Intelligent Control Method for Defogging Efficiency of Aero-engines Based on Multi-Source Data Fusion
By using multi-source data fusion and adaptive reshaping, the problems of control decision deviation and execution unit performance degradation in the fog-eliminating efficiency regulation of aero-engines were solved, and efficient and reliable control of the engine was achieved in complex environments.
Patent Information
- Application Number
- CN202611295511.X
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-08-25
- Publication Date
- 2026-11-17
AI Technical Summary
Existing aero-engine defogging performance control technologies cannot adapt to the dynamic and unsteady evolution of the engine's internal thermodynamic state in complex operating environments, causing control decisions to deviate from the physical optimal solution. Furthermore, they cannot take into account the performance degradation and health status of the defogging actuator, resulting in a decline in control performance.
By fusing multi-source data, we obtain multi-source sensing data of the engine and health status parameters of the defogging execution unit. Based on the heat exchange field distribution data of the combustion chamber, we extract the physical monotonicity constraint set, calculate the health degradation index, and perform adaptive renormalization and weight compensation on the fuzzy membership function to generate a stable fusion weight set and a defogging efficiency control decision vector to drive the defogging execution unit to perform control.
It significantly improves the physical compliance of fog dissipation control decisions and the reliability of command execution, enhances the engine's safety and adaptability across the entire flight envelope, and ensures the integrity of physical constraints and health status from the initial design state to the full life cycle service state.
Smart Images

Figure CN123109872A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of intelligent control technology, and more specifically, to an intelligent control method for the fog-removing efficiency of aero-engines based on multi-source data fusion. Background Technology
[0002] Existing technologies for controlling the fogging performance of aero-engines face multi-dimensional technical bottlenecks when dealing with complex operating environments. In particular, the fusion of multi-source heterogeneous sensor data encounters a model-reality mismatch when processing data that aligns with underlying physical laws. As the engine's operating envelope expands, the inherent coupling relationships, nonlinear characteristics, and time-varying properties of different physical fields (such as temperature, flow, and heat transfer fields) become increasingly pronounced. Traditional control strategies based on fixed mapping relationships or preset fuzzy rules cannot adapt to the dynamic and unsteady evolution of the engine's internal thermodynamic state during actual flight. In complex fogging scenarios, when the control model deviates from the actual heat transfer patterns in the combustion chamber, the system still mechanically infers based on pre-calibrated fuzzy membership functions, potentially leading to control decisions that significantly deviate from the physically optimal solution. To ensure the real-time performance of the control system, existing methods are forced to rely on simplified, decoupled control models, artificially ignoring the verification value of strong constraints such as the physical monotonicity of the combustion chamber heat transfer field. This simplification strategy results in the omission of obvious physical characteristics. In particular, key physical information such as the basic law of monotonically decreasing heat flux density along the flow direction and the reasonable boundary of local secondary flow disturbance cannot be constrained by the control logic online. Frontline engineers often point out that such intelligent control appears advanced but lacks physical robustness.
[0003] Meanwhile, existing methods generally employ static control architectures based on ideal component performance, failing to consider both the performance degradation of the defogging actuator throughout its entire lifecycle and the cross-temporal information of disease evolution during the waiting period. Traditional control logic is constructed only during the design phase based on the rated performance parameters of the actuator, either focusing on the precise issuance of commands while ignoring the dynamic response delay and steady-state accuracy deviation of the actuator, or relying on fault-tolerant design to passively respond to component failures. In particular, the gradual decay of dynamic response capability, the slow drift of valve position accuracy, and the cumulative damage from structural fatigue are intertwined during long-term service. Existing control strategies based on fixed margins, while ensuring safety, sacrifice a significant amount of availability performance, missing the regulatory boundaries that reflect the true achievable capability of component health status. Although some systems have introduced condition monitoring, the lack of comprehensive quantification and online closed-loop application of multi-dimensional performance degradation (dynamic, static, and lifespan) means that monitoring results remain isolated snapshots of health status, failing to establish an adaptive correlation between performance degradation and control strategies. This makes it difficult to compensate for the decline in control effectiveness caused by the gradual degradation of component performance in a timely manner.
[0004] In view of this, the present invention proposes an intelligent control method for the fog-removing efficiency of aero-engines based on multi-source data fusion to solve the above problems. Summary of the Invention
[0005] To overcome the aforementioned deficiencies of the prior art and to achieve the above objectives, the present invention provides the following technical solution: A method for intelligent control of defogging efficiency in aero-engines based on multi-source data fusion includes: Acquire multi-source sensing data of the aero-engine and the current health status parameters of each defogging actuator; the multi-source sensing data includes engine exhaust plume temperature field data, near-field atmospheric humidity and visibility data, combustion chamber heat transfer field distribution data, and intake duct fogging image sequence data, etc.; the health status parameters include response delay, valve position deviation, and cumulative fatigue damage value. Based on the heat transfer field distribution data of the combustion chamber, the physical monotonicity constraint set of the heat transfer field is extracted, and the health degradation index of each demisting actuator is calculated based on the response delay, valve position deviation and cumulative fatigue damage value. Based on the physical monotonicity constraint set of the heat exchange field and the health degradation index, the kernel interval and support interval of the current fog-clearing control fuzzy membership function are physically compatible. Membership function components that violate monotonicity are identified, and the support boundaries of the corresponding membership function components are adaptively contracted and reorganized to output the reorganized fuzzy membership function. Based on the renormalized fuzzy membership degree and the physical monotonicity constraint set of the heat exchange field, the modal data with fusion weight drift in the multi-source sensing data are weighted and a stable fusion weight set is output. Based on the stable fusion weight set, the membership vectors of each modality data are weighted and fused to generate a fog-removing efficiency control decision vector. The fog-removing efficiency control decision vector is then mapped to the target control quantity of each fog-removing execution unit in combination with the deviation of the current operating condition. A fog-removing execution instruction set is then output to drive each fog-removing execution unit to complete the corresponding fog-removing efficiency control operation.
[0006] Furthermore, the set of physical monotonicity constraints for the heat exchange field is used to characterize the physical law of monotonically decreasing heat flux density along the flow direction in the heat exchange field and its permissible boundary constraints.
[0007] Furthermore, the process of acquiring multi-source sensing data includes: Based on the collected engine exhaust plume temperature field data, near-field atmospheric humidity and visibility data, combustion chamber heat transfer field distribution data, and intake duct fogging image sequence data, the credibility of the sensor source to which it belongs is evaluated, and the credibility weight of each acquisition channel is generated. The collected data from each acquisition channel are timestamped and the sampling rate is normalized. In the acquisition channels where the confidence weight is lower than the preset confidence threshold, the missing data of the corresponding acquisition data is repaired by Kriging interpolation based on adjacent trusted channels. Outlier values were removed from the data collected from each acquisition channel after repair, and the collected data were spatially registered and fused in a unified engine coordinate system to obtain complete multi-source sensing data.
[0008] Furthermore, the extraction process of the physical monotonicity constraint set of the heat transfer field includes: The heat transfer field distribution data of the combustion chamber is discretized into several heat transfer sections according to the flow direction, and the heat flux density gradient sequence between adjacent heat transfer sections is calculated. The heat flux density gradient sequence is subjected to a directional consistency test. The section interval corresponding to the heat transfer section where the heat flux density gradient direction reverses is marked as a monotonicity doubt interval. The correlation between the local Reynolds number and Nusselt number of the corresponding heat transfer section is combined to determine whether the monotonicity doubt interval belongs to the physically permissible range of secondary flow disturbance. If it does not fall within the physically permissible range of secondary flow disturbance, extract the upper and lower tolerance boundaries of the heat flux density of the cross-sectional area, construct segmented monotonicity constraints, summarize the segmented monotonicity constraints of the entire flow direction, and form a set of physical monotonicity constraints for the heat transfer field.
[0009] Furthermore, the calculation process for the Health Deterioration Index includes: The dynamic response degradation component, static accuracy degradation component, and structural life degradation component of the defogging actuator are constructed using response delay, valve position deviation, and cumulative fatigue damage value, respectively. The dynamic response degradation component, static accuracy degradation component, and structural life degradation component are normalized respectively, and then weighted and summed according to the influence weight of each component in the fogging control. Apply a fogging execution unit type correction coefficient to the summation result, and output the final fogging execution unit health degradation index for each fogging execution unit.
[0010] Furthermore, the process of performing physical compatibility verification includes: Map the kernel interval endpoints of the membership function components within each fog-eliminating control fuzzy membership function to the heat exchange field's membership coordinate system, and compare them with the tolerance boundary constraints of the corresponding cross section in the heat exchange field's physical monotonicity constraint set. If the endpoints of the kernel interval are outside the tolerance boundary constraints, then the corresponding membership function component is determined to have a monotonicity violation. If the endpoints of the core interval are within the tolerance boundary constraints, the support intervals of the corresponding membership function components are extended for compatibility verification. This verifies whether the support intervals of the corresponding components overlap with the support intervals of adjacent components across the physical constraint boundaries. For membership function components that overlap, they are marked as membership function components that violate monotonicity.
[0011] Furthermore, the output process of the renormalized fuzzy membership function includes: Based on the health degradation index of the execution unit, the reachable control margin of the defogging execution unit associated with the membership function component is determined. The reachable control margin is used to characterize the maximum and minimum control output range that the execution unit can reliably achieve in the current health state. The safe shrinkage amount of the support boundary is determined by jointly using the reachable control margin and the physical monotonicity constraint set of the heat transfer field. The safe shrinkage amount is then calculated by shrinking both ends of the support interval toward the core interval for the membership function components that have monotonicity violations. The slope of the contracted support boundary is smoothed to generate a renormalized fuzzy membership function.
[0012] Furthermore, the process of outputting a stable fusion weight set includes: Based on multi-source sensing data, the renormalized fuzzy membership function is used to fuzzify the data of each modality and generate membership vectors for each modality. The stability of the fusion weights of each mode is monitored in real time by a fusion weight drift detection mechanism. When a fusion weight drift is detected, a weight correction compensation is triggered with the heat exchange field physical monotonicity constraint set as the anchor point, and a stable fusion weight set is output.
[0013] Furthermore, the process of generating the defogging instruction set includes: The membership vectors of each modality are linearly weighted and summed using the weight values corresponding to each modality data in the stable fusion weight set to obtain the fog-eliminating efficiency control decision vector; Based on the design operating condition benchmark parameters of multi-source sensing data and storage, multi-dimensional Euclidean distance is calculated to obtain the current operating condition deviation. The fog-eliminating efficiency control decision vector is defuzzified to obtain the baseline control quantity, which is then interpolated and superimposed with historical compensation records to obtain the target control quantity of the fog-eliminating execution unit for adaptive operation.
[0014] Furthermore, the process of triggering weighted normalization compensation anchored to the physical monotonicity constraint set of the heat transfer field includes: Extract the heat flux density distribution feature vector corresponding to each segment interval in the physical monotonicity constraint set of the heat exchange field, and use it as the physical anchor point for weight positiveing; Calculate the cosine similarity between the membership vector of the modal data that has experienced fusion weight drift at the current time and the physical anchor point, and recalculate the corrected fusion weight of the modal data using the monotonically increasing function of the cosine similarity. The corrected fusion weights are globally normalized, and the compensation range of the weight correction operation is limited to within a preset compensation range threshold, and the stable fusion weight set is output.
[0015] The technical effects and advantages of this invention, which is based on multi-source data fusion for intelligent control of aero-engine fog-removing efficiency, are as follows: This invention deeply integrates the physical laws of the combustion chamber heat exchange field with multi-source heterogeneous sensing data, ensuring that information from exhaust thermal characteristics, environmental meteorological characteristics, combustion heat exchange characteristics, and intake duct fogging morphology characteristics can truly contribute to the accurate characterization of the physical state of the engine's fogging target. This significantly improves the physical conformity of fogging control decisions and the reliability of command execution, enhancing the engine's safety and adaptability across the entire flight envelope. Furthermore, in practical applications, even when facing complex off-design conditions and scenarios of performance degradation of actuators (such as physical constraint boundary drift, component response sluggishness, and decreased control accuracy), it can still maintain the integrity of physical constraints and health status from the initial design state to the full life cycle service state. It no longer needs to rely entirely on simplified control models or fixed safety margins, making fogging performance control decisions more intelligent and precise. In particular, the online renormalization of membership functions based on physical monotonicity constraints and the health index modeling that integrates multi-dimensional degradation information break through the limitations of the traditional static control architecture based on ideal component performance. This eliminates safety hazards such as control decisions violating thermodynamic laws and unexpected decline in control efficiency due to component performance degradation, providing reliable technical support for intelligent closed-loop control and predictive maintenance of aero-engines. Attached Figure Description
[0016] Figure 1 This is a schematic diagram of the intelligent control method for defogging performance of aero-engines based on multi-source data fusion according to the present invention. Figure 2 This is a schematic diagram of the extraction process of the physical monotonicity constraint set of the heat exchange field in this invention; Figure 3 This is a flowchart illustrating the weight correction compensation process in this invention. Detailed Implementation
[0017] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.
[0018] Example 1 Please see Figure 1 As shown in this embodiment, the intelligent control method for defogging efficiency of aero-engines based on multi-source data fusion includes: Step S1: Acquire multi-source sensing data of the aircraft engine and the current health status parameters of each fog dissipation execution unit.
[0019] It should be noted that multi-source sensing data is used to characterize the current physical state of the engine defogging object from different dimensions, such as exhaust thermal characteristics, environmental meteorological characteristics, combustion heat transfer characteristics, and fog morphology characteristics. This data is collected based on a pre-deployed sensor array. Multi-source sensing data includes engine exhaust plume temperature field data, near-field atmospheric humidity and visibility data, combustion chamber heat transfer field distribution data, and intake manifold fog image sequence data. The defogging execution unit refers to the physical component used to perform defogging efficiency control operations, and may include intake manifold electrically heated de-icing components, air curtain blow-off nozzles, and bypass bleed air anti-icing components. Valves, ultrasonic demisting devices, etc.; health status parameters are used to characterize the physical performance degradation of each demisting actuator in the current operating stage, including response delay, valve position deviation, and cumulative fatigue damage value. Response delay refers to the deviation between the time taken for the demisting actuator to actually perform its action after receiving the control command and the design reference response time. Valve position deviation refers to the steady-state deviation between the actual valve position (or nozzle opening, heating power level) of the demisting actuator and the target control quantity. Cumulative fatigue damage value refers to the cumulative structural fatigue damage obtained based on the statistical analysis of the historical number of actions and action amplitude of the demisting actuator.
[0020] In embodiments of the present invention, the process of acquiring multi-source sensing data includes: Based on the collected engine exhaust plume temperature field data, near-field atmospheric humidity and visibility data, combustion chamber heat transfer field distribution data, and intake duct fogging image sequence data, the reliability of their respective sensor sources is assessed, generating a reliability weight for each acquisition channel. Sensor source reliability assessment is a preprocessing step that quantitatively evaluates the data quality of each sensing channel. Exhaust plume temperature field data is acquired by a thermocouple array positioned at the engine exhaust nozzle cross-section; its reliability is assessed based on the consistency deviation of each thermocouple measurement and historical fault records. Near-field atmospheric humidity and visibility data are acquired by a humidity sensor and optical visibility meter located on the outside of the fuselage; their reliability is assessed based on the noise level of the sensor output signal and the drift within the calibration period. Combustion chamber heat transfer field distribution data is acquired by a heat flux meter array embedded in the combustion chamber wall; its reliability is assessed based on the physical reasonableness range and signal stability of each heat flux meter signal. Intake duct fogging image sequence data is acquired by a miniature industrial camera installed on the inner wall of the intake duct lip; its reliability is assessed based on image sharpness scoring and inter-frame continuity indicators. The confidence weights of each sensing channel are normalized to the [0, 1] interval, serving as prior information for quality in subsequent data fusion processing.
[0021] The collected data from each acquisition channel are timestamped and the sampling rate is normalized. For acquisition channels where the confidence weight is lower than the preset confidence threshold, the corresponding acquired data is repaired by Kriging interpolation based on adjacent trusted channels. Timestamp alignment unifies the data from each sensing channel to a synchronized timeline based on the flight computer clock, eliminating phase differences between the sampling trigger times of different sensors. Sampling rate normalization unifies the data from each channel to the system's standard sampling frequency (e.g., 100Hz). For channels with lower sampling rates, cubic spline interpolation is used for upsampling, while for channels with higher sampling rates, low-pass filtering is used for downsampling. When the confidence weight of a certain acquisition channel is lower than a preset confidence threshold (e.g., 0.6), the data acquired by that channel is determined to have serious quality problems and is not directly used for subsequent processing. Instead, the data from adjacent reliable channels is used as the source, and the missing channel's data is reconstructed in physical space using Kriging interpolation. Kriging interpolation uses the known spatial correlation structure between each sensing quantity (e.g., the spatial continuity of the temperature field and the gradual variation characteristics of humidity distribution) to establish a variogram model, generating unbiased optimal estimates at the missing locations. The repaired data is labeled with the estimated variance for reference in subsequent processing.
[0022] Outlier removal was performed on the data collected from each acquisition channel after repair. The removed data were then spatially registered and fused in a unified engine coordinate system to obtain complete multi-source sensing data. Outlier removal combined hard threshold removal based on physical upper and lower limits with statistical removal based on the 3σ criterion. Thermocouple readings exceeding the engine operating temperature range, humidity and visibility values exceeding the physical possibility range, and invalid frames in the image caused by pollution or reflection were all removed. The gaps after removal were filled by linear interpolation of the effective values before and after. Spatial registration and fusion transformed the data from each sensing channel from the local coordinate system of their respective sensors to a unified engine axis coordinate system. A three-dimensional registration coordinate system was established with the center of the engine intake section as the origin and the axis direction as the z-axis. After rotation and translation transformation, the data from each channel were voxelized and aligned in the unified coordinate system to form a complete multi-source sensing data field covering the entire process from the intake to the exhaust nozzle.
[0023] Step S2: Extract the physical monotonicity constraint set of the heat transfer field based on the heat transfer field distribution data of the combustion chamber, and calculate the health degradation index of each demisting actuator based on the response delay, valve position deviation and cumulative fatigue damage value.
[0024] It should be noted that the physical monotonicity constraint set of the heat exchange field is used to characterize the physical law of monotonically decreasing heat flux density along the flow direction in the heat exchange field and its allowable boundary constraints. This physical law is determined based on the fundamental law of heat transfer, which is that the mainstream gas in the combustion chamber gradually exchanges heat with the wall along the flow direction, and energy dissipation leads to an overall decreasing trend in heat flux density. The allowable deviation boundary is used to accommodate non-violation fluctuations such as local secondary flow disturbances. The health degradation index of the defogging actuator is used to comprehensively characterize the performance degradation degree of each defogging actuator in terms of dynamic response, static accuracy, and structural life. The larger the degradation index value, the weaker the current reliable output capability of the defogging actuator. The cross-sectional interval refers to the axial segment composed of a series of continuous heat exchange cross sections.
[0025] Please see Figure 2 As shown, the process of extracting the physical monotonicity constraint set of the heat transfer field includes: The heat transfer field distribution data of the combustion chamber is discretized into several heat transfer sections along the flow direction, and the heat flux density gradient sequence between adjacent heat transfer sections is calculated. The heat transfer field discretization process involves taking a heat transfer section at fixed axial intervals (e.g., every 20 mm) along the engine axis. The heat flux density value on each heat transfer section is the weighted average of all heat flux meter readings within the section range. The weights are determined based on the radial distance of each heat flux meter from the center of the heat transfer section; heat flux meters closer to the center have higher weights, reflecting the dominant role of the core flow region on the average heat transfer characteristics of the heat transfer section. The heat flux density gradient sequence between adjacent heat transfer sections is calculated by dividing the difference in heat flux density between adjacent heat transfer sections by the distance between the sections, forming a spatial gradient sequence of heat flux density distributed along the flow direction. This spatial gradient sequence reflects the changing trend of heat transfer intensity along the flow direction of the combustion chamber. Under normal operating conditions, the heat flux density of the aero-engine combustion chamber gradually converts into gas kinetic energy along the flow direction after the fuel releases heat during combustion in the main combustion zone. Overall, the heat flux density exhibits a monotonically decreasing physical law along the flow direction. This law constitutes the basic basis for the physical monotonicity constraint of the heat transfer field.
[0026] A directional consistency check is performed on the heat flux density gradient sequence. Sections corresponding to heat transfer sections where the heat flux density gradient direction reverses are marked as potentially monotonic intervals. The correlation between the local Reynolds number and Nusselt number of the corresponding heat transfer section is then used to determine whether these potentially monotonic intervals fall within the physically permissible range of secondary flow disturbances. The directional consistency check compares the signs of adjacent gradient values in the heat flux density gradient sequence one by one. If a gradient changes from negative to positive (i.e., a local increase in heat flux density) at a certain section, this section is marked as a potentially monotonic interval, and its axial position range is recorded. For each suspicious interval, the local Reynolds number and Nusselt number at the corresponding heat exchange section are extracted. The deviation ratio between the measured Nusselt number and the value predicted by the Dittus-Belth formula based on the local Reynolds number is calculated. When the deviation ratio is within the known reasonable range of secondary flow disturbance in the combustion chamber (e.g., the deviation ratio is less than 15%), the local increase in heat flux density in the suspicious interval is determined to be a physically permissible secondary flow disturbance caused by secondary flow induced by the cooling hole jets or swirlers on the combustion chamber wall, and does not need to be treated as a monotonicity violation. When the deviation ratio exceeds the reasonable range, the suspicious interval is determined to have a non-physically permissible monotonicity reversal interval.
[0027] If it does not fall within the physically permissible range of secondary flow disturbance, extract the upper and lower tolerance boundaries of the heat flux density of the cross-sectional area, construct segmented monotonicity constraints, summarize the segmented monotonicity constraints of the entire flow direction, and form a set of physical monotonicity constraints for the heat transfer field. For each non-physically permissible monotonic reversal interval, the heat flux density value of the adjacent normal monotonic section upstream of the corresponding interval is taken as the physical permissible upper limit of the heat flux density of that interval, and the heat flux density design lower limit value of the corresponding axial position in the engine design parameters is taken as the physical permissible lower limit. Thus, the upper and lower tolerance boundaries of the heat flux density of that interval are determined. The width of the tolerance boundary takes into account both the sensor measurement error (usually taken as 3 times the standard deviation) and the natural fluctuation range of heat flux density caused by the fluctuation of combustion chamber operating conditions (determined based on historical measurement data statistics) to ensure that the tolerance boundary has sufficient coverage reliability in engineering practice. The decreasing constraint conditions of each normal monotonic interval in the entire flow direction and the tolerance boundary constraint conditions of each reversal interval are summarized in axial position order to form a complete set of physical monotonic constraints for the heat transfer field. The physical monotonic constraints of the heat transfer field are organized by segmented intervals. Each interval records its axial start and end positions, constraint type (decreasing constraint and tolerance boundary constraint) and corresponding heat flux density constraint parameters for subsequent online constraint resetting module to call.
[0028] In an embodiment of the present invention, the calculation process of the health degradation index includes: The dynamic response degradation component, static accuracy degradation component, and structural life degradation component of the defogging actuator are constructed using response delay, valve position deviation, and cumulative fatigue damage value, respectively. The dynamic response degradation component reflects the degree of decline in the speed at which each defogging actuator tracks command signals. The response delay is defined as the difference between the actual response time required for the defogging actuator to reach 90% of the target value after receiving a control command and the factory-specified response time; a larger difference indicates more severe dynamic performance degradation. The static accuracy degradation component reflects the degree of decline in the accuracy of the defogging actuator in maintaining the target valve position under steady-state conditions. The valve position deviation is defined as the root mean square value of the difference between the actual valve position and the commanded valve position under stable control conditions; a larger deviation indicates more severe static accuracy degradation. The structural life degradation component reflects the remaining reliable life level of the mechanical structure of the defogging actuator. The cumulative fatigue damage value is calculated based on Miner's linear damage accumulation rule, according to the stroke amplitude and frequency of each action of the defogging actuator. The closer the cumulative fatigue damage value is to the critical value of 1, the less structural life margin is available.
[0029] The dynamic response degradation component, static accuracy degradation component, and structural life degradation component are normalized respectively, and then weighted and summed according to the influence weight of each component in the fog control. Normalization maps each degradation component to the interval [0, 1]. The maximum allowable degradation of the corresponding parameter of each component is used as the denominator for normalization. The response delay is calculated using the maximum allowable delay time specified in the design specifications as the denominator. The valve position deviation is calculated using the maximum allowable deviation specified in the control accuracy requirements as the denominator. The cumulative fatigue damage value is calculated using the damage threshold value of 1 as the denominator. Each degradation component after normalization is a dimensionless value between 0 and 1. The larger the value, the more severe the degradation of that dimension. The determination of the influence weights takes into account the relative importance of each degradation dimension to the fog control effect: In the fog control scenario, the fog control unit's ability to quickly track commands is crucial for coping with dynamic fog field changes, and the influence weight of the dynamic response degradation component is relatively the highest. The static accuracy degradation component affects the steady-state control accuracy of the fog control unit, and its influence weight is in the middle. The structural life degradation component mainly reflects reliability risk and has the lowest influence weight. The sum of the three influence weights is normalized to 1.
[0030] Apply a fogging execution unit type correction coefficient to the summation result, and output the final fogging execution unit health degradation index for each fogging execution unit. Defogging actuators are categorized into several types based on their function and location within the engine defogging system. These primarily include: anti-icing bleed air regulating valves installed at the leading edge of the engine intake manifold; bypass valves controlling the hot air circulation flow; distribution valves regulating the distribution of hot air along the inner wall of the intake manifold; and mixing valves controlling the amount of cold air mixed in the exhaust nozzle area. The sensitivity of each type of defogging actuator to fog field disturbances varies significantly. The anti-icing bleed air regulating valve, located at the leading edge of the intake manifold, directly affects the hot air supply to the fogging area and has the highest sensitivity to fog field disturbances. Its correction coefficient is relatively the largest, meaning its value is positively correlated with the physical location of the defogging actuator in the defogging channel and its sensitivity level to fog field disturbances. A higher sensitivity level means a more significant impact of performance degradation on overall defogging efficiency. The distribution valve and bypass valve have moderate impacts on defogging efficiency, with correction coefficients taking a middle value. The mixing valve primarily affects the plume temperature in the exhaust nozzle area and has the lowest sensitivity to direct disturbances in the near-field fog field, resulting in a relatively small correction coefficient. The formula for calculating the health degradation index of the defogging actuator is:
[0031] ;in, For the first Health degradation index of each fog dispersing unit For the first Type correction coefficient for each defogging execution unit , , These are the influence weights of the dynamic response degradation component, the static accuracy degradation component, and the structural life degradation component, respectively. , , The first The normalized values of the three degradation components of each defogging actuator. The closer the health degradation index is to 1, the more severe the overall degradation of the defogging actuator and the smaller the control margin that can be achieved.
[0032] As can be seen from the above, the embodiments of this application extract piecewise monotonic physical constraints from the heat transfer field data, transforming the basic laws of heat transfer into explicit constraints that can be used to verify the rationality of fuzzy membership functions. At the same time, by combining the correlation between Reynolds number and Nusselt number, misjudgments of normal secondary flow disturbances are eliminated, avoiding overly conservative constraint set extraction. In addition, the embodiments of this application integrate the degradation information of dynamic response, static accuracy, and structural life of the defogging execution unit, and introduce an execution unit type correction coefficient, realizing a refined quantification of the current true health status of each defogging execution unit, providing a reliable execution unit-side constraint basis for the subsequent adaptive contraction and reorganization of membership functions.
[0033] Step S3: Based on the physical monotonicity constraint set of the heat exchange field and the health degradation index, perform physical compatibility verification on the kernel interval and support interval of the current fog-clearing control fuzzy membership function, identify the membership function components that violate monotonicity, and implement adaptive shrinkage renormalization on the support boundary of the corresponding membership function components, and output the renormalized fuzzy membership function.
[0034] It should be noted that the fuzzy membership function for fog dissipation control refers to the membership distribution function used to fuzzify the sensing data of each modality to characterize the membership of each modality data to different fog dissipation conditions (e.g., light fog, moderate fog, heavy fog). The kernel interval is the interval where the membership function is always 1, which characterizes the deterministic range in which the modality data completely belongs to the corresponding condition level. The support interval is the interval where the membership function is greater than 0, which characterizes the boundary range in which the modality data may belong to the corresponding condition level. The physical compatibility verification process is used to physically verify and correct the rationality of the membership function in real time during the defogging control operation, so as to avoid the membership function from contradicting the actual heat transfer physical law of the combustion chamber and the actual achievable capability of the execution unit due to initial calibration deviation or long-term operation drift. The fuzzy membership function of defogging control is the core reasoning basis of the fuzzy controller of defogging efficiency. It is defined on the domain of the defogging control quantity. Each input variable (such as hot air flow ratio, intake duct wall temperature deviation, near field visibility, etc.) corresponds to a set of fuzzy linguistic values (such as "too low", "moderate", "too high", etc.). Each fuzzy linguistic value corresponds to a membership function curve, that is, the membership function component, which is typically in the form of trapezoidal or triangular membership function.
[0035] In embodiments of the present invention, the process of performing physical compatibility verification includes: Map the endpoints of the kernel intervals of the membership function components within each fog-eliminating control fuzzy membership function to the coordinate system of the heat exchange field, and compare them with the tolerance boundary constraints of the corresponding sections in the physical monotonicity constraint set of the heat exchange field. The mapping process is based on the feature quantity mapping relationship pre-calibrated during the system design phase: the fog-clearing and fuzzy inference system uses the normalized feature quantity obtained after preprocessing multi-source sensing data as the universe of discourse, and the endpoints of the core intervals of each membership function component are represented by the feature quantity coordinates within the universe of discourse; the mapping relationship converts the feature quantity coordinates within the universe of discourse into the heat flux density values of the corresponding cross-sectional intervals in the heat transfer field coordinate system. The conversion function is determined by least squares fitting of historical steady-state operating data and is calculated in real time during system operation using table lookup interpolation; after the mapping is completed, the mapped heat flux density values of each core interval endpoint are compared one by one with the tolerance boundary constraints of the corresponding cross-sectional intervals in the heat transfer field physical monotonicity constraint set. The tolerance boundary constraints are jointly constituted by the allowable upper boundary and allowable lower boundary of heat flux density extracted during the constraint set construction phase; for trapezoidal membership function components with left and right endpoints in the core interval, the left and right endpoints are mapped and compared independently to ensure that both sides of the core interval are included in the physical compatibility verification range.
[0036] If the endpoint of a core interval is outside the tolerance boundary constraint, the corresponding membership function component is determined to have a monotonicity violation. The determination rule is as follows: when the mapped heat flux density value of a core interval endpoint is higher than the allowable upper boundary of the heat flux density tolerance boundary constraint of the corresponding cross-sectional interval, or lower than the allowable lower boundary, the core interval endpoint is determined to be outside the tolerance boundary constraint, and the corresponding membership function component has a monotonicity violation; for trapezoidal membership function components, if either the left or right endpoint is outside the tolerance boundary constraint, the monotonicity violation flag of the component is triggered; for membership function components determined to have a monotonicity violation, its violation magnitude is recorded: calculate the mapped heat flux density value of the core interval endpoint and the nearest tolerance boundary (allowable... The absolute difference between the upper boundary (or the allowed lower boundary) and the normalized ratio of this difference to the width of the corresponding cross-sectional heat flux density tolerance interval is calculated as the normalized violation magnitude. The larger the normalized violation magnitude, the more serious the deviation of the membership function component from the physical monotonicity constraint of the heat transfer field, and the greater the safe shrinkage amount required for subsequent adaptive shrinkage reforming. All membership function components that are determined to have monotonicity violations are summarized to form a preliminary violation component set, and a core interval endpoint over-boundary violation flag is recorded for each violation component for use in the subsequent shrinkage reforming stage.
[0037] If the endpoints of the core interval are within the tolerance boundary constraints, then the support intervals of the corresponding membership function components are subjected to extended compatibility verification to check whether their support intervals overlap with the support intervals of adjacent components across the physical constraint boundary. For membership function components that overlap, they are marked as membership function components with monotonicity violations. The extended compatibility verification is performed sequentially on all membership function components that are determined to be compatible in the core interval endpoint comparison: First, the left and right endpoints of the support interval of the target component are projected onto the heat transfer field coordinate system through the same mapping relationship to obtain the cross-sectional coverage of the support interval in the heat transfer field coordinate system; at the same time, the support intervals of the membership function components adjacent to the target component in the universe of discourse are extracted and similarly projected onto the heat transfer field coordinate system to obtain the cross-sectional coverage of the support intervals of the adjacent components; the overlapping cross-sectional intervals of the target component support intervals and the adjacent component support intervals in the heat transfer field coordinate system are calculated, that is, the intersection of their cross-sectional coverages; the physical constraint boundary is checked on the overlapping cross-sectional intervals, and based on the cross-sectional number boundaries of each segment constraint interval in the heat transfer field physical monotonicity constraint set, it is determined whether the overlapping cross-sectional interval crosses the heat transfer field. Physical monotonicity constraints are marked as non-physically reversible segmented constraint interval boundaries: If the overlapping section interval falls entirely within a single segmented constraint interval, or if all the segmented constraint intervals it crosses belong to the physically permissible secondary flow disturbance range, then the extended compatibility of the target component is satisfied, and the monotonicity violation flag is not triggered; if the overlapping section interval crosses at least one segmented constraint interval boundary marked as non-physically reversible, then the overlapping range between the support interval of the target component and the support interval of the adjacent component is determined to include the non-physically reversible region into the transitional membership segment of fuzzy inference. There is a risk of assigning non-zero membership degrees to heat transfer field states that violate physical monotonicity laws during fuzzy inference. This component is marked as a membership function component with monotonicity violation, and the cross-domain violation flag of the support interval is recorded. The membership function components marked with monotonicity violation in the extended compatibility verification of the support interval are merged with the initial violation component set, and after removing duplicates, a complete violation component set is formed, and the complete monotonicity violation detection result is output.
[0038] In an embodiment of the present invention, the output process of the renormalized fuzzy membership function includes: Based on the health degradation index of the actuator, the reachable control margin of the defogging actuator associated with the membership function component is determined. The reachable control margin characterizes the maximum and minimum control output range that the actuator can reliably achieve under the current health state. The reachable control margin is determined based on the rated control output range of each defogging actuator, and the benchmark range is dynamically compressed in conjunction with the health degradation index of the defogging actuator: For the flow regulating valve on the hot gas ejector pipeline, its rated control output range is calculated with the rated maximum flow and rated minimum flow as the upper and lower endpoints, and the double-end synchronous compression of the rated control output range is calculated using the actuator health degradation index. The upper limit of the compressed reachable control output is the rated maximum flow multiplied by... The lower limit of the achievable control output is the rated minimum flow rate plus the rated control output range width multiplied by [the specified value]. ,in and The upper limit compression coefficient and lower limit lift coefficient are determined by engineering calibration tests for each type of actuator. For proportional control valves at mixing nodes, the maximum and minimum allowable mixing ratios are used as the endpoints of the rated control output range. The compressible control margin is calculated based on the actuator health degradation index in the same way. For on / off valves in auxiliary heating circuits, their achievable control output is binary. When the actuator health degradation index exceeds the preset on / off valve degradation tolerance threshold, the achievable control margin of the on / off valve is downgraded to retain only the current locked state, and no switching operation is performed. The achievable control margin is expressed in range form. The higher the actuator health degradation index, the narrower the achievable control margin range, reflecting the compression effect of actuator health degradation on its actual regulation capability.
[0039] The safe shrinkage amount of the support boundary is determined by jointly using the reachable control margin and the physical monotonicity constraint set of the heat transfer field. For the support intervals with membership function components that violate monotonicity, the safe shrinkage amount is shrunk towards the core interval from both ends. The determination of the safe shrinkage amount is carried out through two independent calculations, with the larger value taken: The first calculation determines the minimum shrinkage amount required by the physical constraints based on the physical monotonicity constraint set of the heat transfer field. For out-of-bounds violation components at the core interval endpoints, the minimum shrinkage amount is the corresponding value after inverse mapping and conversion back to the domain coordinates of the absolute difference between the mapped heat flux density value at the core interval endpoints and the tolerance boundary. For cross-domain violation components in the support intervals, the minimum shrinkage amount is the distance between the support interval endpoint and the nearest non-physically permissible reversal interval boundary in the domain coordinates. The second calculation determines the shrinkage amount required by the reachability constraints of the execution unit based on the reachability control margin. The reachability control margin interval is projected back to the domain coordinates through the inverse transformation of the characteristic quantity mapping relationship to obtain the permissible interval corresponding to the reachability of the execution unit in the domain. The distance between the current support interval endpoint of the violation component and the boundary of the permissible reachability interval of the execution unit is calculated, and the portion exceeding the permissible reachability interval is taken as the shrinkage amount required by the reachability constraints of the execution unit. The safe shrinkage amount is the larger value between the minimum shrinkage amount of the physical constraints and the shrinkage amount of the reachability constraints of the execution unit. For the support intervals of membership function components that violate monotonicity, the safe contraction amount is shrunk towards the core interval from both ends. If the above protective constraints are not satisfied after the contraction is performed according to the safe contraction amount, the safe contraction amount is rolled back to the critical contraction amount that makes the protective constraints just satisfied. At the same time, a constraint conflict alarm is generated and recorded in the control log for offline analysis.
[0040] The slope of the contracted support boundary is smoothed to generate a renormalized fuzzy membership function. After the contraction operation is completed, the slope of the original membership function at the endpoint of the support interval that violates the component changes from the value before contraction to the slope at the new endpoint. If the contracted endpoint is directly truncated, a non-differentiable inflection point of the membership function will be generated at the new support endpoint, affecting the smoothness of subsequent fuzzy inference. The slope smoothing process adopts the cubic spline interpolation method. In the transition section between the left endpoint of the contracted support interval and the left endpoint of the core interval, the original linear membership slope is replaced by a cubic polynomial. The boundary conditions of the cubic polynomial are set as follows: the membership value is 0 and the first derivative is 0 at the left endpoint of the support interval, and the membership value is 1 and the first derivative is continuous with the slope at the original core interval boundary at the left endpoint of the core interval. Symmetrical cubic spline boundary conditions are applied to the transition section between the right endpoint of the contracted support interval and the right endpoint of the core interval. The cubic spline coefficients are determined by solving the above four boundary conditions simultaneously, ensuring that the first derivative of the smoothed membership function is continuous at the right endpoint of the support interval and the right endpoint of the core interval, satisfying the requirement of continuous differentiability. After performing support boundary shrinkage and slope smoothing on all violation components, the renormalized violation components are merged with the compatible components that have not triggered renormalization to form a complete renormalized fuzzy membership function, which is then output to the subsequent multimodal fuzzification processing stage.
[0041] Step S4: Based on the renormalized fuzzy membership degree and the physical monotonicity constraint set of the heat exchange field, perform weight compensation on the modal data with fusion weight drift in the multi-source sensing data, and output a stable fusion weight set.
[0042] It should be noted that modal data refers to the four types of modal data corresponding to the engine exhaust plume temperature field data, near-field atmospheric humidity and visibility data, combustion chamber heat transfer field distribution data, and intake duct fogging image sequence data mentioned above in the multi-source sensing data; fuzzification processing refers to substituting the measured data of each modality into the corresponding renormalized fuzzy membership function to obtain the membership degree value of the modal data to each fog dissipation condition level, and summing them to obtain the membership degree vector of the modality; fusion weight refers to the weight coefficient assigned to each modality when subsequently weighting and fusing the membership degree vectors of each modality, which is used to characterize the relative importance of the modal data in the current fog dissipation performance determination.
[0043] In embodiments of the present invention, the process of outputting a stable fusion weight set includes: Based on multi-source sensing data, a renormalized fuzzy membership function is used to fuzzify the modal data, generating membership vectors for each modality. The multi-source sensing data corresponds to four modalities: exhaust plume temperature field mode, near-field atmospheric humidity and visibility mode, combustion chamber heat transfer field mode, and intake duct fogging image mode. For each modal, the current sensed quantity of that modal is fuzzified and mapped to the renormalized fuzzy membership function. Taking the exhaust plume temperature field mode as an example, the current plume temperature field features (such as plume core temperature, plume boundary temperature gradient, and plume axial attenuation rate) are substituted into the corresponding renormalized fuzzy membership function to calculate the membership degree of each feature on each fuzzy linguistic value, and then arranged according to a fixed feature order. The membership vector for this mode is generated by substituting the current humidity and visibility values into the corresponding renormalized fuzzy membership function for the near-field atmospheric humidity and visibility modes. For the combustion chamber heat transfer field mode, the current heat transfer field features (such as the average heat flux density and heat flux density gradient of each segment interval) are substituted into the renormalized fuzzy membership function to generate the corresponding membership vector. For the inlet fogging image mode, fogging features extracted from the image sequence (such as fogging area ratio, fogging thickness estimate, and fogging growth rate) are substituted into the corresponding renormalized fuzzy membership function to generate the membership vector for this mode. The membership vectors of the four modes together constitute the basic input object for subsequent weighted fusion.
[0044] The stability of the fusion weights for each mode is monitored in real time through a fusion weight drift detection mechanism. When a fusion weight drift is detected, weight correction compensation is triggered using the physical monotonicity constraint set of the heat exchange field as the anchor point, and a stable fusion weight set is output. The fusion weight drift detection continuously collects the fusion weight sequence of each mode using a sliding time window (e.g., the window length is the most recent 20 sampling periods), and calculates the variance and trend slope of each fusion weight within the sliding time window. When the variance of a certain fusion weight exceeds a preset weight stability threshold, or the absolute value of its trend slope exceeds a preset drift rate threshold, it is determined that the fusion weight has drifted for that mode. Further classification analysis of drift is performed to distinguish between normal adjustment drift and abnormal bias drift: if the drift occurrence time closely coincides with the time when flight operating parameters (such as flight altitude, engine thrust setting, and atmospheric temperature) undergo significant changes, it is determined to be normal adjustment drift caused by changes in operating conditions; if the operating parameters do not change significantly when the drift occurs, but the renormalized fuzzy membership function of the corresponding mode has recently undergone a significant contraction and renormalization operation, it is determined to be abnormal bias drift caused by membership function mismatch. For normal adjustment drift, the corresponding operating parameters and weight adjustment records are stored in the off-design condition adaptive history library for reference in subsequent fusion weight identification under off-design conditions; for abnormal bias drift, weight correction compensation is triggered. Please refer to [link to relevant documentation]. Figure 3As shown, the weight correction compensation process includes: extracting the heat flux density distribution feature vector corresponding to each segment constraint interval in the heat exchange field physical monotonicity constraint set, which serves as the physical anchor point for weight correction. The extraction of the heat flux density distribution feature vector is based on the median of the heat flux density tolerance boundary of each segment constraint interval in the heat exchange field physical monotonicity constraint set. The median of the heat flux density tolerance boundary of each segment constraint interval is arranged sequentially according to the cross-section number, and combined to form the heat flux density distribution feature vector. This vector is constructed based on the physical monotonicity law of the combustion chamber heat exchange field along the flow direction, remains relatively stable under current operating conditions, and can objectively characterize the physical reference state of the combustion chamber heat exchange field, possessing the reliability to serve as the physical anchor point for weight correction. The physical anchor point is extracted and cached in the system memory during the initialization phase of the defogging control system, and is synchronously refreshed when the heat exchange field physical monotonicity constraint set is updated.
[0045] The cosine similarity between the membership vector of the modal data experiencing fusion weight drift and the characteristic vector of the heat flux density distribution at the physical anchor point is calculated at the current moment. The corrected fusion weights of the modal data are then recalculated using a monotonically increasing function of cosine similarity. The cosine similarity calculation uses the modal membership vector generated by the modal data experiencing abnormal bias drift within the current control cycle as the object of calculation, and the characteristic vector of the heat flux density distribution at the physical anchor point as the reference object. The cosine similarity is obtained by calculating the ratio of the inner product of the two vectors to the product of their respective moduli. The cosine similarity value ranges from -1 to +1. A higher cosine similarity indicates a stronger directional consistency between the current membership vector of the modal data and the physical reference state of the heat transfer field, and a higher corrected fusion weight should be assigned accordingly. Conversely, a lower cosine similarity indicates a greater directional deviation between the current membership vector of the modal data and the physical reference state of the heat transfer field, and a lower corrected fusion weight should be assigned accordingly. A nonlinear transformation is applied to the cosine similarity using a monotonically increasing function, mapping the cosine similarity to a modified fusion weight value. The form and parameters of the monotonically increasing function are determined based on historical data calibration during the system debugging phase, ensuring that the modified fusion weight increases monotonically with the increase of cosine similarity, and that the output range of the modified fusion weight is limited to a reasonable positive value range. For modal data that did not experience abnormal bias drift during the current control cycle, their fusion weights remain unchanged and directly participate in subsequent global normalization calculations.
[0046] The corrected fusion weights are globally normalized, and the compensation range of the weight correction operation is limited to within a preset compensation range threshold, outputting a stable fusion weight set. Global normalization sums the corrected fusion weights of the abnormal bias drift modal data and the current fusion weights of the non-drift modal data, calculating the sum of the fusion weights of the four modal data. The sum is then used as the denominator to divide each modal data fusion weight individually, ensuring that the sum of the fusion weights of the four modal data is one after normalization, satisfying the probability normalization constraint of the fusion weight set. Before performing global normalization, a compensation amplitude threshold is applied to the adjustment amount of the fusion weights of the abnormal bias drift modal data in a single weight normalization operation. This limits the single adjustment amount of the fusion weights of each drift modal data in this weight normalization operation to within a preset compensation amplitude threshold. When the calculated adjustment amount exceeds the preset compensation amplitude threshold, the adjustment amount is truncated to the preset compensation amplitude threshold before global normalization is performed, preventing excessive single normalization amplitude from causing chain oscillations in the fusion weights of the modal data. After global normalization and compensation amplitude limitation, a stable fusion weight set containing the corresponding fusion weights of the four sensing modal data is output for subsequent weighted fusion stages.
[0047] Step S5: Based on the stable fusion weight set, the membership vectors of each modality data are weighted and fused to generate a fog-removing efficiency control decision vector. The fog-removing efficiency control decision vector is then mapped to the target control quantity of each fog-removing execution unit in combination with the current operating condition deviation. A fog-removing execution instruction set is output to drive each fog-removing execution unit to complete the corresponding fog-removing efficiency control operation.
[0048] It should be noted that the defogging efficiency control decision vector refers to the comprehensive membership vector obtained by weighting and summing the membership vectors of each mode according to the stable fusion weight set. It is used to characterize the overall confidence level of the current engine defogging object belonging to each defogging operating condition level. The current operating condition deviation is used to characterize the degree of deviation between the engine's current actual operating condition and the design reference operating condition. The larger the operating condition deviation, the more the current operating condition deviates from the design reference, and the greater the compensation correction required when mapping the defogging control decision vector to the actuator target control quantity.
[0049] In an embodiment of the present invention, the process of generating the defogging execution instruction set includes: The membership vectors of each modal data are linearly weighted and summed using the weight values corresponding to the data in the stable fusion weight set to obtain the fog-removing efficiency control decision vector. The linear weighted summation uses the fusion weight values corresponding to the multi-source sensing data (including engine exhaust plume temperature field data, near-field atmospheric humidity and visibility data, combustion chamber heat transfer field distribution data, and intake duct fogging image sequence data) as weighting coefficients. These weighted sums are applied to the corresponding components of the membership vectors of the four modalities, with higher fusion weight values contributing more to the fog-removing efficiency control decision vector. Each component of the fused fog-removing efficiency control decision vector corresponds to the comprehensive membership degree of different fog-removing control linguistic variables, comprehensively reflecting the joint evaluation results of the four sensing modalities on the fog-removing state at the current moment.
[0050] The deviation of the current operating condition is obtained by multi-dimensional Euclidean distance calculation based on multi-source sensing data and stored design condition reference parameters. The multi-dimensional Euclidean distance calculation normalizes engine speed, inlet total pressure, atmospheric temperature, and ambient humidity to a dimensionless scale, and then calculates the Euclidean distance between the current parameter vector and the design condition reference parameter vector, serving as a quantitative representation of the current operating condition deviation. A larger current operating condition deviation indicates a more significant deviation between the current flight condition and the fog dissipation system design point, requiring a larger operating condition compensation magnitude for the fog dissipation control strategy. The closest historical compensation record is retrieved from the non-design condition compensation mapping table based on the current operating condition deviation. This non-design condition compensation mapping table is continuously accumulated from normal adjustment drift data during historical fog dissipation control processes, recording the corresponding control compensation amounts under different operating condition deviation conditions. The retrieval process uses the nearest neighbor method to find the closest historical records in the mapping table, and performs weighted interpolation with the reciprocal of the distance as the weight to obtain the historical compensation record corresponding to the current operating condition deviation.
[0051] The fog-eliminating efficiency control decision vector is defuzzified to obtain the baseline control quantity, which is then interpolated and superimposed with historical compensation records to obtain the target control quantity of the fog-eliminating execution unit for adaptive operation. The defuzzification operation uses the centroid method to defuzzify the fog-eliminating efficiency control decision vector, converting the comprehensive membership degree of each fog-eliminating control linguistic variable into the baseline control quantity corresponding to each fog-eliminating execution unit. The baseline control quantity is then interpolated and superimposed with the compensation control quantity in the historical compensation records. The compensation control quantity corrects the systematic deviation of the baseline control quantity under non-design conditions. The superimposed result is the target control quantity of the fog-eliminating execution unit that takes into account both the design point control logic and non-design condition compensation. The target control quantity of each defogging execution unit is saturated and truncated based on the achievable control margin of each unit. When the target control quantity of a certain defogging execution unit exceeds its upper limit of achievable control margin, it is truncated to the upper limit value; when it is below the lower limit of achievable control margin, it is truncated to the lower limit value. The saturation truncation is performed independently at the level of each defogging execution unit, ensuring that the instructions received by each unit are within the range that can be reliably executed under its current health condition. The truncated target control quantities of each defogging execution unit are packaged into instructions, and a defogging execution instruction set is output to drive each defogging execution unit, such as the hot gas ejector flow regulating valve, the mixing proportion regulating valve, and the auxiliary heating on / off valve, to complete the corresponding defogging efficiency regulation operation.
[0052] In embodiments of the present invention, the method further includes: A fog dissipation performance control status monitoring module is set up on the airborne computing platform to calculate and display key status parameters of the current fog dissipation performance control system in real time, including the reliability weight of each sensing channel, the health degradation index of each fog dissipation execution unit, the current stable fusion weight set distribution, the component values of the fog dissipation performance control decision vector, and the current operating condition deviation, providing a panoramic view of the system status for the flight crew and ground support personnel. The status monitoring module is presented in an interface combining digital instruments and graphics: the reliability of each sensing channel is displayed in the form of a bar chart, with channels whose reliability is below the preset threshold marked in yellow and accompanied by a "data repair in progress" prompt; the health degradation index of each fog dissipation execution unit is presented in the form of a dashboard, with fog dissipation execution units whose health degradation index exceeds 0.7 marked in red and triggering a maintenance prompt; the stable fusion weight set is displayed in the form of a pie chart showing the weight ratio of each mode, with modes experiencing weight drift marked with a flashing prompt; the fog dissipation performance control decision vector is displayed in the form of a real-time trend curve showing the historical change trajectory of each control quantity, assisting in judging the overall response status of the fog dissipation control system.
[0053] When the health degradation index of any defogging actuator exceeds a preset maintenance alarm threshold (e.g., 0.75), the defogging performance control status monitoring module automatically generates a defogging actuator maintenance alarm. The alarm content includes the type and number of the defogging actuator, the current health degradation index value, the normalized value of each degradation component, the estimated remaining reliable working life, and recommended maintenance measures. The alarm information is presented in a pop-up window on the airborne flight information management system interface and is simultaneously transmitted to the ground operations control center via data link, allowing ground support personnel to prepare corresponding maintenance plans in advance. If the number of defogging actuators with a health degradation index exceeding 0.9 reaches more than 50% of the total number of similar defogging actuators, the system further generates a defogging system reliability degradation warning, recommending that the flight crew take preventive measures such as limiting flight altitude (avoiding high-humidity foggy weather layers) to prevent the defogging system from failing due to the degradation of multiple defogging actuators.
[0054] A ground-based monitoring instrument panel for defogging performance control is simultaneously installed at the ground operations control center. This panel receives defogging control status data from aircraft via a real-time data link, centrally displaying the defogging performance control status of each aircraft's engine on a fleet-by-flight basis. Flights with consistently below-threshold defogging performance or persistently high health degradation indices of their defogging actuators are highlighted, providing data support for route operation management and preventative maintenance decisions. The ground monitoring instrument panel also features historical data analysis capabilities. It statistically analyzes the deviation and weight drift data recorded during defogging control for each flight, identifying frequently occurring off-design operating conditions under specific route or seasonal conditions. The analysis results are then added to the off-design operating condition compensation mapping table in batches, continuously improving the defogging performance control system's adaptability under off-design conditions.
[0055] This invention achieves intelligent closed-loop control of the fogging performance of aero-engines through: credibility assessment and Kriging interpolation repair of multi-source sensing data; segmented extraction of the physical monotonicity constraint set of the heat transfer field and three-dimensional quantitative modeling of the health degradation index of the fogging execution unit; physical compatibility verification and adaptive shrinkage remodeling of the kernel interval and support interval of the fuzzy membership function by the online constraint remodeling module; fusion weight drift detection and positive compensation mechanism based on physical anchor points; and non-design operating condition compensation mapping driven by operating condition deviation and saturation cutoff mechanism for reachable control margin. This invention explicitly transforms the physical monotonicity of the heat exchange field into a computable constraint and embeds it into the fuzzy inference renormalization process, fundamentally preventing the fogging control decision from violating the thermodynamic physical laws of the combustion chamber. The three-dimensional modeling of the health degradation index of the fogging execution unit dynamically incorporates the current health status of the fogging execution unit into the control constraints, ensuring that the control commands are always within the reliable execution capability range of the fogging execution unit. The weight correction mechanism with the heat exchange field feature vector as the physical anchor point effectively suppresses abnormal weight drift caused by membership function mismatch during multimodal fusion. The online accumulation and update mechanism of the non-design condition compensation mapping table gives the system continuous self-learning ability, enabling the fogging efficiency control performance to continuously improve with the accumulation of operational data. It should be noted that the parameter values involved in this application are only examples for reference.
[0056] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.
[0057] All formulas in this manual are dimensionless and calculated numerically. The formulas are derived from software simulations based on a large amount of collected data to obtain the most recent real-world results. The preset parameters and thresholds in the formulas are set by those skilled in the art according to the actual situation.
[0058] Although embodiments of the invention have been shown and described, those skilled in the art will understand that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the claims and their equivalents.
Claims
1. A method for intelligent control of defogging efficiency in aero-engines based on multi-source data fusion, characterized in that, include: Acquire multi-source sensing data of the aero-engine and the current health status parameters of each defogging actuator; the multi-source sensing data includes engine exhaust plume temperature field data, near-field atmospheric humidity and visibility data, combustion chamber heat transfer field distribution data, and intake duct fogging image sequence data, etc.; the health status parameters include response delay, valve position deviation, and cumulative fatigue damage value. Based on the heat transfer field distribution data of the combustion chamber, the physical monotonicity constraint set of the heat transfer field is extracted, and the health degradation index of each demisting actuator is calculated based on the response delay, valve position deviation and cumulative fatigue damage value. Based on the physical monotonicity constraint set of the heat exchange field and the health degradation index, the kernel interval and support interval of the current fog-clearing control fuzzy membership function are physically compatible. Membership function components that violate monotonicity are identified, and the support boundaries of the corresponding membership function components are adaptively contracted and reorganized to output the reorganized fuzzy membership function. Based on the renormalized fuzzy membership degree and the physical monotonicity constraint set of the heat exchange field, the modal data with fusion weight drift in the multi-source sensing data are weighted and a stable fusion weight set is output. Based on the stable fusion weight set, the membership vectors of each modality data are weighted and fused to generate a fog-removing efficiency control decision vector. The fog-removing efficiency control decision vector is then mapped to the target control quantity of each fog-removing execution unit in combination with the deviation of the current operating condition. A fog-removing execution instruction set is then output to drive each fog-removing execution unit to complete the corresponding fog-removing efficiency control operation.
2. The intelligent control method for defogging efficiency of aero-engines based on multi-source data fusion according to claim 1, characterized in that, The set of physical monotonicity constraints for the heat exchange field is used to characterize the physical law of monotonically decreasing heat flux density along the flow direction in the heat exchange field and its permissible boundary constraints.
3. The intelligent control method for defogging efficiency of aero-engines based on multi-source data fusion according to claim 1, characterized in that, The process of acquiring multi-source sensing data includes: Based on the collected engine exhaust plume temperature field data, near-field atmospheric humidity and visibility data, combustion chamber heat transfer field distribution data, and intake duct fogging image sequence data, the credibility of the sensor source to which it belongs is evaluated, and the credibility weight of each acquisition channel is generated. The collected data from each acquisition channel are timestamped and the sampling rate is normalized. In the acquisition channels where the confidence weight is lower than the preset confidence threshold, the missing data of the corresponding acquisition data is repaired by Kriging interpolation based on adjacent trusted channels. Outlier values were removed from the data collected from each acquisition channel after repair, and the collected data were spatially registered and fused in a unified engine coordinate system to obtain complete multi-source sensing data.
4. The intelligent control method for defogging efficiency of aero-engines based on multi-source data fusion according to claim 1, characterized in that, The process of extracting the physical monotonicity constraint set of the heat transfer field includes: The heat transfer field distribution data of the combustion chamber is discretized into several heat transfer sections according to the flow direction, and the heat flux density gradient sequence between adjacent heat transfer sections is calculated. The heat flux density gradient sequence is subjected to a directional consistency test. The section interval corresponding to the heat transfer section where the heat flux density gradient direction reverses is marked as a monotonicity doubt interval. The correlation between the local Reynolds number and Nusselt number of the corresponding heat transfer section is combined to determine whether the monotonicity doubt interval belongs to the physically permissible range of secondary flow disturbance. If it does not fall within the physically permissible range of secondary flow disturbance, extract the upper and lower tolerance boundaries of the heat flux density of the cross-sectional area, construct segmented monotonicity constraints, summarize the segmented monotonicity constraints of the entire flow direction, and form a set of physical monotonicity constraints for the heat transfer field.
5. The intelligent control method for defogging efficiency of aero-engines based on multi-source data fusion according to claim 1, characterized in that, The calculation process for the Health Deterioration Index includes: The dynamic response degradation component, static accuracy degradation component, and structural life degradation component of the defogging actuator are constructed using response delay, valve position deviation, and cumulative fatigue damage value, respectively. The dynamic response degradation component, static accuracy degradation component, and structural life degradation component are normalized respectively, and then weighted and summed according to the influence weight of each component in the fog control. Apply a fogging execution unit type correction coefficient to the summation result, and output the final fogging execution unit health degradation index for each fogging execution unit.
6. The intelligent control method for defogging efficiency of aero-engines based on multi-source data fusion according to claim 1, characterized in that, The process of performing physical compatibility verification includes: Map the kernel interval endpoints of the membership function components within each fog-eliminating control fuzzy membership function to the heat exchange field's membership coordinate system, and compare them with the tolerance boundary constraints of the corresponding cross section in the heat exchange field's physical monotonicity constraint set. If the endpoints of the kernel interval are outside the tolerance boundary constraints, then the corresponding membership function component is determined to have a monotonicity violation. If the endpoints of the core interval are within the tolerance boundary constraints, the support intervals of the corresponding membership function components are extended for compatibility verification. This verifies whether the support intervals of the corresponding components overlap with the support intervals of adjacent components across the physical constraint boundaries. For membership function components that overlap, they are marked as membership function components that violate monotonicity.
7. The intelligent control method for defogging efficiency of aero-engines based on multi-source data fusion according to claim 1, characterized in that, The output process of the renormalized fuzzy membership function includes: Based on the health degradation index of the execution unit, the reachable control margin of the defogging execution unit associated with the membership function component is determined. The reachable control margin is used to characterize the maximum and minimum control output range that the execution unit can reliably achieve in the current health state. The safe shrinkage amount of the support boundary is determined by jointly using the reachable control margin and the physical monotonicity constraint set of the heat transfer field. The safe shrinkage amount is then calculated by shrinking both ends of the support interval toward the core interval for the membership function components that have monotonicity violations. The slope of the contracted support boundary is smoothed to generate a renormalized fuzzy membership function.
8. The intelligent control method for defogging efficiency of aero-engines based on multi-source data fusion according to claim 1, characterized in that, The process of outputting a stable fused weight set includes: Based on multi-source sensing data, the renormalized fuzzy membership function is used to fuzzify the data of each modality and generate membership vectors for each modality. The stability of the fusion weights of each mode is monitored in real time by a fusion weight drift detection mechanism. When a fusion weight drift is detected, a weight correction compensation is triggered with the heat exchange field physical monotonicity constraint set as the anchor point, and a stable fusion weight set is output.
9. The intelligent control method for defogging efficiency of aero-engines based on multi-source data fusion according to claim 1, characterized in that, The process of generating the defogging instruction set includes: The membership vectors of each modality are linearly weighted and summed using the weight values corresponding to each modality data in the stable fusion weight set to obtain the fog-eliminating efficiency control decision vector; Based on the design operating condition benchmark parameters of multi-source sensing data and storage, multi-dimensional Euclidean distance is calculated to obtain the current operating condition deviation. The fog-eliminating efficiency control decision vector is defuzzified to obtain the baseline control quantity, which is then interpolated and superimposed with historical compensation records to obtain the target control quantity of the fog-eliminating execution unit for adaptive operation.
10. The intelligent control method for defogging efficiency of aero-engines based on multi-source data fusion according to claim 8, characterized in that, The process of triggering weighted normalization compensation anchored to the physical monotonicity constraint set of the heat transfer field includes: Extract the heat flux density distribution feature vector corresponding to each segment interval in the physical monotonicity constraint set of the heat exchange field, and use it as the physical anchor point for weight positiveing; Calculate the cosine similarity between the membership vector of the modal data that has experienced fusion weight drift at the current time and the physical anchor point, and recalculate the corrected fusion weight of the modal data using the monotonically increasing function of the cosine similarity. The corrected fusion weights are globally normalized, and the compensation range of the weight correction operation is limited to within a preset compensation range threshold, and the stable fusion weight set is output.
