Combustion instability boundary prediction system based on complex environment condition analysis

Abstract

The invention relates to the technical field of combustion monitoring, and discloses a combustion instability boundary prediction system based on complex environment condition analysis, and the system comprises a multi-source sensing module, a data fusion module, an anti-interference processing module, an instability feature library, an intelligent decision engine and a digital twinborn body. A dynamic environment interference compensation mechanism is constructed, spectrum correction parameters are configured according to different environment working conditions during real-time prediction of a combustion instability boundary, the problem of spectrum distortion caused by environment disturbance on combustion dynamic parameters is eliminated in real time, spectrum interference deviation of a combustion measurement position can be automatically sensed and corrected, and real-time prediction of the combustion instability boundary is realized. The accuracy of combustion instability sensitive feature extraction is guaranteed, the instability early warning error is reduced, a flame pose real-time tracking system is deployed, fuel control parameter self-adaptive adjustment is triggered immediately when it is detected that the pose is abnormal, autonomous correction of the combustion instability feature recognition position is achieved, and it is guaranteed that the feature capture space is accurately positioned.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of combustion monitoring, in particular to a combustion instability boundary prediction system based on complex environment condition analysis. BACKGROUND

[0002] Combustion instability is characterized by the periodic oscillation coupling of combustion chamber pressure and heat release rate, which is listed as a key scientific problem in the design of aero-engine combustion chambers. Micro- premixed combustion experiments show that equivalence ratio changes can induce the flame shape to change from a thin and long type to an inverted "V" type, accompanied by an increase in pressure pulsation amplitude from 260 Pa to 800 Pa. The essence of combustion instability is the positive feedback coupling of acoustic oscillation and heat release process. When the phase difference between the periodic heat release and the acoustic standing wave in the combustion chamber satisfies the Rayleigh criterion, the system exhibits enhanced oscillation. The low-frequency high-amplitude pressure pulsation generated by the unstable combustion process not only affects the combustion performance, but can also cause damage to the combustion chamber structure in severe cases.

[0003] Currently, due to the operation of the combustion system in complex and variable environmental conditions, when performing real-time prediction of the combustion instability boundary, the environmental disturbance compensation mechanism deployed for frequency spectrum correction of combustion dynamic parameters cannot eliminate the frequency spectrum distortion caused by environmental disturbances in real time. When the frequency spectrum distortion caused by environmental disturbances is too large, it will cause a large error in the extraction of combustion instability sensitive features, and cannot guarantee the accuracy of instability early warning. At the same time, when monitoring the combustion state, it cannot identify the dynamic displacement of the flame flow field in real time, which will cause the spatial position of the combustion instability feature recognition to deviate, and the feature position cannot be automatically corrected when it deviates. When predicting the combustion instability boundary, due to the multi-frequency domain mode coupling characteristics in the combustion chamber, when dynamically predicting the instability boundary, multi-modal hierarchical coupling analysis cannot be achieved, resulting in mode interference misjudgment in the combustion instability prediction, further affecting the accuracy and reliability of the instability boundary prediction.

[0004] Therefore, the present application proposes a combustion instability boundary prediction system based on complex environment condition analysis to solve the above problems. SUMMARY

[0005] In view of the deficiencies of the prior art, the present application provides a combustion instability boundary prediction system based on complex environment condition analysis to solve the problems raised in the background art.

[0006] To achieve the above purpose, the present application provides the following technical solution: a combustion instability boundary prediction system based on complex environment condition analysis, the system comprising: A multi-source perception module: containing a distributed pressure sensor array, a multi-waveband spectrum acquisition unit, and an environmental parameter monitoring unit, which captures combustion chamber physical field data in real time; Data fusion module: performs environmental-burning parameter spatiotemporal alignment, multi-scale feature extraction and coupling matrix generation operations; Anti-interference processing module: built-in environmental disturbance transfer function library, real-time spectral correction to eliminate measurement bias; Instability feature library: stores feature templates and boundary evolution rules of historical instability cases; Intelligent decision engine: uses a reinforcement learning mechanism to drive a bionic optimization algorithm to output an environment-adaptive prediction model; Digital twin: includes a combustion chamber fluid dynamics simulation unit and a real-time early warning unit to generate visual instability boundary and control instructions.

[0007] Preferably, it comprises: The multi-source perception module uses a micro-electro-mechanical system ultrasonic sensor and a high-temperature optical fiber probe, and the sampling frequency is not less than 50 kHz; The anti-interference processing module integrates an adaptive Kalman filter; The intelligent decision engine is deployed with a rule explainability unit to quantify feature contribution through SHAP value analysis; The digital twin forms a closed loop with the actual combustion control system, and the control instruction delay is less than 10 ms.

[0008] The system comprises the following steps: S1, real-time acquisition of combustion chamber environmental parameter data and combustion dynamic parameter data, the environmental parameters including intake pressure pulsation spectrum, fuel component fluctuation rate and environmental turbulence intensity, the combustion dynamic parameters including combustion chamber pressure oscillation waveform, flame radiation intensity and intermediate free radical concentration; S2, multi-source heterogeneous data coupling processing of environmental parameter data and combustion dynamic parameter data to generate environmental-burning coupling feature matrix; S3, based on the environmental disturbance factor dynamic compensation model, frequency spectrum distortion correction processing of the combustion dynamic parameter data to generate anti-interference combustion dynamic feature data; S4, construction of combustion state feature space, extraction of intrinsic mode components of the anti-interference combustion dynamic feature data through nonlinear dimension reduction to generate combustion instability sensitive feature vectors; S5, instability mode matching of the environmental-burning coupling feature matrix with a preset instability mode library, wherein the instability mode library stores standard instability boundary feature templates under different environmental conditions; S6, using a bionic optimization algorithm to search for the optimal instability judgment rule to generate an environment-adaptive instability boundary prediction model; S7, based on the instability boundary prediction model, instability risk quantitative evaluation of the real-time combustion flow field to generate dynamic instability boundary atlas and early warning level data; S8, driving the combustion control parameter adaptive adjustment through the digital twin engine, and outputting a combustion stability optimization control instruction.

[0009] Preferably, the step S1 specifically comprises: S11, arranging a 16-channel high-frequency pressure sensor array circumferentially at the head of the combustion chamber and the flame stabilizer to collect dynamic pressure fluctuation signals at a sampling frequency of 200 kHz, and generating a pressure oscillation distribution tensor with a space-time resolution of 5 mm x 5 mm x 0.1 ms , wherein x, y are sensor coordinates, and t is time; S12, synchronously collecting OH*, CH* free radical radiation intensity by using an ultraviolet-visible fiber spectrometer, acquiring a flame spontaneous light emission image sequence by using a high-speed CMOS camera, and calculating a flame front pulsation velocity field by using an optical flow algorithm , wherein is an image plane coordinate, t is time, OH* is a hydroxyl radical radiation, and CH* is a methyl radical radiation; S13, measuring the inlet end turbulent integral scale by using a three-dimensional ultrasonic wind speed instrument and the turbulent intensity , monitoring the C, H, and O element molar ratio fluctuation amount in the fuel component in real time by using a gas chromatograph-mass spectrometer , and constructing an environmental disturbance feature vector , wherein is the turbulent integral scale, is the turbulent intensity, is the methane molar ratio fluctuation amount, is the oxygen molar ratio fluctuation amount.

[0010] Preferably, the step S2 is realized by the following manner: S21, establishing a phase space reconstruction model of the environmental parameters and the combustion dynamic parameters: wherein ⊕ represents a tensor direct product operation, is a time delay compensation amount, is a flame stabilizer center coordinate, is a coupling feature tensor, is a flame front average position, is a central pressure oscillation, is a flame front pulsation, is a time delay environmental parameter, and FFT(OH∗( t )) is an OH free radical radiation spectrum; is a time variable; S22, strong correlation feature pairs are screened by using the maximum information coefficient MIC, and when the MIC value is greater than 0.75, the feature combination is retained, and an environment-burning coupling feature matrix C with a dimension of 8x12 is constructed, and the elements satisfy: wherein is a combustion dynamic characteristic component, is an environmental disturbance characteristic component, is a coupling matrix element, is an environmental disturbance characteristic component, is a combustion dynamic characteristic component, is an integral time length, is a standard deviation of the combustion dynamic characteristic, is a standard deviation of the environmental disturbance characteristic, is a differential of the integral variable .

[0011] Preferably, the step S3 comprises a spectrum correction mechanism: S31, an environmental disturbance transfer function model is constructed: wherein is an environmental disturbance transfer function, is an environmental parameter vector, is a combustion pressure power spectrum, is an environmental disturbance power spectrum, is a disturbance propagation time delay, is an environmental dependent time delay, =0.15 is a dynamic compensation coefficient, is an environmental change gradient, is a phase delay term, is an imaginary unit; S32, a real combustion signal is reconstructed by using an adaptive inverse filtering algorithm: wherein is a corrected combustion signal, is an inverse Fourier transform, is a regularization factor related to the environmental gradient, is an anti-interference feature matrix, is a transfer function conjugate.

[0012] Preferably, the step S4 adopts a multi-scale feature fusion strategy: S41, a pressure oscillation intrinsic mode is extracted by using a variational mode decomposition: 6 intrinsic mode function IMF components are obtained; Six intrinsic mode function (IMF) components were obtained; in For the center frequency, For time domain envelope, For time derivative operators, For the Dirac function, To construct an analytical signal, To reconstruct the constraints, This is a combustion pressure signal. For Hilbert transform kernel, It is a complex exponential modulation; S42. Calculate the nonlinear coupling exponent matrix: in Here, m is the nonlinear coupling matrix, and m is the number of combustion oscillation modes. Let be the coupling exponent of mode pair i,j, and be the time-domain signal of the i-th oscillation mode. For modality standard deviation For the i-th IMF component, a 12-dimensional combustion instability sensitive feature vector is finally generated. .

[0013] Preferably, step S5 includes a dynamic pattern matching mechanism: S51, Instability Mode Library Store 9 typical operating condition templates, each template containing: Standard instability boundary function ; Environmental operating conditions constraints ; Feature vector reference value ; S52. The matching degree is calculated using the quantized dynamic time warping algorithm: in For quantized similarity, This is a real-time combustion instability sensitive feature vector. For the k-type template in the instability mode library, Let be the projection similarity of the i-th feature component. For the eigenvector norm, The total energy representing the template instability characteristics; S53, when And the current environment When the k-th type of instability boundary prediction mode is activated.

[0014] Preferably, the step S6 is implemented by improving the ant colony algorithm: S61, initialization parameters: number of ants , pheromone matrix , evaporation coefficient , heuristic factor weight , ; S62, construct rule selection probability model: wherein is the feature space distance, is the probability of ant k selecting rule (i→j), is the pheromone concentration of edge (i,j), is the pheromone weight index, is the heuristic factor, is the heuristic factor weight index, is the selectable neighborhood of ant k at node i, , is the real-time mean, standard deviation of feature distance; S63, pheromone update adopts elite strategy: wherein is the pheromone increment, is the number of elite ants, is the enhancement coefficient, is the instability risk predicted by ant k, is the actual combustion state label, is the pheromone evaporation coefficient, t is the time scale.

[0015] Preferably, the step S7 comprises three-dimensional risk field construction: S71, real-time calculation of instability risk index: wherein is the three-dimensional instability risk index, A is the pressure oscillation amplitude, E is the environmental parameter vector, is the pressure oscillation frequency, is the optimal decision rule set, is the extended feature vector, is the decision boundary bias, is the frequency domain feature vector; S72, generate frequency-amplitude-risk three-dimensional boundary surface: wherein is the instability risk region set, Design static pressure for combustion chamber, Design environmental parameter vector, Design three-dimensional real space; S73, according to Value is divided into three early warning levels: Yellow early warning: ; Orange early warning: ; Red early warning: .

[0016] Compared with the prior art, the present application provides a combustion instability boundary prediction system based on complex environmental condition analysis, which has the following beneficial effects: 1. In the present application, by constructing a dynamic environmental disturbance compensation mechanism, when the combustion instability boundary is predicted in real time, the frequency spectrum correction parameters are configured according to different environmental conditions, the frequency spectrum distortion problem caused by the environmental disturbance of the combustion dynamic parameters is eliminated in real time, the spectral interference deviation of the combustion measurement position can be automatically sensed and corrected, the accuracy of the combustion instability sensitive feature extraction is ensured, and the instability warning error is reduced.

[0017] 2. In the present application, by deploying a flame pose real-time tracking system, when the combustion state is monitored, the flame space offset angle is calculated based on high-speed vision analysis, whether the combustion feature detection position deviates from the flow field is automatically identified, and the fuel control parameter is automatically adjusted when the pose is abnormal, the self-correction of the combustion instability feature recognition position is realized, and the positioning accuracy of the feature capture space is ensured.

[0018] 3. In the present application, by developing a frequency domain modal decoupling analysis unit, when the combustion instability boundary is predicted, the multi-frequency domain oscillation mode in the combustion chamber is automatically decoupled based on the variational modal decomposition algorithm, the coupling strength of each mode is independently quantified, and the hierarchical instability risk atlas is generated, the interaction mechanism of the rotating instability and the acoustic standing wave complex mode can be accurately identified, the misjudgment problem caused by modal interference can be avoided, and the prediction reliability is improved. BRIEF DESCRIPTION OF DRAWINGS

[0019] Fig. 1 It is the overall system architecture diagram of the present application; Fig. 2 It is the system step flow chart of the present application; Fig. 3 It is the subsequent flow chart of the feature matrix processing of the present application. DETAILED DESCRIPTION

[0020] With reference to the drawings of the embodiments of the present application, the technical solutions in the embodiments of the present application will be clearly and completely described. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments of the present application. Based on the embodiments of the present application, all other embodiments obtained by those of ordinary skill in the art without creative work fall within the scope of the present application.

[0021] Please refer to Figs. 1-3 The combustion instability boundary prediction system based on complex environment condition analysis comprises: The multi-source sensing module comprises a distributed pressure sensor array, a multi-band spectrum acquisition unit and an environment parameter monitoring unit, and can capture physical field data of the combustion chamber in real time. The data fusion module performs environment-combustion parameter space-time alignment, multi-scale feature extraction and coupling matrix generation operations. The anti-interference processing module has an environment disturbance transfer function library built-in, and can eliminate measurement deviation through real-time spectrum correction. The instability feature library stores feature templates and boundary evolution rules of historical instability cases. The intelligent decision engine adopts a reinforcement learning mechanism to drive a bionic optimization algorithm, and outputs an environment-adaptive prediction model. The digital twin comprises a combustion chamber fluid dynamics simulation unit and a real-time early warning unit, and generates a visual instability boundary and a control instruction.

[0022] It comprises: The multi-source sensing module adopts a micro-electro-mechanical system ultrasonic sensor and a high-temperature optical fiber probe, and the sampling frequency is not less than 50 kHz. The anti-interference processing module integrates an adaptive Kalman filter. The intelligent decision engine is provided with a rule explainability unit, and the feature contribution degree is quantified through SHAP value analysis. The digital twin and the actual combustion control system form a closed loop, and the control instruction delay is less than 10 ms.

[0023] The system comprises the following steps: S1, real-time acquisition of combustion chamber environment parameter data and combustion dynamic parameter data, the environment parameters comprising inlet pressure fluctuation spectrum, fuel component fluctuation rate and environmental turbulence intensity, the combustion dynamic parameters comprising combustion chamber pressure oscillation waveform, flame radiation intensity and intermediate free radical concentration; S2, multi-source heterogeneous data coupling processing of the environment parameter data and the combustion dynamic parameter data to generate an environment-combustion coupling feature matrix; S3, frequency spectrum distortion correction processing of the combustion dynamic parameter data based on an environment disturbance factor dynamic compensation model to generate anti-interference combustion dynamic feature data; S4, construct a combustion state feature space, extract intrinsic mode components of anti-interference combustion dynamic characteristic data through nonlinear dimension reduction, and generate a combustion instability sensitive feature vector; S5, match the environment-combustion coupling feature matrix with a preset instability mode library, wherein the instability mode library stores standard instability boundary feature templates under different environmental conditions; S6, search for an optimal instability judgment rule using a bionic optimization algorithm, and generate an environment-adaptive instability boundary prediction model; S7, based on the instability boundary prediction model, quantitatively evaluate the instability risk of the real-time combustion flow field, and generate a dynamic instability boundary atlas and early warning level data; S8, drive the adaptive adjustment of the combustion control parameters through the digital twin engine, and output the combustion stability optimization control instructions.

[0024] Step S1 specifically includes: S11, arranging 16-channel high-frequency pressure sensor arrays circumferentially at the head of the combustion chamber and the flame stabilizer, collecting dynamic pressure fluctuation signals at a sampling frequency of 200 kHz, and generating a pressure oscillation distribution tensor with a time and space resolution of 5 mm x 5 mm x 0.1 ms , wherein x, y are sensor coordinates, and t is time; S12, synchronously collecting OH*, CH* free radical radiation intensity using an ultraviolet-visible fiber spectrometer, acquiring flame self-luminous image sequences using a high-speed CMOS camera, and calculating flame front pulsation velocity field using an optical flow algorithm , wherein is the image plane coordinate, t is the time, OH* is the hydroxyl radical radiation, and CH* is the methyl radical radiation; S13, measuring the inlet end turbulent integral scale and the turbulent intensity using a three-dimensional ultrasonic wind speed instrument, and monitoring the C, H, and O element molar ratio fluctuations in the fuel components in real time using a gas chromatograph-mass spectrometer , and constructing an environmental disturbance feature vector , wherein is the turbulent integral scale, is the turbulent intensity, is the methane molar ratio fluctuation, is the oxygen molar ratio fluctuation.

[0025] Step S2 is realized by the following method: S21, establish a phase space reconstruction model of environmental parameters and combustion dynamic parameters: wherein ⊕ represents tensor direct product operation, is the time delay compensation, is the flame stabilizer center coordinate, is the coupling feature tensor, is the flame front average position, is the center pressure oscillation, is the flame front fluctuation, is the time-delayed ambient parameter, FFT(OH*(t - Td) ) is the OH radical radiation spectrum, t is the time variable; S22, the strong correlation feature pairs are screened by the maximum information coefficient MIC, and when the MIC value is greater than 0.75, the feature combination is retained, and an environment-burning coupling feature matrix C with a dimension of 8x12 is constructed, and the elements thereof satisfy: wherein is the burning dynamic feature component, is the environmental disturbance feature component, is the coupling matrix element, is the environmental disturbance feature component, is the burning dynamic feature component, is the integral time length, is the standard deviation of the burning dynamic feature, is the standard deviation of the environmental disturbance feature, is the differential of the integral variable .

[0026] Step S3 includes a spectrum correction mechanism: S31, an environmental disturbance transfer function model is constructed: wherein is the environmental disturbance transfer function, is the environmental parameter vector, is the burning pressure power spectrum, is the environmental disturbance power spectrum, is the disturbance propagation time delay, is the environmental dependent time delay, =0.15 is the dynamic compensation coefficient, is the environmental change gradient, is the phase delay term, is the imaginary unit; S32, the real burning signal is reconstructed by using an adaptive inverse filtering algorithm: wherein is the corrected burning signal, is the inverse Fourier transform, is the regularization factor related to the environmental gradient,​ is an anti-interference feature matrix, is a transfer function conjugate.

[0027] Step S4 adopts a multi-scale feature fusion strategy: S41, extract the pressure oscillation intrinsic modal by variational modal decomposition: Obtain 6 intrinsic modal function IMF components; wherein is the center frequency, is the time domain envelope, is the time derivative operator, is the Dirac function, is the analytical signal construction, is the reconstruction constraint condition, is the combustion pressure signal, is the Hilbert transform kernel, is the complex exponential modulation; S42, calculate the nonlinear coupling index matrix: wherein is the nonlinear coupling matrix, m is the number of combustion oscillation modes, is the coupling index of mode pair i, j, is the time domain signal of the i th order oscillation mode, is the standard deviation of the mode , is the i th IMF component, and finally generate a 12-dimensional combustion instability sensitive feature vector .

[0028] Step S5 includes a dynamic pattern matching mechanism: S51, the instability mode library stores 9 typical working condition templates, and each template includes: standard instability boundary function ; environmental condition constraint ; feature vector reference value ; S52, calculate the matching degree by using the quantized dynamic time warping algorithm: wherein is the quantized similarity, is the real-time combustion instability sensitive feature vector, is the k th template in the instability mode library, is the projection similarity of the i th dimensional feature component, is the feature vector norm, total energy of the template instability characteristics; S53、when and the current environment , the kth instability boundary prediction mode is activated.

[0029] Step S6 is implemented by improving the ant colony algorithm: S61, initialize parameters: the number of ants , pheromone matrix , evaporation coefficient , heuristic factor weight , ; S62, construct a rule selection probability model: wherein is the feature space distance, is the probability of ant k selecting rule (i→j), is the pheromone concentration of edge (i,j), is the pheromone weight index, is the heuristic factor, is the heuristic factor weight index, is the selectable neighborhood of ant k at node i, , is the real-time mean and standard deviation of the feature distance; S63, pheromone update adopts the elite strategy: wherein is the pheromone increment, is the number of elite ants, is the enhancement coefficient, is the instability risk predicted by ant k, is the actual combustion state label, is the pheromone evaporation coefficient, and t is the time scale.

[0030] Step S7 includes three-dimensional risk field construction: S71, real-time calculation of instability risk index: wherein is the three-dimensional instability risk index, A is the pressure oscillation amplitude, E is the environmental parameter vector, is the pressure oscillation frequency, is the optimal decision rule set, is the extended feature vector, is the decision boundary bias, is the frequency domain feature vector; S72, generating a frequency-amplitude-risk three-dimensional boundary surface: wherein is a set of instability risk regions, is a combustor design static pressure, is an environmental parameter vector, is a three-dimensional real space; S73, according to the value is divided into three levels of early warning: yellow warning:; orange warning:; red warning:.

[0031] Example one: active inhibition of extreme turbulence working condition instability of aero-engine: A certain type of large thrust aero-engine encountered sudden weather disturbance during transoceanic cruise phase, the monitoring data showed that the inlet end turbulence intensity jumped to 32% instantaneously, and the turbulence integral scale reached 0.35 meters. The system immediately started the environmental disturbance compensation protocol: the three-dimensional ultrasonic anemometer in the multi-source perception module captured the turbulence spectrum characteristics at a sampling rate of 50kHz, the data fusion module called the high-speed inlet turbulence special transfer function to perform real-time spectrum reconstruction on the combustion pressure data collected by the 16-channel pressure sensor array. The anti-interference processing module eliminates the frequency distortion components through an adaptive inverse filtering algorithm, restores the real combustion pressure waveform. The instability feature library synchronously activates the high turbulence working condition template, the intelligent decision engine extracts 6 intrinsic mode components based on variational mode decomposition, among which the energy percentage of 300Hz vortex mode reaches 42%, and the nonlinear coupling index η34 breaks through the critical value of 0.28. The digital twin generates control instructions within 8 milliseconds, reducing the fuel flow of No. 3 combustion zone by 5%, and adjusting the guide vane angle by 2.3 degrees. According to the monitoring of China Aero-engine Commercial Test Bench, the combustion chamber pressure oscillation amplitude is stabilized from the peak value of 92kPa to below 35kPa, successfully avoiding deep thermoacoustic oscillation. This case lasted for 72 hours, and the system false alarm rate was only 1.3% for a total of 17 turbulence mutation events.

[0032] Example two: precise control of heavy gas turbine variable component cold start: A 9F heavy-duty gas turbine encountered severe fluctuations in fuel composition during cold start in extremely cold environments. Gas chromatography-mass spectrometry monitoring showed that the methane molar ratio fluctuation reached 0.12, and the oxygen concentration deviation exceeded the reference value by 7%. The system automatically triggered multiple safeguard mechanisms: first, a high-temperature-resistant optical fiber probe was arranged at the head of the flame tube to collect the flame radiation spectrum, combined with a 1024×768 pixel high-speed CMOS camera to capture the flame front movement trajectory, and the optical flow algorithm accurately calculated the spatial displacement angle of the flame as 4.5 degrees; the digital twin immediately drove the fuel distribution valve to inject 8% premixed gas into nozzle No. 4 to correct the spatial pose. At the same time, the instability feature library accurately matched the fuel-rich start template T4 with a quantum similarity of 0.92, and the intelligent decision engine used frequency domain decoupling technology to implement targeted suppression of the 800 Hz high-frequency oscillation mode, and strengthened the energy distribution in the 50-150 Hz low-frequency stable zone. The entire start-up process was recorded by the national energy efficiency monitoring platform, and the data showed that the combustion efficiency fluctuation amplitude was only 0.8%, the time from ignition to full load operation was shortened to 19 minutes, the combustion chamber transition state temperature gradient decreased by 40%, and the risk of turbine blade thermal stress damage was reduced.

[0033] It should be noted that the relational terms herein such as first and second and the like are used solely to distinguish one entity or action from another entity or action without necessarily requiring or implying any such actual relationship or order between such entities or actions. Moreover, the terms "comprises", "comprising", or any other variations thereof, are intended to cover a non-exclusive inclusion such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but can include other elements not expressly listed or inherent to such process, method, article, or apparatus. Without further limitation, an element preceded by "comprises... " does not, without more limitations, foreclose the existence of additional identical elements in the process, method, article, or apparatus that includes the recited element.

[0034] Although embodiments of the present application have been shown and described, it is to be understood that various modifications, substitutions, replacements and changes can be made to these embodiments without departing from the principles and spirit of the present application, and the scope of the present application is defined by the appended claims and their equivalents.

Claims

1. A combustion instability boundary prediction system based on complex environmental condition analysis, characterized in that: The system includes: Multi-source sensing module: includes a distributed pressure sensor array, a multi-band spectral acquisition unit and an environmental parameter monitoring unit, which captures combustion chamber physical field data in real time; Data fusion module: performs spatiotemporal alignment of combustion parameters, multi-scale feature extraction, and coupling matrix generation operations. Anti-interference processing module: Built-in environmental disturbance transfer function library, which eliminates measurement deviations through real-time spectrum correction; Instability Feature Library: Stores feature templates and boundary evolution patterns of historical instability cases; Intelligent decision engine: It uses a reinforcement learning mechanism to drive a biomimetic optimization algorithm and outputs an environment-adaptive predictive model; Digital twin: Includes a combustion chamber fluid dynamics simulation unit and a real-time early warning unit, generating visualized instability boundaries and control commands.

2. The combustion instability boundary prediction system based on complex environmental condition analysis according to claim 1, characterized in that: include: The multi-source sensing module uses a microelectromechanical system ultrasonic sensor and a high-temperature fiber optic probe, with a sampling frequency of not less than 50kHz. The anti-interference processing module integrates an adaptive Kalman filter; The intelligent decision engine is equipped with a rule interpretability unit, which quantifies the contribution of features through SHAP value analysis. The digital twin forms a closed loop with the actual combustion control system, and the control command delay is less than 10ms.

3. The combustion instability boundary prediction system based on complex environmental condition analysis according to claim 1, characterized in that: The system includes the following steps: S1. Real-time acquisition of combustion chamber environmental parameter data and combustion dynamic parameter data, wherein the environmental parameters include intake pressure pulsation spectrum, fuel composition fluctuation rate and environmental turbulence intensity, and the combustion dynamic parameters include combustion chamber pressure oscillation waveform, flame radiation intensity and intermediate free radical concentration; S2. Perform multi-source heterogeneous data coupling processing on environmental parameter data and combustion dynamic parameter data to generate an environmental-combustion coupling feature matrix; S3. Based on the dynamic compensation model of environmental interference factors, the dynamic combustion parameter data is subjected to spectral distortion correction processing to generate anti-interference dynamic combustion characteristic data. S4. Construct a combustion state feature space, and extract the intrinsic mode components of the anti-interference combustion dynamic feature data through nonlinear dimensionality reduction to generate a combustion instability sensitive feature vector. S5. Match the environment-combustion coupling feature matrix with a preset instability mode library, wherein the instability mode library stores standard instability boundary feature templates under different environmental conditions. S6. Use a biomimetic optimization algorithm to search for the optimal instability judgment rule and generate an environment-adaptive instability boundary prediction model; S7. Based on the instability boundary prediction model, quantitatively assess the instability risk of the real-time combustion flow field and generate dynamic instability boundary maps and early warning level data. S8: Drives combustion control parameters to adaptively adjust via a digital twin engine, and outputs combustion stability optimization control commands.

4. The combustion instability boundary prediction system based on complex environmental condition analysis according to claim 3, characterized in that: Step S1 specifically includes: S11. A 16-channel high-frequency pressure sensor array is arranged around the combustion chamber head and the flame stabilizer periphery to acquire dynamic pressure fluctuation signals at a sampling frequency of 200kHz, generating a pressure oscillation distribution tensor with a spatiotemporal resolution of 5mm×5mm×0.1ms. , where x, y are the sensor coordinates, and t is time; S12. Simultaneously acquire the radiation intensity of OH* and CH* free radicals using an ultraviolet-visible fiber optic spectrometer, obtain flame spontaneous emission image sequences using a high-speed CMOS camera, and calculate the flame front pulsating velocity field using an optical flow algorithm. ,in denoted as the image plane coordinates, t as time, OH* as hydroxyl radical radiation, and CH* as methyl radical radiation; S13. Measure the integral scale of turbulence at the inlet using a three-dimensional ultrasonic anemometer. and turbulence intensity The molar ratio fluctuations of C, H, and O elements in fuel components were monitored in real time using gas chromatography-mass spectrometry. Construct environmental disturbance feature vectors ,in Let be the integral scale of turbulence, and be the turbulence intensity. This represents the fluctuation in the methane molar ratio. This represents the fluctuation in the oxygen molar ratio.

5. The combustion instability boundary prediction system based on complex environmental condition analysis according to claim 3, characterized in that: Step S2 is achieved in the following manner: S21. Establish a phase space reconstruction model of environmental parameters and combustion dynamic parameters: Where ⊕ represents the tensor direct product operation, This is the time delay compensation amount. The coordinates of the center of the flame stabilizer, For coupled feature tensors, For the average position of the flame forward, Oscillations around the central pressure. For the Flame Vanguard Pulse, For time-delay environmental parameters, FFT(OH∗( t The radiation spectrum of OH free radicals is shown in the image. It is a time variable; S22. Use the maximum information coefficient (MIC) to filter strongly correlated feature pairs. When the MIC value is greater than 0.75, retain the feature combination and construct an 8×12 environment-combustion coupling feature matrix C, whose elements satisfy: in For combustion dynamic characteristic components, and for environmental disturbance characteristic components, For elements of the coupling matrix, These are characteristic components of environmental disturbance. For combustion dynamic characteristic components, The integration time is the length of time. The standard deviation of combustion dynamic characteristics, The standard deviation of environmental disturbance characteristics. For integration variables The differential.

6. The combustion instability boundary prediction system based on complex environmental condition analysis according to claim 3, characterized in that: Step S3 includes a spectrum correction mechanism: S31. Constructing an environmental disturbance transfer function model: in Let be the environmental disturbance transfer function. For environmental parameter vectors, The combustion pressure power spectrum, This represents the power spectrum of environmental disturbances. To perturb the propagation delay, Due to environment-dependent latency, =0.15 is the dynamic compensation coefficient. For the gradient of environmental change, For phase delay term, The imaginary unit; S32. Reconstruct the real combustion signal using an adaptive inverse filtering algorithm: in To correct the combustion signal, For inverse Fourier transform, This is a regularization factor related to environmental gradients. For the anti-interference feature matrix, For the transfer function conjugate.

7. The combustion instability boundary prediction system based on complex environmental condition analysis according to claim 3, characterized in that: Step S4 employs a multi-scale feature fusion strategy: S41. Extracting the intrinsic modes of pressure oscillations through variational mode decomposition: Six intrinsic mode function (IMF) components were obtained; in For the center frequency, For time domain envelope, For time derivative operators, For the Dirac function, To construct an analytical signal, To reconstruct the constraints, This is a combustion pressure signal. For Hilbert transform kernel, It is a complex exponential modulation; S42. Calculate the nonlinear coupling exponent matrix: in Here, m is the nonlinear coupling matrix, and m is the number of combustion oscillation modes. Let be the coupling exponent of mode pair i,j, and be the time-domain signal of the i-th oscillation mode. For modality standard deviation For the i-th IMF component, a 12-dimensional combustion instability sensitive feature vector is finally generated. .

8. The combustion instability boundary prediction system based on complex environmental condition analysis according to claim 3, characterized in that: Step S5 includes a dynamic pattern matching mechanism: S51, Instability Mode Library Store 9 typical operating condition templates, each template containing: Standard instability boundary function ; Environmental operating conditions constraints ; Feature vector reference value ; S52. The matching degree is calculated using the quantized dynamic time warping algorithm: in For quantized similarity, This is a real-time combustion instability sensitive feature vector. For the k-type template in the instability mode library, Let be the projected similarity of the i-th feature component. For the eigenvector norm, The total energy representing the template instability characteristics; S53, when And the current environment When the k-th type of instability boundary prediction mode is activated.

9. The combustion instability boundary prediction system based on complex environmental condition analysis according to claim 3, characterized in that: Step S6 is implemented by improving the ant colony algorithm: S61. Initialization parameters: Number of ants pheromone matrix Volatility coefficient heuristic factor weights , ; S62. Construct a rule selection probability model: in For the feature space distance, The probability of ant k choosing rule (i→j). Let (i,j) be the pheromone concentration at edge (i,j). The pheromone weight index As a heuristic factor, As a heuristic factor weight index, Let be the optional neighborhood of ant k at node i. , The real-time mean and standard deviation of the feature distance; S63. Pheromones are updated using an elite strategy: in For pheromone increment, For the number of elite ants, To enhance the coefficient, The instability risk predicted by Ant Group K. Label for actual combustion status. t represents the pheromone evaporation coefficient, and t represents the time scale.

10. The combustion instability boundary prediction system based on complex environmental condition analysis according to claim 3, characterized in that: Step S7 includes the construction of a three-dimensional risk field: S71. Real-time calculation of instability risk index: Where A is the three-dimensional instability risk index, A is the pressure oscillation amplitude, and E is the environmental parameter vector. The frequency of pressure oscillation. This is the optimal set of decision rules. To expand the feature vector, For decision boundary bias, It is a frequency domain eigenvector; S72. Generating a frequency-amplitude-risk three-dimensional boundary surface: in This is a collection of areas at risk of instability. Design static pressure for combustion chamber, For environmental parameter vectors, It is a three-dimensional real number space; S73, according to Values ​​are divided into three levels of early warning: Yellow alert: ; Orange alert: ; Red Alert: .

Images