An engine icing simulation method for flight simulators

Abstract

The present application belongs to the field of flight simulator simulation technology, and discloses an engine icing simulation method for a flight simulator, which solves the problem that the authenticity and training efficiency cannot be achieved simultaneously in the existing engine icing simulation of a flight simulator. The original evolution sequence of the icing thickness of each part and the environmental working condition input parameters are obtained in real time, and an equivalent icing degree index is generated through a weighted comprehensive model. The equivalent icing degree index is mapped into a time scaling factor through a hybrid mapping model, which drives the virtual time accumulation to deduce a virtual time trajectory. The optimal accelerated and smoothed engine performance parameter instruction is solved by a polynomial spline interpolation algorithm in combination with multiple closed-loop physical conformal constraints. A deicing event trigger signal is obtained, and the optimal accelerated and smoothed engine performance parameter instruction is used for complementary state superposition and fusion to generate a global comprehensive engine environmental response parameter, which is then sent to a bottom interface to control the flight simulator simulation. The present application achieves a dynamic balance between authenticity and training efficiency.

Description

Technical Field

[0001] This invention belongs to the field of flight simulator simulation technology, and specifically relates to an engine icing simulation method for flight simulators. Background Technology

[0002] Flight simulators are core equipment for civil aviation pilot training and aircraft type modification, and engine icing emergency handling training is a compulsory subject. Current flight simulator engine icing simulation mainly falls into two technical categories: The first category is phenomenal simulation: it simulates the changes in engine parameters under icing conditions through a pre-set empirical model or lookup table method, without relying on real flight test data. A typical example is CN102522026A "Flight Icing Simulator". The drawback of this type of method is that it is divorced from actual flight icing data, and the parameter change trend deviates from the actual engine icing process, making it difficult to pass the certification of CCAR-60-R1 "Civil Aircraft Flight Simulator Qualification Requirements". The second type is physical timescale simulation driven by real data: based on real flight test and certification data, the icing process is advanced according to the physical real timescale. A typical example is WO2014043844, which uses an orthogonal basis order reduction method to process the icing data. The drawback of this type of method is that the entire icing process is strictly advanced according to the physical timescale. A single icing special situation training takes tens of minutes or even longer, which consumes a lot of training time.

[0003] In summary, the core contradiction of existing technologies is that realism and training efficiency cannot be achieved simultaneously. Currently, there is no technical solution that can achieve controllable acceleration of the icing process through structured time scaling methods while preserving the characteristics of real icing data. Summary of the Invention

[0004] To address the aforementioned problems in existing technologies, such as the slow and inefficient engine icing simulation process driven by real icing data, this invention proposes an engine icing simulation method for flight simulators. This method includes the following steps: The original evolution sequence of icing thickness in different parts of the engine and the current environmental operating conditions input parameters are acquired in real time. The original evolution sequence of icing thickness in different parts of the engine is used to generate the current equivalent icing degree index, which characterizes the icing state of multiple parts of the engine, through a weighted comprehensive model. The current environmental condition input parameters and the current equivalent icing degree index are input into a composite state-space model to deduce the virtual time trajectory of the engine in the acceleration time domain, the changing trend of the time scaling factor, and the current instantaneous curvature and temporal relative proportion of each simulation core parameter. Based on the deviation difference between the current instantaneous curvature and the time-series relative proportion of each simulation core parameter from the historical real engine response characteristics, the correction boundary weight matrix in the multi-closed-loop physical conformal constraint is adaptively and dynamically updated; the time-progression composite state-space model after updating the correction boundary weight matrix is ​​subjected to conformal optimization calculation through polynomial spline interpolation algorithm, and the optimal acceleration smoothing engine performance parameter command is solved and output. After acquiring the anti-icing event trigger signal and processing it with a delay buffer, it is superimposed and fused with the optimal acceleration smoothing engine performance parameter command that has been processed by anti-aliasing low-pass filtering to generate a comprehensive engine environmental response parameter across the entire domain. Based on the changing trends of the virtual time trajectory and time scaling factor, the global integrated engine environmental response parameters are transformed into a dynamic evolution sequence by combining the state management mechanism and sent to the underlying interface to control the flight simulator simulation.

[0005] The beneficial effects of this invention are: 1) Addressing the technical shortcomings of existing real-data-driven engine icing simulations that strictly adhere to physical time scales and require tens of minutes or even hours for a single training session, this invention establishes an automatic linkage mechanism between icing severity and acceleration ratio through a hybrid mapping model driven by an equivalent icing degree index. This compresses the real icing process, which takes tens of minutes or even hours, into a few minutes, significantly increasing the number of training cycles for handling icing emergencies per unit class hour. Simultaneously, through five-fold closed-loop physical conformal constraints, it ensures that the changing trends, characteristic inflection points, curve shapes, and proportional coupling relationships between the five core parameters after acceleration are completely consistent with the original real data. This fundamentally distinguishes it from phenomenal simulation schemes that deviate from real data, ensuring that the acceleration simulation results retain all the dynamic characteristics of the real data while meeting the CCAR-60-R1 airworthiness certification requirements. 2) Supports three de-icing methods: hot air, electric heating, and automatic mechanical de-icing. The icing and de-icing stages feature a smooth transition with cross-fade-in and gain, and seamless re-icing is supported after de-icing. It fully covers the entire emergency response chain from icing occurrence to development, detection, handling, and recovery, meeting the complete training loop needs of pilots from identification and judgment to operational response. Simultaneously, it automatically classifies icing into four levels—light, moderate, heavy, and extreme—based on an equivalent icing severity index. Each level automatically matches a corresponding acceleration factor, and supports continuously adjustable custom time scaling factors (1~16x), meeting the tiered training needs from initial familiarization to rapid response capability enhancement, and adapting to diverse teaching scenarios at different training stages and with varying pilot skill levels. 3) By using standardized icing data sequence files, aircraft parameter configuration tables, and multiple linear regression calibration procedures, it can be easily adapted to different aircraft engines without modifying the core algorithm, effectively reducing system porting costs. Through time scaling, it quickly presents icing fault characteristics (thrust decline, abnormal EGT increase, rotor speed decrease, reduced surge margin, etc.), significantly shortening the fault evolution waiting time. This allows pilots to gain more opportunities to identify, judge, and handle icing within limited training hours, strengthening muscle memory and situational awareness, and improving pilots' ability to identify and handle abnormal engine conditions and respond to emergencies under complex weather conditions. It has good aviation safety application value and industrialization prospects. Attached Figure Description

[0006] Other features, objects, and advantages of this application will become more apparent from the following detailed description of non-limiting embodiments with reference to the accompanying drawings.

[0007] Figure 1 This is a flowchart of the steps of an engine icing simulation method for flight simulators according to the present invention. Detailed Implementation

[0008] The present application will now be described in further detail with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are for illustrative purposes only and are not intended to limit the invention. Furthermore, it should be noted that, for ease of description, only the parts relevant to the invention are shown in the accompanying drawings.

[0009] It should be noted that, unless otherwise specified, the embodiments and features described in this application can be combined with each other. This application will now be described in detail with reference to the accompanying drawings and embodiments.

[0010] To more clearly explain the engine icing simulation method for flight simulators of the present invention, the following will be combined with... Figure 1 The steps in the embodiments of the present invention will be described in detail below.

[0011] The first embodiment of the present invention provides a method for simulating engine icing in flight simulators. (See also...) Figure 1 The method includes the following steps: S1, real-time acquisition of the original evolution sequence of icing thickness in different parts and the current environmental operating condition input parameters, the original evolution sequence of icing thickness in different parts is used by a weighted comprehensive model to generate the current equivalent icing degree index characterizing the icing state of multiple parts of the engine; In this embodiment, the original evolution sequence of icing thickness at different locations is extracted from real aircraft flight test certification data. The corresponding physical locations include the inlet wall, the leading edge of the fan blades, and the low-pressure compressor (LPC) stator blades; the icing thickness at the inlet wall is used as an example. Thick ice on the leading edge and surface of the fan blades LPC stator leaf ice thickness As a measure of the freezing state, the ice thickness ranges at the three locations are as follows: ∈[0,50]mm、 ∈[0,30]mm、 ∈[0,20]mm, initial values ​​are all 0; The ice thickness at each location is updated using a discrete iterative method: ; Where n is the simulation step number and the initial conditions , The change in ice thickness at the i-th location within the n-th simulation step is read from the standardized icing data sequence; The standardized icing data sequence is derived from real aircraft flight test certification data, including the icing growth phase (ice thickness monotonically increasing), the stabilization phase (ice thickness growth rate decreasing), and the melting / de-icing phase (ice thickness monotonically decreasing). Different aircraft models require corresponding standardized icing data sequences (files), with one data sequence file reference for each model, containing the storage path of the aircraft model's flight test certification data and the corresponding environmental and operating condition index keys. First, an icing data sequence index library is established, indexed in multiple dimensions according to environmental conditions (Temp, liquid water content LWC, average droplet diameter MVD) and engine operating conditions (intake flow rate Wa, fan speed N1). The index keys are: ; The current environmental input parameters are the real-time values ​​of the above five dimensions. During simulation, the data are collected in real time and the nearest neighbor matching method is used to find the closest index key and call the corresponding ice growth / melting data sequence. The valid range and anomaly handling rules for the above input parameters are as follows: Ambient temperature (°C), range -60 to +30°C, clamped to the boundary value when abnormal; Liquid water content (g / m³), range 0 to 3g / m³, set to 0 when below 0, clamped to 3 when above 3; Average water droplet diameter (μm), range 5 to 50μm, the nearest neighbor valid value is used to replace when abnormal; Engine intake airflow (kg / s), alarm triggered when the deviation from the rated value of the model exceeds ±20%; Fan speed N1, (%), range 0 to 105%, clamped to the boundary value when abnormal. The automatic phase switching logic is as follows: During the ice growth phase, the ice thickness monotonically increases over time. >0; when the ice thickness growth rate is less than 0 for three consecutive simulation steps. (A threshold value of 0.01 mm / s is used to determine whether the ice thickness growth rate has entered a stable phase.) The system enters a stable phase when the ice thickness growth rate reaches a stable level; after triggering de-icing, the system automatically switches to the de-icing data sequence. ; The weighted synthesis model normalizes the ice thickness evolution values ​​of each physical location, then nonlinearly weights the normalized ice thickness ratios of each location according to a preset aerodynamic influence sensitivity weights to obtain the dimensionless Equivalent Severity Index (ESI); wherein, each location Value, weighting coefficient For aircraft-related parameters, they need to be pre-configured according to the aircraft model: For each aircraft model, collect the actual flight test icing data of that aircraft model, calculate the correlation between ice thickness and net thrust reduction in each part, and determine the basic weights; Net thrust attenuation: ,in, This is the reference thrust when there is no icing. Pearson correlation coefficient calculation for the correlation between ice thickness and thrust attenuation at various locations: ; in, for The mean, for The mean; Initial weights based on correlation: ; Further optimization of the weights was achieved through multiple linear regression: a linear regression model of thrust decay and ice thickness was established. ; The regression coefficients a1, a2, and a3 are solved using the least squares method, and the final weights are determined as follows: ; The following weighting schemes are all based on the model parameter configurations obtained from the above calibration process; The normalization formula is: ; in, Let i ∈ {intake duct, fan, compressor} be the maximum permissible ice thickness for each part. After normalization, the ice thickness ratio for each part is a dimensionless value, ranging from [0,1]. The formula for weighted average ESI calculation is: ; Wherein, the weights satisfy The default scheme uses empirically fixed weights: (Air intake) (fan), (LPC), highlighting the characteristics of LPC as the main area affected by icing; in addition, a configurable dynamic weighting scheme can be used to combine engine physical characteristics and fault modes to achieve a more accurate comprehensive characterization of icing degree, for example: A sensitivity-driven dynamic weighting scheme is adopted to calculate the partial derivatives of engine thrust with respect to ice thickness at various locations in real time. The weight of each location at the current moment is determined by the proportion of the absolute value of each partial derivative to the sum of the absolute values ​​of all partial derivatives. This scheme does not require predefined weights and is dynamically adjusted entirely based on the engine's physical characteristics, thus best reflecting the actual impact of the current icing state on performance. An adaptive weighting scheme is adopted, which predefines a weight matrix based on the current flight stage, flight altitude, Mach number, and engine thrust status. The weight of the impact of icing on performance of different parts varies under different operating conditions. For example, the weight of LPC is increased to 0.6 under low-speed, high-thrust conditions, and the weight of the fan is increased to 0.4 under high-speed cruise conditions. Another approach is to use a fault mode identification (FMI) weighting scheme: dynamically adjust the weight distribution based on the current ESI level. Mild icing (ESI<0.25): Weights are balanced across different areas. ); Moderate icing (0.25≤ESI<0.5): Fan and LPC weights dominate ( ); Severe and extreme icing (ESI≥0.5): LPC weights absolutely dominate ( ); More accurately reflect the core risks of LPC stall and engine rollback under extreme icing conditions; Feature extraction is performed on the standardized icing data sequence to identify key inflection points. The feature types of key inflection points, also known as feature inflection points, include: icing start point (ESI monotonically increases from 0, and all parameters begin to deviate from the baseline), parameter decay inflection point (the second derivative of the thrust decrease rate reaches its maximum value, and thrust and N1 begin to decrease significantly), stabilization point (the ice thickness growth rate drops below the threshold, and the parameter change rate decreases), de-icing trigger point (instructor command or automatic de-icing trigger, ESI begins to monotonically decrease), and recovery completion point (ESI drops to zero and stabilizes, and all parameters recover to the baseline). A phased five-dimensional adaptive dynamic threshold system is adopted for key inflection point identification, abandoning the traditional single threshold logic, realizing differentiated inflection point detection in all stages of icing, and improving feature recognition accuracy. The threshold calculation formula is as follows: ; In the formula, This represents the noise standard deviation of the parameters within the current sliding window. is the average absolute value of the first derivative of the parameter within the current sliding window; k1 and k2 are the basic adjustable sensitivity coefficients, with typical values ​​of 3 and 0.2, respectively; The icing stage discrimination coefficient is set to 1.0 for the growth stage, 0.4 for the stable stage, and 0.7 for the de-icing stage. It is used to match the parameter change characteristics of different icing stages and solve the false detection problem caused by the uniform threshold across all stages. The parameter sensitivity weights are dynamically assigned based on the real-time change rates of core parameters (Fnet, EGT, N1, N2) to enhance the priority of inflection point identification for key fault parameters. The dynamic weight calculation formula is as follows: ; Where, j∈{Fnet, EGT, N1, N2}, The absolute value of the rate of change of each parameter after averaging over a 5-step sliding window is used. For example, in the early stage of icing, the rate of change of EGT is 5℃ / s, while the rate of change of thrust is only 0.5. After normalization, the weight of EGT is calculated as (5 / 5.5)×4≈3.6. After limiting, the upper limit of 2.0 is taken, while the lower limit of 0.5 is taken. At this time, the threshold responds first to the inflection point of EGT. When the rate of change of thrust reaches 8 in the later stage of icing and EGT is only 1, the thrust weight automatically rises to the upper limit of 2.0, realizing the automatic transfer of sensitive targets from EGT to thrust. This is a time-drift correction term used to offset the threshold error caused by parameter baseline drift during long-term simulations. A first-order low-pass filter is used to estimate the baseline drift in real time. ,in, This is the drift correction factor (typical value 0.001). These are the current simulation parameter values. This is the reference baseline value for this parameter under non-icing conditions; k3 and k4 are newly added adjustable coefficients, with typical values ​​of 0.5 and 0.3 respectively, which together with the base coefficients form a five-parameter coordinated control system; Judgment logic: Distinguish between three stages of icing growth, stabilization, and de-icing, and match differentiated thresholds for each stage. Only when three consecutive sampling points meet the stage adaptation threshold and the trend of the second derivative changes matches the stage characteristics is it determined to be a valid inflection point. This effectively solves the problems of missed detection, false detection, and lag in traditional threshold algorithms. After completing the inflection point identification, the annotation file of the feature inflection point is output for subsequent interpolation calculations. Based on publicly available engine icing airworthiness test data and NASA C-MAPSS40k simulation research results: Low-pressure compressor (LPC) icing blockage is the main influencing mechanism of engine icing. When the blockage reaches 20%, engine rollback can be observed. Typical performance changes: total intake pressure loss coefficient can reach 13.3%, compressor efficiency decreases by 10.4%, EGT increases by more than 40°C, and net thrust loss is 15.8%; Control system response: After icing reduces core flow, the control system automatically increases Wf to maintain EPR, causing N2 and EGT to rise first, and then the thrust eventually rolls back as icing worsens. Typical flight conditions for icing: altitude 11,500~39,000ft, ISA+6℃ and above, HIWC conditions, critical temperature for supercooled water droplet icing -9℃~-6℃; Three common de-icing methods are used: Hot air anti-icing: The triggering method is manual triggering from the instructor's platform or automatic triggering from the system. Automatic triggering threshold: ESI≥0.15 and ambient temperature∈[-15℃,0℃], lasting for 500ms; or ice thickness at any part reaches 30% of the maximum allowable value; operating parameters: bleed air volume accounts for about 5% of the total engine flow, heating temperature is about 200℃, performance impact: about 5% thrust loss, no obvious recovery lag; Electric heating anti-icing: Triggering methods are manual triggering from the instructor's station or automatic triggering from the system. Automatic triggering threshold: ESI ≥ 0.2 and ambient temperature ∈ [-20℃, -5℃], lasting 500ms; or ice thickness on fan blades ≥ 8mm. Operating parameters: heating power is randomly configured, using a periodic on / off mode. Performance impact: Only power consumption is generated, with no thrust loss and no significant recovery lag. Mechanical automatic de-icing: The triggering method is purely automatic (no manual triggering). The automatic triggering threshold meets any of the following conditions: ice thickness h1 in the intake duct ≥ 25mm; ice thickness h2 in the fan blades ≥ 15mm; ice thickness h3 in the LPC stator blades ≥ 10mm; ice thickness in any two parts simultaneously reaches 80% of the corresponding threshold, and engine N1 speed > 80%; Working principle: Relying on the centrifugal force generated by engine rotation to remove the ice layer; Performance impact: engine vibration increases briefly (vibration value increases by about 20%~30%), and there may be a 1~2s performance recovery lag; Re-icing: If the ESI does not drop to zero after de-icing, re-icing is allowed. Starting from the current ESI value, the icing sequence continues. For hot gas anti-icing, slow icing may still occur during de-icing, which is reflected by the non-monotonic decrease of ESI. S2, input the current environmental conditions and the current equivalent icing index into the composite state-space model to deduce the virtual time trajectory of the engine in the acceleration time domain, the changing trend of the time scaling factor, and the current instantaneous curvature and temporal relative proportion of each simulation core parameter; In this embodiment, the time-progressive composite state-space model is a hybrid mapping model that includes the current equivalent icing index (ESI) and the instantaneous time scaling factor (α). Specifically, the mapping relationship between ESI and the time scaling factor is first established: based on the ESI value range [0,1], it is divided into four levels: light icing (ESI∈[0,0.25]) corresponds to α=1, moderate icing (ESI∈(0.25,0.5]) corresponds to α=4, heavy icing (ESI∈(0.5,0.75]) corresponds to α=8, and extreme icing (ESI∈(0.75,1]) corresponds to α=16. The hybrid mapping model is a weighted combination of an exponential function and a sigmoid function, and the higher the current equivalent icing degree exponent, the larger the value of the instantaneous time scaling factor; the formula of the hybrid mapping model is as follows: ; in, , , represents the upper and lower bounds of the time scaling factor; β is the mixing coefficient, typically 0.4, which can be adjusted according to the training scenario: It is a pure exponential type (suitable for rapid processing training). The time is a pure S-shape (suitable for progressive recognition training); is the exponential coefficient, typically 2; k is the slope of the S-curve, typically 10; The midpoint of the S-curve, with a typical value of 0.5; Calculate the current ESI in real time and obtain the current instantaneous time scaling factor. The current instantaneous time scaling factor is smoothed using a first-order inertial filter, and the time scaling factor is output. The filtering method is as follows: ; in, =0.15, the transition time is about 20 simulation steps (200ms); the time scaling factor α is dynamically evolved with the real-time change of the equivalent icing index ESI, and its evolution trajectory is driven by the hybrid mapping model and smoothed by the first-order inertial filter to form a continuous trend of the time scaling factor change. The method for simulating the virtual time trajectory of the engine in the acceleration time domain is as follows: Starting from the initial moment of the simulation, the virtual time of the previous simulation step is used as the starting virtual time of the current simulation step. ; Obtain the smoothed time scaling factor for the current simulation step. The time scaling factor is compared with the simulation step size. Multiply them to obtain the virtual time increment for the current simulation step; Add the starting virtual time to the virtual time increment to obtain the virtual time of the next moment of the current simulation step; ; in, Let n be the simulation time for the nth simulation step. The simulation step size is typically 10ms. The physical meaning of virtual time is the actual time increment in the icing data sequence. To ensure the accuracy of virtual time, the following two measures are adopted: The first-order inertial filtering stage described above is used to smooth the instantaneous time scaling factor, avoiding the time scaling factor being affected by ESI measurement noise or level boundary jitter. Frequent jumps; Using the predictive-correction integral (trapezoidal method): Using the time scaling factor of the current simulation step, forward prediction is performed to obtain the predicted virtual time increment, which is also the prediction step: ; The predicted virtual time increment is weighted and corrected using the average of the time scaling factors of the current simulation step and the previous simulation step. The corrected virtual time increment replaces the original increment, completing the final accumulation of virtual time, which yields the correction step. ; The final cumulative evolution process of this virtual time is used to characterize the virtual time trajectory in the engine icing acceleration time domain; Start and end point alignment rules: The starting point of the freezing stage is fixed at... The termination point is the last moment of the data sequence, and the starting point of the de-icing stage is the virtual time position corresponding to the de-icing trigger moment; cross-segment processing: the virtual time axis of each stage of freezing → de-icing → recovery is maintained independently, and the virtual time is not reset when switching stages; Freezing stage: , ( (Maximum time for the icing growth data sequence). De-icing phase: Recorded when de-icing is triggered The virtual time is counted independently during the de-icing phase: ; Recovery Phase: After de-icing is completed, the virtual time of the recovery phase is counted independently. ; Each stage's virtual timeline is maintained independently, and the global virtual time counter is not reset when switching stages. The method for obtaining the current instantaneous curvature of each simulation core parameter is as follows: For each core parameter's discrete sequence, calculate its first and second derivatives in real time. Divide the absolute value of the second derivative by the square of the first derivative plus one raised to the power of two thirds to obtain the current instantaneous curvature of that core parameter. : ; Among them, the time series curve of icing parameters , The first derivative of the interpolation curve (the rate of change of the parameters). The second derivative (acceleration due to parameter change); The method for obtaining the temporal relative proportion is as follows: synchronously obtain the parameter value of the core parameter in the current simulation step, calculate its ratio with the corresponding core parameter value in the original real data at the same virtual time, and obtain the current temporal relative proportion of the core parameter; S3, based on the deviation difference between the current instantaneous curvature and the time-series relative proportion of each simulation core parameter from the historical real engine response characteristics, adaptively and dynamically update the correction boundary weight matrix in the multi-closed-loop physical conformal constraint; perform conformal optimization calculation on the time-progression composite state-space model after updating the correction boundary weight matrix using a polynomial spline interpolation algorithm, solve and output the optimal acceleration smooth engine performance parameter command; In this embodiment, the standardized icing data sequence is first stored as a time array according to the parameter type. and parameter value array Use binary search to find the solution that satisfies the condition. Index k; Boundary handling: Pick , Pick ; interval The polynomial within the range, specifically the cubic spline polynomial, is as follows: ; Solving the tridiagonal system of equations using natural boundary conditions (M[0]=0, M[N]=0) yields the second derivative M[k], which in turn allows for the calculation of coefficients: ; ; ; ; Will Substituting into the polynomial yields the interpolation result. ; The specific mechanism by which the polynomial spline interpolation algorithm performs calculations is as follows: Using the virtual time trajectory as a data index, the engine performance parameters are calculated by default using a cubic spline interpolation function in the time-progression composite state-space model; When the evolution node of the virtual time trajectory approaches the characteristic inflection point caused by the environmental conditions, the polynomial spline interpolation algorithm is switched to piecewise cubic Hermite interpolation without disturbance, and monotonicity preservation constraints are applied to cut off possible spurious oscillations or performance parameter overshoot. When the virtual time passes through a pre-labeled feature inflection point, the system switches to piecewise cubic Hermite interpolation to ensure that the inflection point position and change trend are consistent with the original data. Four control points t[k-1], t[k], t[k+1], t[k+2] and their corresponding parameter values ​​are obtained by taking one sampling point before and after the inflection point. Hermite interpolation is then performed using the inflection point parameter values ​​and the rate of change as constraints. The formula for calculating the rate of change m[k] at the inflection point is: ; ; Three Hermite interpolations were applied to the interpolation intervals before and after the inflection point to ensure that the values ​​and slopes at the inflection point were completely consistent with the original data. ; To avoid physical distortion in Hermite interpolation, multiple closed-loop physical conformal constraints are applied. Unlike traditional single, independent constraint rules, this is a progressive, interconnected verification architecture where constraints are nested, mutually verified, and corrected in a closed loop. It addresses simulation distortion issues in scenarios with 1-16x global acceleration by constraining the interpolation curve shape from multiple dimensions. These multiple closed-loop physical conformal constraints include: (1) Monotonicity preservation constraint: If the real data monotonically increases / decreases within the interval of adjacent sampling points, the interpolation curve maintains the same monotonicity throughout the entire process, prohibiting overshoot or oscillation, and conforming to the inherent physical change law of parameters such as ice thickness and thrust loss during the icing process, providing the basic physical boundary for all subsequent constraints; (2) Physical extreme value constraint: The value of the interpolation curve in any interval shall not exceed the maximum and minimum values ​​of the sampling points at both ends of the interval. Based on the physical prior that the icing parameter has no sudden change, the Runge oscillation phenomenon is eliminated and the parameter output boundary is locked. (3) Second derivative continuity constraint: By dynamically adjusting the interpolation boundary conditions, the parameter value, first derivative and second derivative at the inflection point are triple continuous, eliminating the curve sharp point and sudden change problem caused by manual interpolation, ensuring the smoothness of the parameter change curve after acceleration, and conforming to the gradual change characteristics of the real engine working condition. (4) Constant curvature constraint at inflection points: During the interpolation process, the curvature at the inflection points is kept consistent with the original real data; (5) Parameter proportional closed-loop constraint: Maintain the proportional relationship between the five core parameters of engine net thrust (Fnet), exhaust temperature (EGT), fan speed (N1), core engine speed (N2), and fuel flow (Wf) consistent with the original real data. The deviation judgment and correction method are described in the constraint update section. The specific proportional relationship includes: Ratio: Physically, it represents the change in net engine thrust corresponding to a unit change in Equivalent Erosion Index (ESI); the maintenance mechanism is that all parameters use the same virtual time interpolation, so the ratio is naturally maintained. Vib / ESI Ratio: Physically, it represents the ratio of engine vibration value to Equivalent Erosion Index; the maintenance mechanism is that all parameters use the same virtual time interpolation, so the ratio is naturally maintained. EGT / Fnet Ratio: Physically, it represents the ratio of engine exhaust temperature to net thrust; the maintenance mechanism is that all parameters use the same virtual time interpolation, so the ratio is naturally maintained. N1 / N2 Ratio: Physically, it represents the ratio of fan speed (N1) to core engine speed (N2); the maintenance mechanism is that all parameters use the same virtual time interpolation, so the ratio is naturally maintained. Wf / Fnet Ratio: Physically, it represents the ratio of engine fuel flow rate to net thrust; the maintenance mechanism is that all parameters use the same virtual time interpolation, so the ratio is naturally maintained. Maintenance Mechanism: All parameters use the same virtual time axis interpolation, and the interpolation point corresponds to the same virtual time. Therefore, the ratio of any two parameters is completely consistent with the ratio at the same time in the original data. The above five constraints constitute a progressive closed-loop verification architecture of "basic boundary → smooth shape → feature fidelity → global linkage": First, monotonicity and extreme value constraints lock the underlying boundary of the physical rationality of parameters; then, the second derivative continuity constraint optimizes the basic shape of the curve; next, curvature constraints accurately preserve the core features of the fault inflection point; finally, parameter ratio constraints complete the global parameter linkage calibration. Each constraint mutually restricts and iteratively corrects the others. Preceding constraints provide a compliance basis for subsequent constraints, and subsequent constraints in turn correct subtle deviations in preceding constraints, forming a complete closed loop. The correction boundary weight matrix in the multi-loop physical conformal constraint is adaptively and dynamically updated based on the deviation difference between the current instantaneous curvature and the temporal relative proportion of each simulation core parameter and the historical real engine response characteristics. Specifically: Calculate the instantaneous curvature of each core simulation parameter in the current simulation step, and calculate the deviation of this curvature from the curvature of the original real data at the same virtual moment to obtain the curvature deviation value; The specific implementation of the constant curvature constraint at the inflection point is as follows: taking the marked feature inflection point as the center, a predetermined number of standard simulation steps before and after it are taken as the curvature comparison interval; the curvature deviation value between the accelerated simulation curve curvature and the original real curve curvature is calculated; when the curvature deviation value is greater than the first preset threshold, the slope constraint weight at the Hermite interpolation inflection point is dynamically fine-tuned, and the value of the second derivative is iteratively optimized to force the curvature deviation value to converge to within the first preset threshold; specifically, taking the marked feature inflection point as the center, the standard simulation steps before and after the inflection point are taken as the center, and the curvature deviation value is ... Centered on the curve, five standard simulation steps (totaling 100ms) are extracted before and after the curve as the curvature comparison interval. This range can completely cover the core feature interval of the icing fault abrupt change, balancing accuracy and computational efficiency; the curvature of the accelerated simulation curve is calculated. Curvature of the original true curve Curvature deviation value: ; when When the curvature deviation is greater than 5%, the curvature closed-loop correction is initiated: the slope constraint weight at the Hermite interpolation inflection point is dynamically fine-tuned, the second derivative value is iteratively optimized, and the curvature deviation is forced to converge to within 5%, so as to ensure that the sudden change characteristics and curve shape of the icing fault inflection point under high-speed acceleration scenario are consistent with the actual working conditions. Calculate the ratio of the parameter value of each simulation core parameter in the current simulation step to the corresponding parameter value in the original real data at the same virtual time, obtain the temporal relative ratio, and calculate the deviation of this ratio from the corresponding parameter ratio in the original real data, obtain the ratio deviation value; specifically, synchronously acquire the parameter value of the core parameter in the current simulation step, calculate its ratio to the corresponding core parameter value in the original real data at the same virtual time, obtain the current temporal relative ratio of the core parameter; take the current temporal relative ratio of any two sets of core parameters, calculate the deviation between the two sets of ratios, and use it as the ratio deviation value of the two sets of core parameters; lock the five core parameters of engine net thrust (Fnet), exhaust temperature (EGT), fan speed (N1), core engine speed (N2), and fuel flow (Wf), and define the real-time ratio deviation of any two sets of core parameters a and b: ; The preset proportional deviation threshold is 3%. When any set of core parameters... When the value is greater than 3%, a global closed-loop correction is initiated: the physical parameters with stronger change trends (ice thickness, ESI) are fixed first, the interpolation output values ​​of the five core simulation parameters are finely adjusted in reverse, the parameter change rate is corrected synchronously, and all parameter ratio deviations are uniformly converged to within the 3% threshold to ensure that the parameter coupling relationship between the accelerated simulation and the real icing condition is completely consistent. The curvature deviation value and the proportional deviation value are used as the deviation judgment criteria for the inflection point curvature constant constraint and the parameter proportional closed-loop constraint, respectively: when the curvature deviation value of any simulation core parameter is greater than the first preset threshold (typical value 5%), the second derivative constraint weight term of the parameter at the discrete interpolation point corresponding to the virtual time trajectory is increased, and the first derivative constraint weight term is decreased to suppress curvature distortion; when the proportional deviation value of any two sets of simulation core parameters is greater than the second preset threshold (typical value 3%), the interpolation output weight of the parameter value of the one with weaker change trend rigidity among the two sets of parameters is increased to force the proportional relationship to be restored. The correction boundary weight matrices of each constraint are maintained independently. The correction results of the preceding constraints serve as the input boundary conditions of the subsequent constraints. The correction results of the subsequent constraints are used to back-verify whether the correction of the preceding constraints leads to new deviations exceeding the limits, forming a progressive linkage closed-loop correction system. Each constraint restricts and iteratively corrects itself, together forming a complete verification link from basic boundary constraints to global parameter linkage calibration. Raw data sampling rate uniformity: Resampling is used to unify the sampling rate to a standard sampling rate fs (typically 100Hz). During resampling, an anti-aliasing low-pass filter (cutoff frequency fs / 2) is used for pre-filtering. Accelerated resampling: The output sampling rate is consistent with the analog master clock (100Hz). Cubic spline interpolation naturally completes resampling; aliasing avoidance: When α>4, an optional first-order low-pass filter step (cutoff frequency fs) is added after interpolation. Parameter range constraint: The interpolated parameter must not exceed the maximum and minimum values ​​of the parameter in the original data; if it does, it will be truncated to the boundary value. Parameter rate of change constraint: The rate of change of parameters over time after acceleration must not exceed α times the maximum rate of change of the original data. ; Alarm trigger timing preservation: The alarm trigger times after acceleration maintain the same order as the alarm trigger times in the original data; S4. After obtaining the anti-icing event trigger signal and processing it with a delay buffer, it is superimposed and fused with the optimal acceleration smoothing engine performance parameter command that has been processed by anti-aliasing low-pass filtering to generate the comprehensive engine environmental response parameter of the whole domain. In this embodiment, the anti-icing event trigger signal is either manually triggered by the instructor's console or automatically triggered by the system. The automatic trigger threshold is executed based on the trigger conditions of the three de-icing methods in S1; After obtaining the anti-icing event trigger signal, it also includes abnormal interruption fidelity control for handling de-icing conditions: When the anti-icing event trigger signal indicates that the de-icing state has been entered, the data sequence that maps the physical attenuation of de-icing is switched to be interpolated and advanced. The time scaling factor for the de-icing stage is calculated by multiplying the scaling factor value corresponding to the equivalent icing degree index of the same level in the icing stage by N times before execution. If the current equivalent icing degree index is still greater than zero after the de-icing command terminates, the remaining equivalent icing degree index recorded by the controller and the virtual time position of the freeze are used as the new starting point to seamlessly transition to the re-icing evolution sequence. Specifically, the instantaneous time scaling factor for the de-icing stage is taken as 0.8 times the scaling factor value corresponding to the same equivalent icing degree index in the icing stage, and the transition is smoothed using the same inertial filtering method. When the equivalent icing index drops below the icing level threshold, the time scaling factor is adjusted after a 5-step delay. If the equivalent icing index does not drop to zero after de-icing, the simulation continues from the current equivalent icing index value and virtual time position to achieve re-icing simulation. Anti-aliasing low-pass filtering: The optimal acceleration smoothing engine performance parameter command input is given to a Butterworth anti-aliasing low-pass filter with an adaptive cutoff frequency to filter out high-frequency numerical noise that may be generated during the interpolation process; The complementary state superposition and fusion includes: inputting the optimal acceleration smoothing engine performance parameter command into a Butterworth anti-aliasing low-pass filter with an adaptive cutoff frequency; extracting the high-frequency disturbance response caused by the intercepted anti-icing event trigger signal; using a cross-fade-in / fade-out gain algorithm (using a cross-fade-in / fade-out gain algorithm for smooth transition, taking the icing stage parameter command and de-icing stage parameter command within each preset time before and after the state switching time as the center, calculating the weighted sum according to the method of the icing stage weight gradually changing from 1 to 0 and the de-icing stage weight gradually changing from 0 to 1, replacing the step jump generated by hard switching); and adding the weighted low-frequency base of the filtered stable performance parameter with the dynamic high-frequency response of the anti-icing system to obtain the comprehensive engine environmental response parameter covering the entire icing and de-icing process. S5. Based on the changing trend of the virtual time trajectory and time scaling factor, the global integrated engine environmental response parameters are transformed into a dynamic evolution sequence by combining the state management mechanism and sent to the underlying interface to control the flight simulator simulation. In this embodiment, the comprehensive engine environmental response parameters include: net thrust Fnet: unit is lbf / N, update frequency is 100Hz, output to the aerodynamic model and instrument display system; fan speed (N1): unit is % rpm, update frequency is 100Hz, output to the instrument display and engine model; core engine speed N2: unit is % rpm, update frequency 100Hz, output to instrument display system; Vibration value: unit as specified for corresponding model, update frequency 100Hz, output to instrument and engine vibration system; Exhaust temperature EGT: unit is ℃, update frequency 100Hz, output to instrument display system; Engine pressure ratio EPR: dimensionless, update frequency 100Hz, output to engine model; Fuel flow rate Wf: unit is lbm / s, update frequency 100Hz, output to engine model; Equivalent icing index ESI: dimensionless, update frequency 100Hz, output to instructor display system; Alarm status: Boolean type, update frequency 100Hz, output to instrument and audible alarm system; Synchronization mechanism adopts simulator master clock interrupt mode, total delay does not exceed one simulation step (100ms). The alarm types and configurations output during the engine icing simulation are as follows: EGT over-temperature alarm: Triggering condition is that EGT is greater than EGTlimit, debouncing time is 20ms, and priority is high. Low thrust alarm: Triggered when Fnet is less than Fnetmin, debouncing time is 20ms, and priority is high; Surge warning: The triggering condition is that the surge margin is less than SMmargin_min, the dejitter time is 20ms, and the priority is medium. Freezing warning: The triggering conditions are ESI greater than 0.1, debouncing time 40ms, and low priority; The alarm thresholds EGTlimit, Fnetmin, and SMmargin_min are aircraft-specific parameters and must be pre-configured according to the aircraft model. The parameter values ​​corresponding to the actual trigger time of each alarm in the original icing data sequence for that aircraft model are directly used. For example, if an EGT alarm is triggered 180 seconds after the start of icing in the test flight data for that aircraft model, corresponding to EGT=850℃, then EGTlimit for that aircraft model is 850℃. Combination logic: When multiple alarms trigger simultaneously, they are displayed in descending order of priority, and alarms of the same priority are not triggered repeatedly. The state management mechanism includes a four-state finite state machine, the four states being standby, icing, deicing, and recovery. The state transition rules are as follows: Upon receiving an icing command while in standby mode, the system switches to icing mode and executes the following action: Initialize virtual time. =0, Equivalent Icing Index (ESI)=0, load the icing data sequence corresponding to the working condition; Upon receiving a de-icing command while in an icing state, the system switches to de-icing state and executes the following actions: Records the current ESI value and virtual time. Load the corresponding de-icing data sequence; When the Equivalent Icing Index (ESI) reaches its upper limit (ESI≥1.0) while the engine is in an icing state, switch to the recovery state and perform the following actions: trigger an extreme icing alarm and start the engine parameter recovery process. After the equivalent icing index drops below zero during de-icing and remains below zero for a predetermined period of time ( ESI If the value is less than 0.01 and lasts for 10 seconds, or after receiving a stop icing command, switch to the recovery state and execute the following actions: stop the de-icing process and start the engine parameter recovery process; upon receiving a reset command, unconditionally switch back to the standby state and execute the following actions: reset all simulation parameters to the engine baseline value; after all engine parameters are restored to the baseline value in the recovery state, switch back to the standby state; in any state, upon receiving a reset command, unconditionally switch back to the standby state. Once all engine parameters are restored to baseline values ​​during the recovery process, the system switches back to standby mode and performs the following actions: outputs a simulation recovery completion signal; and unconditionally switches back to standby mode upon receiving a reset command. Upon receiving a command to set the level or time scaling factor in any state, maintain the current state and execute the following actions: update the time scaling factor α value, freeze the current ESI and The simulation continues from the current state as a new starting point; if a reset command is received in any state, the system will unconditionally switch back to the standby state. The training control interface includes at least the following commands: Start Icing (START_ICING), Stop Icing (STOP_ICING), Set Icing Level (SET_LEVEL, parameter level∈{1,2,3,4}), Set Custom Time Scaling Factor (SET_ALPHA, parameter α∈[1,16], step size 0.5), Trigger De-icing (TRIGGER_DEICE), and Reset (RESET). Therefore, by acquiring the original evolution sequence of icing thickness in different parts of the engine from actual aircraft flight testing and certification, and the input parameters of the current environmental conditions, the ice thickness of each part is normalized and weighted to generate a dimensionless equivalent icing degree index, so as to uniformly represent the comprehensive icing state of multiple parts of the engine. At the same time, a multi-dimensional index library is established according to ambient temperature, liquid water content, average water droplet diameter, intake air flow rate, and fan speed. A phased five-dimensional adaptive dynamic threshold system is used to extract and label the icing start point, parameter decay inflection point, stability point, de-icing trigger point, and recovery completion point, forming a standardized icing data sequence and characteristic inflection points. The equivalent icing degree index is input into a hybrid mapping model consisting of a weighted combination of exponential and sigmoid functions, mapped to an instantaneous time scaling factor. After smoothing with a first-order inertial filter, a continuously changing time scaling factor is generated, achieving dynamic linkage where the acceleration multiplier increases with the severity of icing. The virtual time accumulation is driven by the time scaling factor, and the virtual time trajectory is derived using a prediction-correction integral method. During the advancement of the virtual time axis, the instantaneous curvature and temporal relative proportion of each simulation core parameter are obtained in real time by calculating the first and second derivatives, which are used for subsequent shape preservation deviation determination. Using the virtual time trajectory as an index, the engine performance parameters are calculated by default using cubic spline interpolation in the standardized icing data sequence. When the evolution node approaches the feature inflection point, it switches to piecewise cubic Hermite interpolation without disturbance to ensure that the parameter values ​​and rate of change at the inflection point are consistent with the original data. On this basis, five physical conformal constraints are applied: monotonicity preservation, physical extrema, second derivative continuity, constant inflection point curvature, and parameter proportional closed loop. These constraints are used to prevent overshoot oscillation of the interpolation curve, limit the parameter output boundary, eliminate abrupt changes at the curve apex, maintain the curvature characteristics of the inflection point, and lock the proportional coupling relationship between the five core parameters. The curvature deviation and proportional deviation between the accelerated simulation curve and the original real curve are used as the judgment criteria. The correction boundary weight matrix of each constraint is adaptively updated. After conformal optimization calculation, the optimal acceleration smoothing engine performance parameter command is output so that the parameter change trend, inflection point characteristics, and proportional relationship between parameters after acceleration are consistent with the real data. Real-time monitoring of anti-icing trigger signals; smooth transition at the boundary between icing and de-icing stages using cross-fade-in and fade-out gain; weighted sum calculated with the state switching moment as the center, with the weight of the icing stage gradually changing from 1 to 0 and the weight of the de-icing stage gradually changing from 0 to 1, replacing the step jump caused by hard switching; splicing the timing of performance parameter commands for the icing and de-icing stages to generate a comprehensive engine environmental response parameter covering the entire icing and de-icing process. Based on the virtual time trajectory and the changing trend of the time scaling factor, the simulation process is managed by a finite state machine containing four states: standby, icing, de-icing, and recovery. State transitions are driven and corresponding actions are executed according to the instructor's console control commands and the equivalent icing degree index. The comprehensive engine environmental response parameters are transformed into a dynamic evolution sequence and sent to the underlying interface to control the flight simulator simulation. Simultaneously, net thrust, fan speed, core engine speed, vibration value, exhaust temperature, engine pressure ratio, fuel flow rate, equivalent icing degree index, and alarm status are output to each subsystem of the flight simulator at a 100Hz update frequency. Hierarchical alarm logic is output according to priority. Training control interfaces are constructed through commands to initiate icing, stop icing, set icing level, set a custom time scaling factor, trigger de-icing, and reset, supporting real-time intervention by the instructor's console.

[0012] Those skilled in the art will understand that, for the sake of convenience and brevity, the specific working process and related explanations of the methods described above can be found in the corresponding processes in the foregoing system embodiments, and will not be repeated here.

[0013] Those skilled in the art will recognize that the modules and method steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of both. The programs corresponding to the software modules and method steps can be placed in random access memory (RAM), main memory, read-only memory (ROM), electrically programmable ROM, electrically erasable programmable ROM, registers, hard disks, removable disks, CD-ROMs, or any other form of storage medium known in the art. To clearly illustrate the interchangeability of electronic hardware and software, the components and steps of the various examples have been generally described in terms of functionality in the foregoing description. Whether these functions are implemented in electronic hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of the invention.

[0014] The terms “first”, “second”, etc., are used to distinguish similar objects, not to describe or indicate a specific order or sequence.

[0015] The term "comprising" or any other similar term is intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus / device that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent in such process, method, article, or apparatus / device.

[0016] The technical solution of the present invention has been described above with reference to the preferred embodiments shown in the accompanying drawings. However, it will be readily understood by those skilled in the art that the scope of protection of the present invention is obviously not limited to these specific embodiments. Without departing from the principles of the present invention, those skilled in the art can make equivalent changes or substitutions to the relevant technical features, and the technical solutions after such changes or substitutions will all fall within the scope of protection of the present invention.

Claims

1. A method for simulating engine icing in flight simulators, characterized in that, The method includes the following steps: The original evolution sequence of icing thickness in different parts of the engine and the current environmental operating conditions input parameters are acquired in real time. The original evolution sequence of icing thickness in different parts of the engine is used to generate the current equivalent icing degree index, which characterizes the icing state of multiple parts of the engine, through a weighted comprehensive model. The current environmental condition input parameters and the current equivalent icing degree index are input into a composite state-space model to deduce the virtual time trajectory of the engine in the acceleration time domain, the changing trend of the time scaling factor, and the current instantaneous curvature and temporal relative proportion of each simulation core parameter. Based on the deviation difference between the current instantaneous curvature and the time-series relative proportion of each simulation core parameter from the historical real engine response characteristics, the correction boundary weight matrix in the multi-closed-loop physical conformal constraint is adaptively and dynamically updated; the time-progression composite state-space model after updating the correction boundary weight matrix is ​​subjected to conformal optimization calculation through polynomial spline interpolation algorithm, and the optimal acceleration smoothing engine performance parameter command is solved and output. After acquiring the anti-icing event trigger signal and processing it with a delay buffer, it is superimposed and fused with the optimal acceleration smoothing engine performance parameter command that has been processed by anti-aliasing low-pass filtering to generate a comprehensive engine environmental response parameter across the entire domain. Based on the changing trends of the virtual time trajectory and time scaling factor, the global integrated engine environmental response parameters are transformed into a dynamic evolution sequence by combining the state management mechanism and sent to the underlying interface to control the flight simulator simulation.

2. The engine icing simulation method for flight simulators according to claim 1, characterized in that, The original evolution sequence of icing thickness in each part is extracted from real aircraft flight test certification data. The corresponding physical parts include the inlet wall, the leading edge of the fan blades, and the low-pressure compressor stator blades. The weighted synthesis model normalizes the icing thickness evolution values ​​of each physical part, and then nonlinearly weights the normalized icing thickness ratios of each part according to a preset aerodynamic influence sensitivity weight to obtain the dimensionless current equivalent icing degree index.

3. The engine icing simulation method for flight simulators according to claim 1, characterized in that, The time-progression composite state-space model is a hybrid mapping model that includes the current equivalent icing degree index and the instantaneous time scaling factor. The hybrid mapping model is a weighted combination of an exponential function and a sigmoid function, and the higher the current equivalent icing degree index, the larger the value of the instantaneous time scaling factor. The instantaneous time scaling factor is smoothed using a first-order inertial filter to output the time scaling factor.

4. The engine icing simulation method for flight simulators according to claim 3, characterized in that, The method for simulating the virtual time trajectory of the engine in the acceleration time domain is as follows: Starting from the initial moment of the simulation, the virtual time of the previous simulation step is used as the starting virtual time of the current simulation step; Obtain the smoothed time scaling factor for the current simulation step, and multiply the time scaling factor by the simulation step size to obtain the virtual time increment for the current simulation step. Add the starting virtual time to the virtual time increment to obtain the virtual time of the next moment of the current simulation step; Using the time scaling factor of the current simulation step, forward prediction is performed to obtain the predicted virtual time increment; The predicted virtual time increment is weighted and corrected by the average of the time scaling factors of the current simulation step and the previous simulation step. The corrected virtual time increment replaces the original increment, and the final accumulation of virtual time is completed. The final accumulation and evolution process of virtual time is used to characterize the virtual time trajectory in the engine icing acceleration time domain.

5. The engine icing simulation method for flight simulators according to claim 1, characterized in that, The correction boundary weight matrix in the multi-loop physical conformal constraint is adaptively and dynamically updated based on the deviation difference between the current instantaneous curvature and the temporal relative proportion of each simulation core parameter and the historical real engine response characteristics. Specifically: Calculate the instantaneous curvature of each core simulation parameter in the current simulation step, and calculate the deviation of this curvature from the curvature of the original real data at the same virtual moment to obtain the curvature deviation value; Calculate the ratio of the parameter value of each simulation core parameter in the current simulation step to the corresponding parameter value in the original real data at the same virtual time step to obtain the time-series relative ratio, and calculate the deviation of this ratio from the corresponding parameter ratio in the original real data to obtain the ratio deviation value. The curvature deviation value and the proportional deviation value are used as the deviation judgment criteria for the inflection point curvature constant constraint and the parameter proportional closed-loop constraint, respectively: when the curvature deviation value of any simulation core parameter is greater than the first preset threshold, the second derivative constraint weight term of the parameter at the discrete interpolation point corresponding to the virtual time trajectory is increased, and the first derivative constraint weight term is decreased to suppress curvature distortion; when the proportional deviation value of any two sets of simulation core parameters is greater than the second preset threshold, the interpolation output weight of the parameter value of the one with the weaker change trend rigidity among the two sets of parameters is increased to force the proportional relationship to be restored. The correction boundary weight matrices of each constraint are maintained independently. The correction results of the preceding constraints serve as the input boundary conditions for the subsequent constraints. The correction results of the subsequent constraints are used to back-verify whether the correction of the preceding constraints leads to new deviations exceeding the limits, thus forming a progressive linkage closed-loop correction system.

6. The engine icing simulation method for flight simulators according to claim 5, characterized in that, The specific implementation of the constant curvature constraint at the inflection point is as follows: taking the marked feature inflection point as the center, extracting a predetermined number of standard simulation steps before and after as the curvature comparison interval; calculating the curvature deviation value between the accelerated simulation curve curvature and the original real curve curvature; when the curvature deviation value is greater than the first preset threshold, dynamically fine-tuning the slope constraint weight at the Hermite interpolation inflection point, iteratively optimizing the value of the second derivative, and forcibly converging the curvature deviation value to within the first preset threshold.

7. The engine icing simulation method for flight simulators according to claim 1, characterized in that, The specific mechanism by which the polynomial spline interpolation algorithm performs calculations is as follows: Using the virtual time trajectory as a data index, the engine performance parameters are calculated by default using a cubic spline interpolation function in the time-progression composite state-space model; When the evolution node of the virtual time trajectory approaches the characteristic inflection point caused by the environmental conditions, the polynomial spline interpolation algorithm is switched to piecewise cubic Hermite interpolation without disturbance, and monotonicity preservation constraints are applied to cut off possible spurious oscillations or performance parameter overshoot.

8. The engine icing simulation method for flight simulators according to claim 1, characterized in that, After obtaining the anti-icing event trigger signal, it also includes abnormal interruption fidelity control for handling de-icing conditions: When the anti-icing event trigger signal indicates that the de-icing state has been entered, the data sequence that maps the physical attenuation of de-icing is switched to be interpolated and advanced. The time scaling factor for the de-icing stage is N times the scaling factor value corresponding to the equivalent icing degree index of the same level in the icing stage. If the current equivalent icing degree index is still greater than zero after the de-icing command is terminated, the remaining equivalent icing degree index recorded by the controller and the virtual time position of the freeze are used as the new starting point to seamlessly transition to the re-icing evolution sequence.

9. The engine icing simulation method for flight simulators according to claim 1, characterized in that, The superposition and fusion of complementary states includes: The optimal acceleration smoothing engine performance parameter command is input into a Butterworth anti-aliasing low-pass filter with an adaptive cutoff frequency; The high-frequency disturbance response triggered by the intercepted anti-icing event is extracted, and a cross-fade-in / fade-out gain algorithm is used to add the low-frequency base of the filtered stable performance parameters to the dynamic high-frequency response of the anti-icing system with weights.

10. The engine icing simulation method for flight simulators according to claim 1, characterized in that, The state management mechanism includes a four-state finite state machine, comprising a standby state, an icing state, a de-icing state, and a recovery state. Upon receiving an icing start command in the standby state, the system switches to the icing state. In the icing state, upon receiving a de-icing trigger command or when the equivalent icing degree index reaches its upper limit, the system switches to either the de-icing state or the recovery state. In the de-icing state, after the equivalent icing degree index drops below zero and remains below zero for a predetermined time, or upon receiving a stop icing command, the system switches to the recovery state. In the recovery state, after all engine parameters return to baseline values, the system switches back to the standby state. Upon receiving a reset command in any state, the system unconditionally switches back to the standby state.

Images