translational oscillator with rotational actuator) system, and can enable the multiple-TORA system to return to a balanced position quickly and accurately and keep stable when given an initial position or influenced; and meanwhile, an update law can carry out online compensation on the influence of the unknown parameters and the uncertainty factors on the system. Experiment results show that the method has good control performance.

The invention discloses a speed reducer, which comprises a power input gear shaft, a first-level planet gear transmission component, an eccentric shaft of which the axis is deflected to the axis of the power input gear shaft in parallel, a second-level planet gear transmission component, a power output gear shaft and a speed reducer shell. The speed reducer employs the planet gear transmission mechanism, so the structure is compact; the floating planet gear transmission mechanism can adaptively compensate manufacturing and assembling errors and abnormal abrasion, and has small transmission error and high precision. In the transmission process, the differential transmission way of the rare-tooth planet gear of the speed reducer increases the number of meshed teeth in a transmission process, ensures strong bearing capability, stable transmission, low transmission noise and high transmission efficiency, and solves the nonlinear coupling problem of the speed reducer under an extreme working condition. In the transmission process, cambered drum teeth of the speed reducer can filter fluctuations caused by processing and installation errors and the bad working condition, so the abrasion is lowered, the transmission is more stable, the noise is lower, and the service life is longer.

The invention discloses a gait feature fusion method based on nonlinear coupling metric learning, and belongs to the technical field of pattern recognition. The technical problem that a gait characterization method in the prior art is low in recognition performance is solved. The gait feature fusion method disclosed by the invention comprises the following steps of: 1, acquiring a binary contour sequence of a person from a gait video stream by adopting a codebook detection method, and standardizing and centralizing each frame of image; 2, detecting a non-front gait period according to the periodicity of the separation degree of the two legs of the person in walking, and extracting gait energy diagram characteristics and active energy diagram characteristics in one period; Step 3, performing nonlinear coupling metric learning on the gait energy diagram features and the active energy diagram features, and projecting the features to a nuclear coupling space to obtain two new features; Andstep 4, carrying out weighted fusion on the two groups of obtained new feature vectors to obtain new gait features of the nuclear coupling space. The gait feature fusion method is effective for non-frontal periodic gait sequences and high in recognition rate.

The invention relates to a coupling analysis method of multi-channel electroencephalogram based on multi-scale multi-variate transfer entropy and belongs to the technical field of non-linear dynamic causal systems and cerebral sensory motion network researches. The method comprises the following steps: 1, collecting a multi-channel electroencephalogram signal by adopting 32-channel Neuracle equipment; 2, respectively performing processing of removing baseline drift, electromyogram interference, eye movement interference and 50Hz power line interference on the collected electroencephalogram signal by adopting matlab software; 3, performing 20 different resolution decompositions on the multi-channel electroencephalogram signal by adopting a coarse-grained analysis method; and 4, analyzing the electroencephalogram signal under different resolutions among different time frequencies by adopting a multi-variate transfer entropy method, and quantitatively characterizing nonlinear coupling andinformation transfer characteristics among different cerebral sections. The coupling analysis method has the effects of describing the nonlinear characteristics among cerebral motion sensory cortex and deeply exploring the coupling intensity and information transfer among different regions of a brain.

The invention discloses an optical communication MIMO detection method and system. The detection method comprises the following steps: acquiring an input signal d(t) and a channel matrix H of a transmitting end of an optical communication system, and a port output Rx(t) of a receiving end; constructing an M-layer full-connection deep neural network which is expressed as a mapping function f(x0;theta):RD0->RDM, and converting an input vector x0 belonging to RD0 into an output vector xM belonging to RDM through M times of iteration; training the full-connection deep neural network by adopting different CN data; in the training process, a first-order optimization algorithm is adopted for gradient calculation based on a cost function, a parameter theta is dynamically updated, an expected valueof J(theta) is minimized, an input signal d(t) belonging to BN is trained in step S3 to obtain the fully connected deep neural network output d'(t) belonging to BN, which completes the detection of the MIMO signal. The present invention adopts the first-order optimization algorithm based on the gradient calculation of the cost function to realize the training of the deep neural network, and the nonlinear coupling problem of multimode optical communication is solved by using the strong nonlinear learning ability of the deep neural network.