Aerodynamic instability warning method and system for multi-stage high-speed compressor based on deterministic learning

Abstract

The present invention proposes a multi-stage high-speed compressor aerodynamic instability early warning method and system based on deterministic learning. Sensor measuring points are arranged in a multi-angle manner combining full ring and sector in the multi-stage high-speed compressor to obtain pulsating pressure data in the aerodynamic instability test. The internal nonlinear dynamic system of the aerodynamic instability is learned based on the pulsating pressure data using the deterministic learning modeling method. The learning result is a holographic expression of the fast time-varying instability mode, which can depict the more essential and comprehensive dynamic characteristics of the development process of the complex dynamic phenomenon of aerodynamic instability and has good interpretability. On this basis, an online early warning method for aerodynamic instability is developed. Only the pulsating pressure data of a single measuring point of a multi-stage high-speed compressor can give a warning signal before the instability occurs. Therefore, it has certain application value and is expected to provide a real-time monitoring method with practical significance for the safe and stable operation of high-performance aircraft engines.

Description

Technical Field

[0001] The present invention belongs to the technical field related to aerodynamic instability warning of a multi-stage high-speed compressor, and in particular to a method and system for aerodynamic instability warning of a multi-stage high-speed compressor based on deterministic learning. Background Art

[0002] The statements in this section merely provide background information related to the present invention and do not necessarily constitute prior art.

[0003] Turbofan engines are the most commonly used power units for military and large civil aircraft. The load capacity and aerodynamic stability of their core components, the multi-stage high-speed axial compressor, are crucial to the engine's operating efficiency and safe operation. With the increase in the thrust-to-weight ratio of turbofan engines and the reduction in the number of core engine stages, the increase in compressor load has made aerodynamic instability problems (rotating stall and surge, etc.) more prominent, becoming one of the main bottlenecks restricting engine performance improvement. Therefore, there is an urgent need to effectively warn and control the aerodynamic instability of multi-stage high-speed axial compressors to improve the reliability and stability of turbofan engine operation.

[0004] E.M. Greitzer of MIT conducted a systematic and comprehensive review of the mechanism of aerodynamic instability, pointing out that the initial surge of a low-speed single-stage or multi-stage axial flow compressor is always accompanied by rotating stall, so rotating stall is considered to be a precursor to surge, and timely capture of rotating stall signals is very important for instability warning. In the rotating stall test, two stall precursors were found: modal waves and spike waves. The former is a large-scale, small-amplitude disturbance rotating along the circumference of the compressor, and the propagation speed does not exceed 50% of the rotor speed; the latter is a small-scale disturbance with a faster propagation speed, which appears as a pulse in the signal. Once it appears, it expands rapidly and causes the compressor to enter a stall. However, in the test of multi-stage high-speed compressors, the situation becomes more complicated. The compressor has a spike-type stall at low speed, a modal wave stall at intermediate speed, and a spike-type stall or sudden surge at full speed. This shows that there are a large number of stall / surge modes in multi-stage high-speed compressors.

[0005] In order to solve the problem of identifying and detecting the precursors of stall and surge, people often use frequency domain, time domain or time-frequency analysis methods to directly analyze the characteristics of the initial disturbance of stall, such as harmonic Fourier coefficient method, traveling wave energy method, wavelet analysis method, correlation measurement method, etc. These methods extract partial characteristics such as amplitude, phase, and energy when rotating stall occurs, which makes it difficult to fully characterize the dynamic characteristics of rotating stall. In addition, the initial disturbance of stall is diverse due to different compressors, so it is difficult to obtain statistically significant precursors. In addition, people try to indirectly predict the occurrence of stall based on the changes in the compressor blade channel signal. For example, M. Dhingra et al. of Georgia Institute of Technology used a correlation measurement method based on random process analysis to track this change, and achieved the detection of the precursor of the sudden tip type stall 30ms in advance in the GE high-speed compressor test. However, the irregularity of the blade channel signal is affected by the tip clearance and concentricity. These parameters change continuously during the flight cycle, and the stall warning based on the irregularity of the blade channel signal may be invalid.

[0006] In summary, although the research on aerodynamic instability warning of axial compressors has made certain progress, in actual complex scenarios, especially instability warning of multi-stage high-speed compressors, it still faces many challenges. As Professor IJ Day, a fellow of the British Academy of Engineering and Cambridge University, pointed out in a recent review, after 75 years of research, we still cannot accurately predict the instability characteristics of a new compressor. Especially compared with single-stage low-speed compressors, multi-stage high-speed compressors have a more complex mechanism of aerodynamic instability precursors due to their high pressure ratio and fast acceleration, and they change extremely rapidly. The diversity and complexity of aerodynamic instability modes of multi-stage high-speed compressors make rotating stall and surge warnings face huge challenges in theory and engineering. Summary of the invention

[0007] In order to overcome the shortcomings of the above-mentioned prior art, the present invention provides a multi-stage high-speed compressor aerodynamic instability early warning method and system based on deterministic learning. Sensors are installed at multi-angle measuring points combining the full ring and the sector to collect the pulsating pressure signals of the multi-stage high-speed axial flow compressor. The internal nonlinear dynamic system of the aerodynamic instability is learned according to the pulsating pressure data using a deterministic learning modeling method. The learning results can characterize the more essential and comprehensive dynamic characteristics of the development process of the complex dynamic phenomenon of aerodynamic instability, and have good interpretability. The early warning method developed on this basis can realize online early warning of aerodynamic instability using only single measuring point data.

[0008] In order to achieve the above object, the present invention adopts the following technical solution:

[0009] In a first aspect, the present invention provides a multi-stage high-speed compressor aerodynamic instability early warning method based on deterministic learning, comprising:

[0010] The sensor measuring points are arranged in a multi-angle manner combining full ring and sector on a multi-stage high-speed compressor to obtain pulsating pressure test data of different instability types under different operating conditions;

[0011] Based on the deterministic learning theory, the RBF neural network is used to train the pulsating pressure test data of the normal mode and the instability precursor mode, and the constant RBF neural network approximate model of the nonlinear dynamics inside the compressor is obtained.

[0012] According to various instability types under different operating conditions, a dynamic mode library is constructed using the constant RBF neural network approximate model of normal mode and instability precursor mode;

[0013] The pulsating pressure signal of a single measuring point of the compressor to be tested is used to design a dynamic estimator based on the constructed dynamic pattern library to evaluate the operating status of the compressor to be tested and realize online early warning of aerodynamic instability of a multi-stage high-speed compressor.

[0014] In a second aspect, the present invention provides a multi-stage high-speed compressor aerodynamic instability warning system based on deterministic learning, comprising:

[0015] An instability data acquisition module is configured to: arrange sensor measuring points in a multi-angle manner combining full ring and sector on a multi-stage high-speed compressor to acquire pulsating pressure test data of different instability types under different operating conditions;

[0016] An operation mode training module is configured to: based on deterministic learning theory, use an RBF neural network to learn and train the pulsation pressure test data of the normal mode and the instability precursor mode, and obtain a constant RBF neural network approximate model of the nonlinear dynamics inside the compressor;

[0017] A pattern library construction module is configured to: construct a dynamic pattern library using a constant RBF neural network approximate model of a normal mode and an instability precursor mode according to various instability types under different operating conditions;

[0018] The aerodynamic instability warning module is configured to: utilize the single-measurement-point pulsating pressure signal of the compressor to be tested, design a dynamic estimator based on the constructed dynamic pattern library, evaluate the operating state of the compressor to be tested, and realize online warning of aerodynamic instability of multi-stage high-speed compressors.

[0019] In a third aspect, the present invention provides an electronic device comprising a memory and a processor, and computer instructions stored in the memory and executed on the processor, wherein when the computer instructions are executed by the processor, the method described in the first aspect is performed.

[0020] In a fourth aspect, the present invention provides a computer-readable storage medium for storing computer instructions, wherein when the computer instructions are executed by a processor, the method described in the first aspect is performed.

[0021] In a fifth aspect, the present invention provides a computer program product, comprising a computer program, which, when executed by a processor, implements the method described in the first aspect.

[0022] One or more of the above technical solutions have the following beneficial effects:

[0023] In the present invention, a multi-angle measurement point arrangement scheme combining full-ring and sector is proposed, and high-frequency response sensors are used to perform multi-channel, high-sampling-rate parallel real-time acquisition of blade tip pulsation pressure data at all levels, thereby realizing the capture of the circumferential propagation characteristics of modal waves and spike wave stall precursors, and solving the problem of capturing the dynamic data of aerodynamic instability precursors of multi-stage high-speed axial flow compressors in complex environments.

[0024] In the present invention, by determining the learning theory modeling method, the internal nonlinear dynamic system of aerodynamic instability can be learned according to the pulsating pressure data of a multi-stage high-speed axial flow compressor. The learning result, as a holographic expression of the fast time-varying instability mode, can characterize the more essential and comprehensive dynamic characteristics of the development process of the complex dynamic phenomenon of aerodynamic instability, and has good interpretability.

[0025] Compared with the instability detection technology in existing projects, the method proposed in the present invention only uses the pulsating pressure data of a single measuring point of a multi-stage high-speed compressor to give a warning signal before instability occurs. Therefore, it has certain application value and is expected to provide a real-time monitoring method with practical significance for the safe and stable operation of high-performance aircraft engines.

[0026] Advantages of additional aspects of the present invention will be given in part in the following description, and in part will become obvious from the following description, or will be learned through practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] The accompanying drawings in the specification, which constitute a part of the present invention, are used to provide a further understanding of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute improper limitations on the present invention.

[0028] Figure 1 This is a schematic diagram of the installation of sensors for multi-angle measurement points in the first embodiment of the present invention;

[0029] Figure 2 Schematic diagram of first-order derivative estimation of pulsating pressure signal in Embodiment 1 of the present invention;

[0030] Figure 3 Schematic diagram of spatial trajectories in different modes in Embodiment 1 of the present invention;

[0031] Figure 4 Schematic diagram of the convergence process of the neural network estimation weight in the first embodiment of the present invention;

[0032] Figure 5 is the single-measurement-point flow data φ of the compressor under test in the first embodiment of the present invention r (k) Schematic diagram of decision indicators corresponding to each dynamic estimator;

[0033] Figure 6 This is a flow chart of a multi-stage high-speed compressor aerodynamic instability warning method based on deterministic learning in Example 1 of the present invention. DETAILED DESCRIPTION

[0034] It should be noted that the following detailed descriptions are exemplary and are intended to provide further explanation of the present invention. Unless otherwise specified, all technical and scientific terms used herein have the same meanings as those commonly understood by those skilled in the art to which the present invention belongs.

[0035] It should be noted that the terms used herein are for describing specific embodiments only and are not intended to be limiting of exemplary embodiments according to the present invention.

[0036] In the absence of conflict, the embodiments of the present invention and the features of the embodiments may be combined with each other.

[0037] Embodiment 1

[0038] like Figure 6 As shown, this embodiment discloses a multi-stage high-speed compressor aerodynamic instability warning method based on deterministic learning, including:

[0039] Step 1: Arrange sensor measuring points in a multi-angle manner combining full ring and sector on a multi-stage high-speed compressor to obtain pulsating pressure test data of different instability types under different operating conditions;

[0040] Step 2: Based on the deterministic learning theory, the RBF neural network is used to train the pulsating pressure data of the normal mode and the instability precursor mode to obtain a constant RBF neural network approximate model of the nonlinear dynamics inside the compressor;

[0041] Step 3: According to various instability types under different operating conditions, a dynamic mode library is constructed using the constant RBF neural network approximate model of the normal mode and the instability precursor mode;

[0042] Step 4: Using the single-point pulsating pressure signal of the compressor to be tested, a dynamic estimator is designed based on the constructed dynamic pattern library to evaluate the operating status of the compressor and realize online early warning of aerodynamic instability of multi-stage high-speed compressors.

[0043] The following is a detailed description of this embodiment:

[0044] In step 1 of this embodiment, according to the circumferential disturbance propagation characteristics of modal waves and spike wave stalls, a multi-angle measurement point installation scheme combining full-ring and sectoral is adopted to achieve the acquisition of dynamic pressure data of high-order aerodynamic instability precursors under engineering environment, thereby constructing an aerodynamic instability data set for different instability types and different operating conditions using the surge test data.

[0045] The rotating stall and surge of the axial compressor of a turbofan engine are essentially complex dynamic phenomena produced by an infinite-dimensional distributed parameter system described by partial differential equations. The following circumferentially discretized high-order Moore-Greitzer model can be regarded as a finite-dimensional approximation of the infinite-dimensional partial differential equation system:

[0046]

[0047] Among them, ψ p represents the average compressor pressure increase, φ∈R 2N+1 represents the state vector composed of the flow coefficients of 2N+1 measuring points along the circumferential position of the compressor, and the aerodynamic instability interference micro-group is described by the circumferential n-order harmonic approximation. c , σ and b represent the compressor geometry parameters, throttling parameters and Greitzer stability parameters respectively, and the nonlinear function ψ c (·) is the compressor characteristic function, and E, F, G and H are coefficient matrices.

[0048] By selecting different system matrices and compressor characteristic functions, the model can simulate the aerodynamic instability development process of various compressors and produce simulation results that are consistent with the experimental measurement data within a certain accuracy range.

[0049] In order to observe the propagation of circumferential disturbances of modal waves and small-scale spike wave stall, this embodiment comprehensively considers the high-order modal properties of aerodynamic instability and the limitations of compressor structure and cost, and proposes a multi-angle measurement point arrangement scheme combining full ring and sector.

[0050] Different numbers of high-frequency response sensors are installed at different angles at each stage of the compressor to achieve multi-channel, high-sampling-rate parallel real-time acquisition of pulsating pressure signals at the blade tip. For example, when full-ring uniform distribution is performed at a certain stage, if N = 7th-order harmonics are selected to describe the aerodynamic instability interference micro-group, corresponding to 2N + 1 = 15 basis functions (respectively 1, sinx, cosx, ..., sin7x, cos 7x), at the circumferential position 15 sensors are evenly spaced. When sensors are arranged at different angles in two sectors at a certain level, one sector can be equipped with 5 sensors corresponding to the 9th-order harmonic to describe the stall micro-group, and the interval between adjacent sensors is Another sector can be equipped with 5 sensors corresponding to the 8th order harmonics to describe the stalled micro-cluster, with the interval between adjacent sensors being like Figure 1 As shown, the harmonic order, full ring uniform distribution, sector position and sensor interval can be adjusted according to actual conditions and are not limited to the specific values ​​in this example.

[0051] Based on the above measurement point arrangement scheme, an aerodynamic instability data set for different operating conditions (such as different intake conditions, different speeds, different temperatures, different pressures, different intake rates, etc.) is constructed using the pulsating pressure data during the multi-stage high-speed axial flow compressor surge test.

[0052] In step 2 of this embodiment, the pulsating pressure data in step 1 is preprocessed, and a dynamic identification algorithm based on deterministic learning is designed to achieve local accurate approximation of the internal nonlinear dynamic system of the pneumatic instability.

[0053] First, according to the physical characteristics of compressor surge, this embodiment adopts preprocessing processes such as filtering, downsampling, observer dimension expansion and normalization.

[0054] Specifically, filtering: According to the frequency spectrum characteristics of aerodynamic instability, this embodiment uses a Butterworth filter to filter out high-frequency noise in the compressor test data, and retains the frequency components of aerodynamic instability, especially precursor signals.

[0055] Downsampling: Downsample the filtered data to reduce the data volume while retaining the overall characteristics of the original data, thereby reducing the challenges that dense data brings to real-time warning tasks.

[0056] Data expansion: This embodiment uses the following sampling high gain observer to estimate the pulsating pressure data φ at each measuring point i Derivative information of:

[0057]

[0058] Where ψ(k)∈R n , Indicates pulsating pressure data φ i The estimated results of the derivatives of each order of (k) are as follows: D=diag[1,∈,…,∈ n-1 ], H o =[h 1 ,…,h n ] T , A o =hH o C, I∈R (n-1)×(n-1) is the identity matrix, C = [1,0,…,0], T is the sampling period, ∈ is a small positive constant, and H is selected o Make s n +h 1 s n-1 +…+h n-1 s+h n = 0 has a negative real part. In this embodiment, n = 2, H o =[4,3] T , α=1.5, The first-order derivative estimation result of the pulsating pressure data is as follows: Figure 2 shown.

[0059] Normalization: For the expanded data, a sliding window is used for normalization. This embodiment uses the following formula for normalization:

[0060]

[0061] in, and Respectively represent the sliding window The minimum and maximum values ​​of .

[0062] Then, according to the preprocessing results of the gasping test data, this embodiment uses the deterministic learning theory to perform dynamic modeling of aerodynamic instability.

[0063] Intercept training mode: For the pre-processed wheezing data, intercept data in different time periods as training patterns, including normal mode and precursor mode.

[0064] Construct a neural network model: According to the pulsating pressure data corresponding to the training mode and its derivative estimation results, a spatial trajectory z(k)=[z 1 (k),…,z n (k)]∈R n , and arrange RBF neurons in the trajectory distribution area. Use Gaussian function as the neural network basis:

[0065]

[0066] Among them, η is the receptive field width, ξ l is the neuron center, and L is the number of neurons.

[0067] like Figure 3 As shown, this embodiment uses pulsating pressure data and its first-order derivative to form a trajectory in a two-dimensional plane, and evenly distributes L=289 neuron centers in the area of ​​[-1.6, 1.6]×[-1.6, 1.6], and η=0.2.

[0068] Dynamic modeling of training mode: Based on the above RBF neural network model, an identifier is constructed to learn the internal nonlinear dynamics of the pulsating pressure data:

[0069]

[0070] Where v is the state of the identifier, z n is the n-1th order derivative of the pulsating pressure data, the identifier gain is α = 0.1, is the estimated value of the neural network weight, and the initial weight is Used to study the nonlinear dynamics of normal or precursory modes.

[0071] The neural network weight adjustment law is:

[0072]

[0073] Where T is the sampling period, z n is the n-1th order derivative of the pulsating pressure data. In this embodiment, the learning gain is γ=200, σ=0.005.

[0074] According to the determined learning theory, such as Figure 4 As shown, the RBF subvector S composed of neurons near the trajectory ζ (z) satisfies the continuous excitation condition, corresponding to the neuron weight sub-vector exponentially converges to the ideal true value; for neurons far away from the trajectory, the weight subvector remains at 0, which means that as Adjustment, The nonlinear dynamics can gradually approach the normal or precursory modes in the trajectory neighborhood.

[0075] In step 3 of this embodiment, for various instability types under different working conditions, the identification results in step 2 are used to perform a holographic expression of the dynamic characteristics of aerodynamic instability, thereby constructing a dynamic model library that describes the instability characteristics of a multi-stage high-speed compressor.

[0076] Estimated weights after selecting transient state Compute constant-valued neural network weights:

[0077]

[0078] Among them, [Tk a ,Tk b ] represents the time period after the transient state. The constant RBF neural network obtained It is possible to achieve an approximate expression of the internal dynamics of aerodynamic instability within the trajectory neighborhood, which means that the learning result contains the holographic characteristics of the state and dynamics of the multi-stage high-speed compressor system.

[0079] Furthermore, for aerodynamic instability data under different instability types and different operating scenarios, a dynamic mode library is constructed using the constant neural network learning results of the normal mode and the instability precursor mode:

[0080]

[0081] Among them, j is used to distinguish the instability type and operation scenario, m=0 indicates the normal mode, and m=1 indicates the instability precursor mode.

[0082] In addition, dynamic modeling can be performed on more gasping test data to expand the model library, which is conducive to the comprehensive expression of the aerodynamic instability development characteristics of multi-stage high-speed compressors.

[0083] In step 4 of this embodiment, a dynamic estimator is designed using the pattern library in step 3, and the single-point pulsating pressure signal of the measured compressor is input into the estimator. The operating status of the compressor is quickly identified based on the estimator residual, thereby achieving online real-time early warning of aerodynamic instability.

[0084] Flow data φ at any measuring point of a multi-stage high-speed axial flow compressor r (k), the preprocessing process in step 2 is used for online processing, and the result is recorded as z r (k)∈R n . Use the constant neural network model in the pattern library Φ to construct multiple dynamic estimators:

[0085]

[0086] in, is the state of the estimator, and the preprocessing result of the single measuring point data of the tested compressor is r (k) is fed into all estimator systems in parallel and in real time, Indicates z r For all dynamic estimators, the same gain parameter b is set. In this embodiment, b=0.9.

[0087] Defining the estimator residuals Satisfies the following equation:

[0088]

[0089] Among them, d(k) is the error associated with the observer data expansion, f r (·) is the stream data φ r The true system dynamics of (k) is an unknown nonlinear function.

[0090] According to the above estimator residual system, if it is stored in The dynamic knowledge in is effectively activated, along the spatial trajectory z of the test pattern r (k) It can reflect the dynamic difference between the actual operating mode of the tested compressor and each training mode in the mode library. It is used as a similarity measure of the operating mode. Similar modes mean smaller estimator residuals. Therefore, the estimator residual can be regarded as an external manifestation of the similarity between the actual mode of the compressor and different training modes, which can be used to monitor the actual operating status of the compressor.

[0091] In order to perform the online warning task of aerodynamic instability, this embodiment takes the average L1 norm of the above-mentioned estimator residual as the decision indicator:

[0092]

[0093] Among them, T e =200 is the length of the sliding window. r (k), the decision indicators calculated by each estimator are as follows Figure 5 shown.

[0094] from Figure 5 It can be seen that when the compressor is operating normally, the dynamic difference between the actual operation mode and the precursor mode is large, resulting in a larger warning index corresponding to the precursor mode; as the compressor approaches and eventually enters the unstable operation state, the dynamic difference between the actual operation mode and the precursor mode gradually decreases, resulting in a smaller warning index corresponding to the precursor mode and a larger warning index corresponding to the normal mode, and at t = 258.80s, the warning index of the precursor mode is less than that of the normal mode. Therefore, based on the minimum residual principle, the occurrence of the instability precursor mode can be quickly detected, and the present invention uses it as an alarm signal to achieve online real-time warning of aerodynamic instability of multi-stage high-speed axial flow compressors.

[0095] In summary, when the decision indicators of each estimator are used to monitor the actual operating status of a multi-stage high-speed axial flow compressor, the dynamic knowledge in the determined learning model can be used to evaluate the dynamic characteristics of the compressor system in real time, thereby achieving high-performance early warning of aerodynamic instability of a multi-stage high-speed compressor.

[0096] Embodiment 2

[0097] The purpose of this embodiment is to provide a multi-stage high-speed compressor aerodynamic instability warning system based on deterministic learning, including:

[0098] An instability data acquisition module is configured to: arrange sensor measuring points in a multi-angle manner combining full ring and sector on a multi-stage high-speed compressor to acquire pulsating pressure test data of different instability types under different operating conditions;

[0099] An operation mode training module is configured to: based on deterministic learning theory, use an RBF neural network to learn and train the pulsating pressure data of the normal mode and the instability precursor mode, and obtain a constant RBF neural network approximate model of the nonlinear dynamics inside the compressor;

[0100] A pattern library construction module is configured to: construct a dynamic pattern library using a constant RBF neural network model of a normal mode and an instability precursor mode according to various instability types under different operating conditions;

[0101] The aerodynamic instability warning module is configured to: utilize the single-measurement-point pulsating pressure signal of the compressor to be tested, design a dynamic estimator based on the constructed dynamic pattern library, evaluate the operating state of the compressor to be tested, and realize online warning of aerodynamic instability of multi-stage high-speed compressors.

[0102] In further embodiments, there is also provided:

[0103] An electronic device includes a memory and a processor, and computer instructions stored in the memory and executed on the processor, wherein when the computer instructions are executed by the processor, the method described in Embodiment 1 is performed. For the sake of brevity, no further description is given here.

[0104] It should be understood that in this embodiment, the processor may be a central processing unit CPU, and the processor may also be other general-purpose processors, digital signal processors DSP, application-specific integrated circuits ASIC, off-the-shelf programmable gate arrays FPGA or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc.

[0105] The memory may include a read-only memory and a random access memory, and provide instructions and data to the processor. A portion of the memory may also include a non-volatile random access memory. For example, the memory may also store information about the device type.

[0106] A computer-readable storage medium is used to store computer instructions. When the computer instructions are executed by a processor, the method described in embodiment 1 is completed.

[0107] The method in the first embodiment can be directly embodied as a hardware processor, or a combination of hardware and software modules in the processor. The software module can be located in a mature storage medium in the field such as a random access memory, a flash memory, a read-only memory, a programmable read-only memory, or an electrically erasable programmable memory, a register, etc. The storage medium is located in the memory, and the processor reads the information in the memory and completes the steps of the above method in combination with its hardware. To avoid repetition, it will not be described in detail here.

[0108] A computer program product includes a computer program, and when the computer program is executed by a processor, the method described in the first embodiment is implemented.

[0109] The present invention also provides at least one computer program product tangibly stored on a non-transitory computer-readable storage medium. The computer program product includes computer executable instructions, such as instructions included in a program module, which are executed in a device on a real or virtual processor of the target to perform the process / method as described above. Typically, program modules include routines, programs, libraries, objects, classes, components, data structures, etc. that perform specific tasks or implement specific abstract data types. In various embodiments, the functions of program modules can be combined or divided between program modules as needed. Machine executable instructions for program modules can be executed in local or distributed devices. In distributed devices, program modules can be located in local and remote storage media.

[0110] The computer program code for implementing the method of the present invention can be written in one or more programming languages. These computer program codes can be provided to the processor of a general-purpose computer, a special-purpose computer or other programmable data processing device, so that the program code, when executed by the computer or other programmable data processing device, causes the function / operation specified in the flow chart and / or block diagram to be implemented. The program code can be executed completely on a computer, partially on a computer, as an independent software package, partially on a computer and partially on a remote computer or completely on a remote computer or server.

[0111] In the context of the present invention, computer program codes or related data may be carried by any appropriate carrier to enable a device, apparatus or processor to perform the various processes and operations described above. Examples of carriers include signals, computer readable media, etc. Examples of signals may include electrical, optical, radio, acoustic or other forms of propagation signals, such as carrier waves, infrared signals, etc.

[0112] Those of ordinary skill in the art will appreciate that the units and algorithm steps of each example described in conjunction with this embodiment can be implemented in electronic hardware or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.

[0113] Although the above describes the specific implementation mode of the present invention in conjunction with the accompanying drawings, it is not intended to limit the scope of protection of the present invention. Those skilled in the art should understand that various modifications or variations that can be made by those skilled in the art on the basis of the technical solution of the present invention without creative work are still within the scope of protection of the present invention.

Claims

1. A multi-stage high-speed compressor aerodynamic instability early warning method based on deterministic learning, characterized in that: include: Arrange sensor points in a multi-angle manner combining full ring and sector on a multi-stage high-speed compressor to obtain pulsating pressure test data of different instability types under different operating conditions; Based on the deterministic learning theory, the RBF neural network is used to train the pulsating pressure test data of the normal mode and the instability precursor mode, and the constant RBF neural network approximate model of the nonlinear dynamics inside the compressor is obtained, which is specifically: The pulsating pressure data corresponding to the training pattern and the estimation results of each order of derivatives thereof form a spatial trajectory; In the process of using RBF neural network to identify the internal dynamics of the compressor, the RBF subvector composed of the RBF neural network nodes in the neighborhood of the spatial trajectory meets the continuous excitation condition, and the corresponding neuron weight subvector exponentially converges to the optimal value; the RBF neural network nodes far away from the spatial trajectory are not excited, and the corresponding neuron weight subvector remains zero; According to various instability types under different operating conditions, the dynamic mode library is constructed using the constant RBF neural network approximate model of the normal mode and the instability precursor mode, specifically: Calculate the constant RBF neural network weights based on the estimated weights of the transient post-RBF neural network; For the pulsating pressure test data under different instability types and different operating conditions, the dynamic mode library is constructed using the constant RBF neural network corresponding to the normal mode and the instability precursor mode. The pulsating pressure signal of a single measuring point of the compressor to be tested is used to design a dynamic estimator based on the constructed dynamic pattern library to evaluate the operating status of the compressor to be tested and realize online early warning of aerodynamic instability of a multi-stage high-speed compressor.

2. The multi-stage high-speed compressor aerodynamic instability early warning method based on deterministic learning as claimed in claim 1, characterized in that: By installing different numbers of sensors at each stage of a multi-stage high-speed compressor in a multi-angle manner combining full ring and sector, multi-channel and high sampling rate collection of pulsating pressure signals at the blade tip is performed.

3. The multi-stage high-speed compressor aerodynamic instability early warning method based on deterministic learning as claimed in claim 1, characterized in that: The pulsating pressure signal of a single measuring point of the compressor to be tested is preprocessed, and multiple dynamic estimators are constructed using the constant RBF neural network in the dynamic pattern library. The estimation residuals are calculated according to the dynamic estimator, and the average L1 norm of the estimator residuals is used as the decision indicator to monitor the actual operating status of the compressor to be tested.

4. The method for early warning of aerodynamic instability of a multi-stage high-speed compressor based on deterministic learning as claimed in claim 3, characterized in that: The preprocessing includes filtering, down sampling, observer dimension expansion and normalization processing.

5. The multi-stage high-speed compressor aerodynamic instability early warning method based on deterministic learning as claimed in claim 3, characterized in that: The estimated residual is used to measure the similarity between the actual operating mode of the compressor to be tested and different training modes.

6. A multi-stage high-speed compressor aerodynamic instability warning system based on deterministic learning, characterized in that: include: An instability data acquisition module is configured to: arrange sensor measuring points in a multi-angle manner combining full ring and sector on a multi-stage high-speed compressor to acquire pulsating pressure test data of different instability types under different operating conditions; The operation mode training module is configured as follows: based on the deterministic learning theory, the RBF neural network is used to train the pulsating pressure data of the normal mode and the instability precursor mode, and obtain the constant RBF neural network approximate model of the nonlinear dynamics inside the compressor, specifically: The pulsating pressure data corresponding to the training pattern and the estimation results of each order of derivatives thereof form a spatial trajectory; In the process of using RBF neural network to identify the internal dynamics of the compressor, the RBF subvector composed of the RBF neural network nodes in the neighborhood of the spatial trajectory meets the continuous excitation condition, and the corresponding neuron weight subvector exponentially converges to the optimal value; the RBF neural network nodes far away from the spatial trajectory are not excited, and the corresponding neuron weight subvector remains zero; The mode library construction module is configured to construct a dynamic mode library according to various instability types under different operating conditions using a constant RBF neural network approximate model of a normal mode and an instability precursor mode, specifically: Calculate the constant RBF neural network weights based on the estimated weights of the transient post-RBF neural network; For the pulsating pressure test data under different instability types and different operating conditions, the dynamic mode library is constructed using the constant RBF neural network corresponding to the normal mode and the instability precursor mode. The aerodynamic instability warning module is configured to: utilize the single-measurement-point pulsating pressure signal of the compressor to be tested, design a dynamic estimator based on the constructed dynamic pattern library, evaluate the operating state of the compressor to be tested, and realize online warning of aerodynamic instability of multi-stage high-speed compressors.

7. An electronic device, characterized in that: The method comprises a memory and a processor, and computer instructions stored in the memory and executed on the processor, wherein when the computer instructions are executed by the processor, the method according to any one of claims 1 to 5 is completed.

8. A computer-readable storage medium, characterized in that: Used to store computer instructions, which, when executed by a processor, complete the method described in any one of claims 1 to 5.

Images