A temperature monitoring control method and system for a gas generator set

Abstract

The present application belongs to the technical field of gas generator control, and relates to a kind of gas generator set temperature monitoring control method and system, it is first by in key thermal area distributedly embedded temperature sensor array, reconstruct temperature field in conjunction with three-dimensional structure information, to identify local hot spot and quantify its temperature difference with background;Then by fitting hot spot temperature and the time series response of load parameter, calculate thermal inertia coefficient;Subsequently, the real-time mapping relationship of multidimensional characteristic parameter and hot spot temperature difference, thermal inertia coefficient and operating parameter is established respectively, and the operating parameter of cooling component is adjusted based on fuzzy control rule self-adaption;Finally, according to the historical data of temperature gradient change rate, temperature uniformity index and thermal inertia coefficient, multi-level early warning threshold is set, early fault warning is realized;Realize the adaptive, advanced adjustment of gas generator set temperature monitoring control, and build a closed-loop system from perception to intelligent regulation and control, improve the overall operation safety.

Description

Technical Field

[0001] This invention belongs to the field of gas generator control technology, specifically a method and system for temperature monitoring and control of gas generator sets. Background Technology

[0002] As an important energy conversion device, the reliability and efficiency of gas generator sets are closely related to the effectiveness of the thermal management system. Excessive operating temperature or local overheating can accelerate component aging and cause surface damage. Therefore, accurate temperature monitoring and timely control intervention in key hot areas are crucial.

[0003] Currently, temperature monitoring and control of gas generator sets mainly rely on deploying a limited number of discrete temperature sensors in key areas, and using PID control based on fixed thresholds or simple on / off control strategies to adjust the operating parameters of the cooling components based on the measured values ​​of these discrete points.

[0004] However, existing technologies still have the following limitations: 1. Existing methods focus more on absolute temperature values ​​and lack the identification and quantitative analysis of the temperature difference between local hot spots and their surrounding background. They do not fully consider the lag characteristics of temperature changes relative to the dynamic response of unit load, resulting in a mismatch between control response and thermal state changes. In addition, traditional control strategies are mostly based on fixed thresholds or simple PID control, which are difficult to adapt to the complex nonlinear changes in thermal state under variable operating conditions of the unit, and the adjustment is coarse and the adaptive capability is poor.

[0005] 2. Existing early warning methods mostly rely on the absolute threshold of a single or a few temperature measurement points, failing to fully integrate the temporal change rate of the temperature gradient, temperature uniformity index, and thermal inertia coefficient analysis, resulting in insufficient timeliness and accuracy of early warnings. Summary of the Invention

[0006] To overcome the shortcomings of the prior art, embodiments of the present invention provide a method and system for temperature monitoring and control of gas generator sets, which can effectively solve the problems involved in the prior art.

[0007] The objective of this invention can be achieved through the following technical solution: In a first aspect, this invention provides a method for temperature monitoring and control of a gas generator set, comprising: distributively embedding an array of temperature sensors in the key hot areas of the gas generator set.

[0008] Based on the periodic acquisition of temperature data from each monitoring point by the sensor array, and combined with the three-dimensional structural geometric information of the unit, the three-dimensional temperature field of the key thermal area is reconstructed.

[0009] Multidimensional feature parameters, including the highest temperature, lowest temperature, average temperature, and temperature uniformity index, are extracted from the three-dimensional temperature field. Local hotspots are identified and their differences from the average temperature of the background neighborhood are calculated to construct the local hotspot temperature difference.

[0010] The time-series response changes of local hot spot temperature and unit load parameters are fitted, and the time constant of hot spot temperature change lagging behind load change is calculated to obtain the thermal inertia coefficient.

[0011] Establish a real-time mapping relationship between multidimensional characteristic parameters and local hot spot temperature difference, and a real-time mapping relationship between thermal inertia coefficient and unit operating parameters, and adaptively adjust the operating parameters of cooling components based on fuzzy control rules.

[0012] Based on the calculation of the time-series change rate of the temperature gradient in the three-dimensional temperature field, and combined with historical data of temperature uniformity index and thermal inertia coefficient, multi-level early warning thresholds are set to execute fault early warning.

[0013] Secondly, the present invention also provides a temperature monitoring and control system for a gas generator set, comprising: a sensor embedding module, a three-dimensional field reconstruction module, a feature extraction module, a thermal inertia calculation module, a fuzzy adjustment module, and a multi-level early warning module.

[0014] The sensor embedding module is connected to the three-dimensional field reconstruction module, the three-dimensional field reconstruction module is connected to the feature extraction module, the feature extraction module is connected to the thermal inertia calculation module, the thermal inertia calculation module is connected to the fuzzy adjustment module, and the fuzzy adjustment module is connected to the multi-level early warning module.

[0015] The sensor embedding module is a distributed array of temperature sensors embedded in the key hot areas of the gas generator set.

[0016] The three-dimensional field reconstruction module reconstructs the three-dimensional temperature field of key thermal areas by periodically collecting temperature data from each monitoring point based on the sensor array and combining it with the three-dimensional structural geometric information of the unit.

[0017] The feature extraction module extracts multi-dimensional feature parameters such as the highest temperature, lowest temperature, average temperature, and temperature uniformity index from the three-dimensional temperature field, identifies local hotspots, calculates the difference between their relative average temperature and the background neighborhood temperature, and constructs the local hotspot temperature difference.

[0018] The thermal inertia calculation module fits the time-series response changes of local hot spot temperature and unit load parameters, calculates the time constant of hot spot temperature change lagging behind load change, and obtains the thermal inertia coefficient.

[0019] The fuzzy control module establishes a real-time mapping relationship between multi-dimensional feature parameters and local hot spot temperature differences, and a real-time mapping relationship between thermal inertia coefficient and unit operating parameters, and adaptively adjusts the operating parameters of the cooling components based on fuzzy control rules.

[0020] The multi-level early warning module calculates the temporal rate of change of the temperature gradient based on the three-dimensional temperature field, combines historical data of temperature uniformity index and thermal inertia coefficient, and sets multi-level early warning thresholds to execute fault warnings.

[0021] Compared with the prior art, the embodiments of the present invention have at least the following advantages or beneficial effects: (1) The present invention pre-identifies the priority monitoring area in the key hot area through numerical simulation, and non-uniformly deploys the sensor array according to the temperature gradient distribution, and reconstructs the three-dimensional temperature field accordingly, effectively covering the high gradient and easily overheated areas, realizing the identification of temperature spatial distribution and local hot spots and the quantitative analysis of temperature difference, reducing or eliminating the monitoring blind spot from the source, and improving the perception ability of temperature spatial distribution.

[0022] (2) By establishing a real-time mapping between thermal characteristic parameters and temperature difference, thermal inertia coefficient and operating parameters, and by using fuzzy control to adaptively adjust the cooling system, the present invention achieves adaptive and advanced adjustment of the cooling system, making the control response more compatible with the nonlinear changes in thermal state.

[0023] (3) By integrating the temperature gradient change rate, temperature uniformity index and thermal inertia coefficient, this invention constructs a multi-level early warning mechanism based on historical data, which improves the fault identification capability, forms a closed-loop system from perception to intelligent control, and improves the overall operation and maintenance safety. Attached Figure Description

[0024] The present invention will be further described with reference to the accompanying drawings, but the embodiments in the drawings do not constitute any limitation on the present invention. For those skilled in the art, other drawings can be obtained based on the following drawings without creative effort.

[0025] Figure 1 This is a schematic diagram of the method flow of the present invention.

[0026] Figure 2 The flowchart illustrates the method for constructing local hotspot temperature differences according to the present invention.

[0027] Figure 3 This is a schematic diagram of the module connection of the present invention. Detailed Implementation

[0028] 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.

[0029] Reference Figure 1As shown, in a first aspect, the present invention provides a temperature monitoring and control method for a gas generator set, comprising: S1, distributively embedding a temperature sensor array in the key thermal region of the gas generator set, the specific process of which is: based on the three-dimensional geometric model and operating condition parameters of the gas generator set, a computational fluid dynamics and heat transfer coupled simulation method is used for simulation; during the simulation, boundary conditions are set according to the thermal properties parameters such as thermal conductivity and specific heat capacity of the actual material, and the steady-state and transient temperature distribution of the gas generator set under typical load and extreme conditions is simulated, and the spatial distribution data of temperature in the key thermal region is obtained by iterative solution, which is used as the expected distribution of the thermal field.

[0030] The critical hot zone is pre-defined during initialization based on the design characteristics of the gas generator set. Specifically, it consists of spatial parts with high temperature, high heat flux density, or significant temperature gradient caused by the action of high-temperature gas, intense frictional heat generation, or insufficient cooling. These include, but are not limited to, the combustion chamber head, the high-temperature exhaust manifold, and the high-temperature end bearing housing supporting the rotor.

[0031] Using the condition that the temperature in the expected distribution of the thermal field exceeds the preset temperature threshold, the priority monitoring area is identified and delineated. This area usually corresponds to the high-temperature core area shown in the simulation. The preset temperature threshold is determined based on the allowable operating temperature of the material of the component involved in the key thermal area, and is usually set to a value that is lower than the maximum allowable temperature of the material by a certain safety margin.

[0032] By prioritizing the spatial distribution of temperature gradient in the monitoring area, temperature sensors are deployed according to the rule that the temperature gradient value is inversely proportional to the sensor spacing. Specifically, discrete three-dimensional temperature field data generated by numerical simulation in this area are acquired. For each internal grid node, the absolute value of the temperature difference between it and its adjacent nodes in each coordinate axis direction is calculated, and the maximum value in the three directions is taken as the temperature gradient value of the node.

[0033] Meanwhile, based on the physical dimensions of the sensors and the requirements of the installation process, the minimum and maximum allowable spacing are preset. The minimum spacing must ensure that adjacent sensors do not interfere with each other in terms of physical structure, and its value is not less than the maximum outline dimension of the sensor body. The maximum spacing must meet the spatial resolution requirements for the temperature gradient change in the priority monitoring area, and its value is not greater than the spatial dimension of the priority monitoring area.

[0034] For any point within the region, the difference between its temperature gradient value and the minimum gradient value in the region is divided by the difference between the maximum gradient value and the minimum gradient value in the region to obtain the scaling factor.

[0035] The difference between the maximum and minimum deployment spacing is linearly weighted using this proportional coefficient, and the weighted result is subtracted from the maximum deployment spacing to obtain the sensor deployment spacing at that location point. This allows the temperature sensor spacing to decrease as the temperature gradient increases, thereby achieving adaptive matching between the monitoring network density and the severity of thermal changes.

[0036] Temperature sensors are deployed at specific characteristic locations outside the priority monitoring area.

[0037] It should be noted that specific characteristic locations include, but are not limited to: the geometric center of key thermal components on the three-dimensional structural model; locations on the heat conduction path adjacent to the priority monitoring area, such as the physical junction boundary between the intake or exhaust manifold and the cylinder head; specific locations where temperature anomalies have occurred in the past; their selection takes into account structural representativeness, thermal significance, and installation feasibility.

[0038] The sensor points generated adaptively based on temperature gradients within the priority monitoring area are merged with specific feature locations selected outside the priority monitoring area. The spatial coordinates of all points are unified, integrated, and verified to generate a non-uniform spatial distribution configuration that covers the entire key thermal area and matches the spatial density with the intensity of thermal changes. Based on the spatial coordinates of each point in this non-uniform spatial distribution configuration, the actual positioning and embedding of all temperature sensors are completed, thus obtaining the final temperature sensor array. This method achieves the optimized spatial configuration of monitoring resources and achieves a balance between high-density coverage of key areas and global representative sampling.

[0039] This invention uses numerical simulation to pre-identify priority monitoring areas in key hot regions, and deploys sensor arrays non-uniformly according to temperature gradient distribution to reconstruct a three-dimensional temperature field, effectively covering high gradient and easily overheated areas. This enables the identification and quantitative analysis of temperature spatial distribution and local hot spots, reducing or eliminating monitoring blind spots from the source and improving the ability to perceive temperature spatial distribution.

[0040] It should also be noted that the type of temperature sensor can be selected according to the actual monitoring needs, such as one or more combinations of fiber optic temperature sensors or infrared temperature sensors.

[0041] S2. Based on the periodic acquisition of temperature data from each monitoring point by the sensor array, and combined with the three-dimensional structural geometric information of the unit, the three-dimensional temperature field of the key thermal area is reconstructed.

[0042] The specific process of reconstructing the three-dimensional temperature field of the key thermal region is as follows: based on the deployment position of each sensor in the temperature sensor array, the real-time temperature data collected by them is bound to the corresponding spatial coordinates in the three-dimensional structural geometric model of the gas generator set to generate a discrete temperature point set with spatial coordinates.

[0043] Using a discrete set of temperature points, a three-dimensional Delaunay triangulation is performed on the key thermal region, dividing the space of the key thermal region into multiple non-overlapping triangular units. Each triangular unit stores the spatial coordinates of its three vertices and real-time temperature data.

[0044] Select any cell from the set of triangular cells generated by the subdivision as the current cell, and calculate the signed area of ​​the point to be found relative to each side of the current triangle. The sign of this value indicates the orientation of the point to be found relative to that side.

[0045] By dividing the signed area of ​​each of the above sub-triangles by the area of ​​the entire triangle, a set of normalized parameters is obtained and combined to form the centroid coordinates. If all the calculated centroid coordinate components are non-negative, then the point to be located is located inside or on the boundary of this triangle.

[0046] Otherwise, if the location point to be found is located outside the face that generates a negative distance, the search is moved to the cell adjacent to that face, and the calculation and judgment are repeated until the triangular cell containing the location point to be found is located, and the centroid coordinates of the location point to be found in the corresponding triangular cell are calculated.

[0047] Linear interpolation is performed on the temperature data of the three vertices based on the centroid coordinates. The vectors starting from the point to be determined and ending at each vertex of the candidate triangular unit are calculated, and the corresponding vectors are used as the temperature contribution weights of the corresponding vertices. The real-time temperature data of each vertex is multiplied by its corresponding weight, and the three weighted temperature values ​​are summed. The result is the interpolated temperature of the point to be determined at the current time.

[0048] For all the points to be determined in the three-dimensional structural geometric model of the key thermal region, the aforementioned method is used to determine the triangular element to which they belong and calculate their centroid coordinates within that element. Then, based on these centroid coordinates, linear temperature interpolation is performed to finally obtain the temperature value of that point.

[0049] Once the temperature values ​​of all the locations to be determined have been calculated, they together form a continuous three-dimensional temperature field that covers the entire key thermal region and reflects the spatial distribution of temperature at the current moment.

[0050] S3. Extract multidimensional feature parameters including the highest temperature, lowest temperature, average temperature and temperature uniformity index from the three-dimensional temperature field, identify local hotspots and calculate their differences from the average temperature of the background neighborhood to construct the local hotspot temperature difference.

[0051] The identification of local hotspots specifically includes: taking the spatial range of the three-dimensional temperature field as the boundary, dividing the space into multiple voxel grids of the same size according to the preset voxel resolution, and assigning a representative temperature value to each voxel through three-dimensional interpolation to form a voxel temperature matrix.

[0052] The preset voxel resolution is determined based on the sensor density and, in principle, matches the average spacing between the sensors. For example, if the average spacing between sensors in a certain area is 10 mm, then the voxel resolution for that area can be set to 10 mm accordingly.

[0053] Traverse the voxel temperature matrix to determine the global maximum temperature of the three-dimensional temperature field, and calculate the arithmetic mean of all voxel temperatures as the temperature average of the three-dimensional temperature field. Calculate the standard deviation of all voxel temperatures as the temperature uniformity index of the three-dimensional temperature field.

[0054] The highest global temperature is used as the upper limit of the filtering temperature range, and the average temperature is used as the lower limit of the filtering temperature range. All voxels whose temperature values ​​fall within the filtering temperature range are searched to form a candidate voxel set.

[0055] Traverse the candidate voxel set. If the temperature value of a candidate voxel is greater than the sum of the average temperature and the temperature uniformity index, then mark the voxel as a hot spot voxel.

[0056] Connectivity analysis is performed on all hotspot voxels, which aggregate hotspot voxels that are face-adjacent to each other in 3D space to form several connected regions. Each connected region is marked as a local hotspot. Face adjacency means that two voxels share a complete common face.

[0057] Reference Figure 2 As shown, in a preferred embodiment of the present invention, the specific process of constructing the local hot spot temperature difference is as follows: for each local hot spot, the mean value of the coordinates of all voxel center points contained therein is calculated, which is taken as the representative center point of the hot spot. With the representative center point as the center of the sphere and a preset length as the radius, a spherical spatial region is defined in the three-dimensional structural geometric model as the spherical neighborhood.

[0058] From all voxels contained in the spherical neighborhood, remove the voxels that belong to the local hotspot itself, and define the region consisting of all the remaining voxels in the spherical neighborhood as the background neighborhood of the local hotspot.

[0059] The average temperature of all voxels within the local hotspot is calculated as the hotspot average temperature, and the average temperature of all voxels within the background neighborhood is calculated as the background neighborhood average temperature.

[0060] The difference between the average temperature of the hotspot and the average temperature of the background neighborhood is calculated as the local hotspot temperature difference.

[0061] It should be noted that the preset length is determined based on the spatial scale of the hotspot itself, for example, it is taken as 1.5 to 3 times the maximum geometric size of the hotspot in space.

[0062] S4. Fit the time-series response changes of local hot spot temperature and unit load parameters, calculate the time constant of hot spot temperature change lagging behind load change, and obtain the thermal inertia coefficient.

[0063] The thermal inertia coefficient specifically refers to: acquiring the temperature change sequence of local hot spots in the key hot area within a preset time window, and simultaneously acquiring the unit load parameter change sequence related to the key hot area; the preset time window is determined based on the typical load change cycle of the gas generator set, and usually covers at least one complete load rise and fall process.

[0064] The unit load parameter change sequence and the temperature change sequence of local hot spots are time-stamped and preprocessed for data smoothing.

[0065] In a preferred embodiment of the present invention, timestamp synchronization may specifically include: aligning the unit load parameter change sequence and the local hot spot temperature change sequence with time based on a unified clock reference, identifying and processing data points with abnormal timestamps, and using linear interpolation to align data points with deviations in sampling time.

[0066] Data smoothing preprocessing uses a moving average filtering method.

[0067] After the above processing, a discrete-time first-order difference equation is established to determine the dynamic response relationship between the unit load parameters and the local hot spot temperature. The first-order difference equation includes the response gain parameter, the attenuation parameter, and the time offset parameter.

[0068] It should be noted that the first-order difference equation is in the form of: .

[0069] in, This is a time offset parameter, representing the delay time in which load changes are transmitted to temperature changes; The response gain parameter reflects the intensity of the effect of unit load on temperature; This is the attenuation parameter, reflecting the degree of influence of the previous temperature on the current temperature; Represents the current discrete time point. For constant terms, for Local hotspot temperature at any given moment In order to be in Before the moment Unit load parameter values ​​for each sampling time interval; This indicates the temperature of the local hotspot at the previous moment.

[0070] In the above formula, This indicates the delayed effect of unit load on temperature. The larger the load, the more significant the effect of load changes on temperature; This indicates the sustained effect of temperature. The closer it is to 1, the more gradual the temperature change and the greater the influence of historical temperatures.

[0071] This equation comprehensively reflects the combined effects of historical load, historical temperature, and inherent system characteristics on temperature, and is suitable for describing the temperature change behavior of thermodynamic systems with delayed and cumulative effects.

[0072] Based on the first-order difference equation, the least squares method is used to calculate the response gain parameter, attenuation parameter, time offset parameter, and constant term to obtain the attenuation parameter. .

[0073] Obtain attenuation parameters After obtaining the parameter values, and combining them with a fixed sampling time interval, the time constant is calculated based on existing first-order inertial system theory. Finally, the calculated time constant is defined as the thermal inertia coefficient of the local hotspot within the current preset time window.

[0074] S5. Establish a real-time mapping relationship between multidimensional characteristic parameters and local hot spot temperature difference, and a real-time mapping relationship between thermal inertia coefficient and unit operating parameters, and adaptively adjust the operating parameters of the cooling components based on fuzzy control rules.

[0075] The specific process of constructing a real-time mapping relationship between multidimensional feature parameters and local hotspot temperature differences is as follows: real-time acquisition of updated multidimensional feature parameters and corresponding local hotspot temperature differences.

[0076] Collect temperature field data of key thermal regions of gas generator sets under historical normal operating conditions, covering different typical operating conditions; extract time series of the highest temperature, lowest temperature, average temperature and temperature uniformity index from these historical data; calculate the statistical mean and standard deviation of each time series, and define the historical normal fluctuation range of the parameter as the interval determined by the mean plus or minus 3 times the standard deviation, and the upper limit of the interval is the normal threshold of the corresponding parameter.

[0077] For any parameter among the multidimensional feature parameters, when its value exceeds its corresponding normal threshold, the start and end times of the event are recorded to define an abnormal period in time series. Within the abnormal period, the Pearson correlation coefficient between the time series data of this parameter and the time series data of local hotspot temperature difference in the same period is calculated. By traversing all feature parameters and all abnormal periods, multiple correlation coefficients corresponding to each parameter are obtained.

[0078] Simultaneously, curve segments of the same time period are extracted from the time series curve of local hot spot temperature difference. The curve segment of this parameter during the abnormal period and the local hot spot temperature difference curve segment of the same period are combined to form a pair of related curve segments that are strictly aligned in time.

[0079] For each pair of correlation curve segments, the time series data point set of the parameter and the time series data point set of the local hotspot temperature difference are extracted from the multidimensional feature parameters. Based on the two data point sets, the Pearson correlation coefficient is calculated as the correlation coefficient between the parameter and the local hotspot temperature difference in the correlation curve segment.

[0080] By traversing all associated curve segments and their corresponding characteristic parameters, all obtained correlation coefficients are collected to obtain a set of correlation coefficients between each dimension of characteristic parameters and local hotspot temperature differences.

[0081] In the set of correlation coefficients, for each characteristic parameter such as the highest temperature, the arithmetic mean of the multiple correlation coefficients calculated in all correlation curve segments is taken to obtain the correlation coefficient between each characteristic parameter and the local hot spot temperature difference.

[0082] A linear weighted fusion is performed based on the importance weights of each feature parameter to obtain a comprehensive correlation coefficient between the multidimensional feature parameters and the local hotspot temperature difference, which is used to characterize the real-time mapping relationship between the multidimensional feature parameters and the local hotspot temperature difference.

[0083] The importance weights are determined based on a comprehensive assessment of expert experience. Specifically, this involves collecting multiple sets of clearly defined multidimensional characteristic parameters, historical data samples, and their performance analysis reports in typical overheating failure cases. A Delphi method or weighted scoring method is used for quantitative evaluation. Experts are asked to score the relative importance of each characteristic parameter in indicating overall thermal anomalies and the risk of triggering local overheating. After collecting all expert scores, the scores are statistically processed and assigned values. For example, if the calculation results show that the highest temperature has the highest average score, it is assigned a weight of 0.4; the average temperature has the second highest score, assigned a weight of 0.3; the lowest temperature has a relatively low score, assigned a weight of 0.2; and the temperature uniformity index has the lowest score, assigned a weight of 0.1.

[0084] In a preferred embodiment of the present invention, the real-time mapping relationship between the thermal inertia coefficient and the unit operating parameters specifically involves: acquiring the real-time updated thermal inertia coefficient and simultaneously collecting multiple unit operating parameters.

[0085] The unit operating parameters include, but are not limited to, a set of unit output power, fuel flow rate and ambient temperature.

[0086] Using the same calculation logic as the Pearson correlation coefficient mentioned above, the correlation coefficient between the thermal inertia coefficient and the time series of operating parameters of each unit was calculated.

[0087] The operating parameter with the largest absolute value of the correlation coefficient with the thermal inertia coefficient is selected as the dominant operating parameter.

[0088] The correlation coefficient between the dominant operating parameters and the thermal inertia coefficient is used to characterize the real-time mapping relationship between the thermal inertia coefficient and the unit operating parameters.

[0089] In a preferred embodiment of the present invention, the specific process of adaptively adjusting the operating parameters of the cooling component based on fuzzy control rules is as follows: the comprehensive correlation coefficient reflecting the predicted local hot spot temperature difference level of multi-dimensional feature parameters is defined as the first fuzzy input variable, namely the thermal anomaly correlation degree; and the coefficient reflecting the correlation between the thermal inertia coefficient and the dynamic characteristics of the dominant operating parameters is defined as the second fuzzy input variable, namely the thermal dynamic coupling degree.

[0090] Three linguistic values—low, medium, and high—are defined for thermal anomaly correlation, and five linguistic values—negative strong, negative medium, weak, positive medium, and positive strong—are defined for thermal dynamic coupling.

[0091] The fuzzification process is implemented through a membership function. In this embodiment, a triangular membership function is used. This method sets a vertex with a membership degree of 1 and two adjacent endpoints with a membership degree of 0 for each fuzzy set within a defined universe of discourse, forming a triangular membership degree distribution. The triangular membership function is then applied to convert the real-time collected comprehensive correlation coefficient and correlation coefficient into the membership degree of each corresponding language value.

[0092] Based on a pre-set fuzzy rule base, the membership degree is inferred to obtain the fuzzy quantity of cooling regulation output.

[0093] The fuzzy rule base stores the fuzzy rule mapping relationship between thermal anomaly correlation degree, thermal dynamic coupling degree and cooling regulation output fuzzy quantity. The mapping relationship is represented by fuzzy rules in IF-THEN form.

[0094] Among them, the thermal anomaly correlation degree is divided into three linguistic values ​​of low, medium and high according to the numerical range of the comprehensive correlation coefficient, which are used to characterize the level of the predicted local hot spot temperature difference. The cooling regulation output fuzzy quantity is divided into three linguistic values ​​of small, medium and large according to the adjustment requirements of cooling power or flow rate.

[0095] The thermal dynamic coupling degree is divided into five linguistic values ​​based on the numerical range of the correlation coefficient: negative strong, negative medium, weak, positive medium, and positive strong. These values ​​are used to characterize the response characteristics of the system's thermal inertia to changes in the dominant operating parameters. The cooling regulation output fuzzy quantity is divided into five linguistic values ​​based on the adjustment requirements of cooling power or flow rate: decrease, maintain, slightly increase, moderately increase, and significantly increase.

[0096] The output fuzzy quantity is converted into a specific cooling power or flow rate value through defuzzification. The specific process is as follows: the linguistic values ​​and their membership degrees corresponding to the output fuzzy quantity are converted according to the predefined output domain and scaling factor.

[0097] The output domain is determined based on the actual physical adjustment capability and safe operating boundary. Its lower limit usually corresponds to the lowest allowable speed or minimum flow rate of the component, while its upper limit corresponds to its highest speed or maximum design flow rate.

[0098] The scaling factor is the result of a trial-and-error method to determine the normal adjustment range of cooling power or flow rate required to maintain temperature stability under typical operating conditions.

[0099] A precise scalar output value is calculated using the center of gravity method. Based on this scalar output value and the mapping relationship set by the system, the target speed adjustment of the cooling fan or the target flow rate setting of the coolant is determined, and control commands are generated accordingly.

[0100] It should be noted that if multiple rules are triggered simultaneously, a weighted average method is used for defuzzification.

[0101] The fuzzy rule base is constructed through the following steps: First, during historical operation, multidimensional feature parameters, thermal inertia coefficient, dominant operating parameters, and the operating parameters of the cooling components at the corresponding time and their control effect evaluation are collected to form a set of historical control cases.

[0102] Secondly, for each historical control case, the corresponding comprehensive correlation coefficient is calculated as the actual value of the thermal anomaly correlation degree, and the correlation coefficient between the thermal inertia coefficient and the dynamic characteristics of the dominant operating parameters is calculated as the actual value of the thermal dynamic coupling degree.

[0103] Then, based on the actual value distribution of thermal anomaly correlation degree and thermal dynamic coupling degree, and combined with the control effect evaluation, the optimal cooling regulation output fuzzy quantity corresponding to different input combinations is summarized.

[0104] For each linguistic value of thermal anomaly correlation degree and each linguistic value of thermal dynamic coupling degree, their mapping relationship with the output fuzzy quantity in historical cases is statistically analyzed. Input-output correspondences with high frequency and good control effect are extracted to form a preliminary fuzzy rule set.

[0105] Finally, the preliminary fuzzy rule set is structurally represented in the form of IF-THEN rules and stored in the rule base. The completeness and consistency of the rule base are verified through simulation or historical data backtesting. Conflicting rules are eliminated and missing rules are added to form the final fuzzy rule base. The verification of the fuzzy rule base is based on existing technology and will not be elaborated here.

[0106] This invention establishes a real-time mapping between thermal characteristic parameters and temperature difference, and between thermal inertia coefficient and operating parameters, and adopts fuzzy control to adaptively adjust the cooling system, thereby achieving adaptive and proactive adjustment of the cooling system and making the control response more compatible with the nonlinear changes in the thermal state.

[0107] S6. Calculate the time-series change rate of the temperature gradient based on the three-dimensional temperature field, and combine historical data of temperature uniformity index and thermal inertia coefficient to set multi-level early warning thresholds to execute fault early warning.

[0108] In a preferred embodiment of the present invention, the specific process of setting multi-level early warning thresholds is as follows: obtaining historical data sequences of temperature gradient temporal change rate, temperature uniformity index and thermal inertia coefficient of key thermal areas under historical normal operating conditions.

[0109] For each parameter sequence, calculate the arithmetic mean and standard deviation of all its historical data points. Subtract the mean from each original data value in the sequence and then divide by the standard deviation to obtain a standardized sequence with a mean of 0 and a standard deviation of 1.

[0110] Based on the standardized sequence, it is aligned by timestamp and arranged into a matrix with the behavior time points and the columns being the three feature parameters mentioned above.

[0111] Calculate the covariance matrix of the matrix, and perform eigenvalue decomposition on the covariance matrix to solve for its eigenvalues ​​and corresponding eigenvectors.

[0112] Arrange all eigenvalues ​​in descending order, and the eigenvector corresponding to the largest eigenvalue is the direction of the first principal component. Projecting the original three-dimensional standardized data onto this first principal component direction yields a new one-dimensional time series, which is the first dynamic principal component sequence. This first dynamic principal component sequence is established as a comprehensive early warning index. This index integrates dynamic information from three dimensions: temperature gradient change rate, field uniformity, and thermal inertia, and can most effectively reflect the overall dynamic deviation of the operating state of key thermal regions.

[0113] Add or subtract the arithmetic mean of the historical data points calculated above. The standard deviation is used to determine the baseline range for normal fluctuations, where... This is a constant greater than zero selected based on the distribution characteristics of historical data.

[0114] Starting from the upper limit of the baseline interval, a set of strictly increasing warning thresholds is generated by progressively adding different multiples of the standard deviation. Specifically, the first-level warning threshold can be set as the arithmetic mean plus... The second-level warning threshold is set to the arithmetic mean plus the standard deviation. The third-level hazard warning threshold is set to the arithmetic mean plus the standard deviation. Double the standard deviation.

[0115] in ,multiple , , The values ​​were obtained through experimental calibration, specifically by retrieving clearly recorded abnormal events or fault cases from historical operation records and extracting a comprehensive early warning indicator sequence for a specific time window before the event occurred, such as one hour before the fault.

[0116] Calculate the quantiles of the statistical distribution of the comprehensive early warning index sequence, and set the 85th percentile of the sequence as a multiple corresponding to the first-level attention warning threshold. The 90th percentile is set as a multiple of the second-level warning threshold, based on the benchmark value. The 95th percentile is set as a multiple of the third-level danger warning threshold, based on the benchmark value. The percentile setting is an adaptive setting by the implementer based on the system's accuracy requirements.

[0117] If the comprehensive early warning index value calculated in real time exceeds the corresponding early warning threshold level, the corresponding level of fault warning will be executed.

[0118] This invention integrates temperature gradient change rate, temperature uniformity index and thermal inertia coefficient to construct a multi-level early warning mechanism based on historical data, which can identify fault signs earlier and form a closed-loop system from perception to intelligent control, thereby improving the overall operation and maintenance safety.

[0119] Reference Figure 3 As shown, in a second aspect, the present invention also provides a temperature monitoring and control system for a gas generator set, comprising: a sensor embedding module, a three-dimensional field reconstruction module, a feature extraction module, a thermal inertia calculation module, a fuzzy adjustment module, and a multi-level early warning module.

[0120] The sensor embedding module is connected to the 3D field reconstruction module, the 3D field reconstruction module is connected to the feature extraction module, the feature extraction module is connected to the thermal inertia calculation module, the thermal inertia calculation module is connected to the fuzzy adjustment module, and the fuzzy adjustment module is connected to the multi-level early warning module.

[0121] The sensor embedding module is a distributed array of temperature sensors embedded in the key hot areas of the gas generator set.

[0122] The three-dimensional field reconstruction module reconstructs the three-dimensional temperature field of key thermal areas by periodically collecting temperature data from each monitoring point based on the sensor array and combining it with the three-dimensional structural geometric information of the unit.

[0123] The feature extraction module extracts multi-dimensional feature parameters such as the highest temperature, lowest temperature, average temperature, and temperature uniformity index from the three-dimensional temperature field, identifies local hotspots, calculates the difference between their relative average temperature and the background neighborhood temperature, and constructs the local hotspot temperature difference.

[0124] The thermal inertia calculation module fits the time-series response changes of local hot spot temperature and unit load parameters, calculates the time constant of hot spot temperature change lagging behind load change, and obtains the thermal inertia coefficient.

[0125] The fuzzy control module establishes a real-time mapping relationship between multi-dimensional feature parameters and local hot spot temperature differences, and a real-time mapping relationship between thermal inertia coefficient and unit operating parameters, and adaptively adjusts the operating parameters of the cooling components based on fuzzy control rules.

[0126] The multi-level early warning module calculates the temporal rate of change of the temperature gradient based on the three-dimensional temperature field, combines historical data of temperature uniformity index and thermal inertia coefficient, and sets multi-level early warning thresholds to execute fault warnings.

[0127] The above description is merely an example and illustration of the structure of the present invention. Those skilled in the art can make various modifications or additions to the specific embodiments described, or use similar methods to replace them, as long as they do not deviate from the structure of the invention or exceed the scope defined by the present invention, they should all fall within the protection scope of the present invention.

Claims

1. A method for temperature monitoring and control of a gas generator set, characterized in that, include: A distributed array of temperature sensors is embedded in the key thermal areas of the gas generator set. Based on the periodic acquisition of temperature data from each monitoring point by the sensor array, and combined with the three-dimensional structural geometric information of the unit, the three-dimensional temperature field of the key thermal area is reconstructed. Multidimensional feature parameters, including the highest temperature, lowest temperature, average temperature, and temperature uniformity index, are extracted from the three-dimensional temperature field. Local hotspots are identified and their differences from the average temperature of the background neighborhood are calculated to construct the local hotspot temperature difference. The time-series response changes of local hot spot temperature and unit load parameters are fitted, and the time constant of hot spot temperature change lagging behind load change is calculated to obtain the thermal inertia coefficient. Establish a real-time mapping relationship between multidimensional characteristic parameters and local hot spot temperature difference, and a real-time mapping relationship between thermal inertia coefficient and unit operating parameters, and adaptively adjust the operating parameters of cooling components based on fuzzy control rules; Based on the calculation of the time-series change rate of the temperature gradient in the three-dimensional temperature field, and combined with historical data of temperature uniformity index and thermal inertia coefficient, multi-level early warning thresholds are set to execute fault early warning.

2. The method for temperature monitoring and control of a gas generator set according to claim 1, characterized in that, The specific process of the distributed embedded temperature sensor array in the key thermal region is as follows: The expected thermal field distribution of key thermal areas is obtained through numerical simulation. Priority monitoring areas are identified and delineated based on the thermal field temperature exceeding a preset temperature threshold. Temperature sensors are deployed according to the rule that the temperature gradient value is inversely proportional to the sensor spacing, based on the spatial distribution of the temperature gradient in the priority monitoring area. Temperature sensors are deployed at specific characteristic locations outside the priority monitoring area; By merging the locations of all sensors inside and outside the priority monitoring area, a non-uniform spatial distribution configuration is generated, thereby obtaining the temperature sensor array arrangement.

3. The method for temperature monitoring and control of a gas generator set according to claim 1, characterized in that, The identification of local hotspots specifically includes: Using the spatial range of the three-dimensional temperature field as the boundary, the space is uniformly divided into multiple voxel grids of the same size according to the preset voxel resolution, and each voxel is assigned a corresponding temperature value through three-dimensional interpolation to form a voxel temperature matrix. Traverse the voxel temperature matrix to determine the global maximum temperature, average temperature, and temperature uniformity index of the three-dimensional temperature field. The screening temperature range is set based on the global highest temperature and average temperature, and voxels whose temperature values ​​fall within the screening temperature range are retrieved to form a candidate voxel set. Traverse the candidate voxel set. If the temperature value of a candidate voxel is greater than the sum of the average temperature and the temperature uniformity index, then mark the candidate voxel as a hot spot voxel. Connectivity analysis is performed on all hotspot voxels, which aggregate hotspot voxels that are adjacent to each other in 3D space to form connected regions, and each connected region is marked as a local hotspot.

4. The method for temperature monitoring and control of a gas generator set according to claim 3, characterized in that, The specific process for constructing local hotspot temperature differences is as follows: For each local hot spot, the mean value of the coordinates of the center points of all voxels contained therein is calculated and used as the representative center point of the hot spot. With the representative center point as the center of the sphere and the preset length as the radius, a spherical neighborhood is defined in the three-dimensional structural geometric model. Exclude the area occupied by local hotspots from the spherical neighborhood and use the remaining area as the background neighborhood; The average temperature of all voxels within the local hot spot is calculated as the hot spot average temperature, and the average temperature of all voxels within the background neighborhood is calculated as the background neighborhood average temperature. The difference between the average temperature of the hotspot and the average temperature of the background neighborhood is calculated as the local hotspot temperature difference.

5. The method for temperature monitoring and control of a gas generator set according to claim 3, characterized in that, The thermal inertia coefficient is specifically: Acquire the temperature change sequence of local hot spots in the key hot area within the preset time window, and simultaneously collect the unit load parameter change sequence related to the key hot area; Timestamp synchronization and data smoothing preprocessing are performed on the unit load parameter change sequence and the temperature change sequence of local hot spots; Based on the preprocessed data, a discrete-time first-order difference equation is established to determine the dynamic response relationship between unit load parameters and local hot spot temperature. The first-order difference equation includes response gain parameters, attenuation parameters, and time offset parameters. The parameters of the difference equation are jointly estimated to obtain the attenuation parameters. Combined with a fixed sampling time interval, the thermal inertia coefficient of each local hot spot within a preset time window is calculated.

6. The method for temperature monitoring and control of a gas generator set according to claim 1, characterized in that, The process of establishing a real-time mapping relationship between multidimensional feature parameters and local hotspot temperature differences includes: Real-time acquisition of updated multidimensional feature parameters and corresponding local hotspot temperature differences; The upper limit of the historical normal fluctuation range of each of the multidimensional feature parameters is determined based on historical data and used as its normal threshold. Based on the time series data of multidimensional feature parameters and local hot spot temperature difference, construct their respective time series variation curves, extract the curve segment where any parameter in the multidimensional feature parameters exceeds its corresponding normal threshold, and simultaneously extract the local hot spot temperature difference curve segment within the same time period to form a pair of related curves. Calculate the correlation coefficient of each pair of related curve segments to obtain the set of correlation coefficients between the characteristic parameters of each dimension and the local hot spot temperature difference; The correlation coefficients corresponding to the same feature parameter in the correlation coefficient set are arithmetically averaged to obtain the correlation coefficients between each feature parameter and the local hot spot temperature difference. A linear weighted fusion is performed based on the importance weights of each feature parameter to obtain a comprehensive correlation coefficient between the multidimensional feature parameters and the local hotspot temperature difference, which characterizes the real-time mapping relationship between the multidimensional feature parameters and the local hotspot temperature difference.

7. The method for temperature monitoring and control of a gas generator set according to claim 5, characterized in that, The real-time mapping relationship between the thermal inertia coefficient and the unit operating parameters is as follows: It acquires the real-time updated thermal inertia coefficient, as well as the synchronously collected operating parameters of multiple units; Calculate the correlation coefficient between the thermal inertia coefficient and the time series of operating parameters for each unit; The operating parameter with the largest absolute value of the correlation coefficient with the thermal inertia coefficient is selected as the dominant operating parameter; The correlation coefficient between the dominant operating parameters and the thermal inertia coefficient is used to characterize the real-time mapping relationship between the thermal inertia coefficient and the unit operating parameters.

8. The method for temperature monitoring and control of a gas generator set according to claim 7, characterized in that, The specific process of adaptively adjusting the operating parameters of the cooling component based on fuzzy control rules is as follows: The real-time mapping relationship between the multidimensional feature parameters that predict the local hot spot temperature difference level and the real-time mapping relationship between the thermal inertia coefficient that reflects the dynamic characteristics of the dominant operating parameters are used as inputs; The input is fuzzified and converted into membership degrees with corresponding linguistic values; Based on a pre-set fuzzy rule base, the membership degree is inferred to obtain the fuzzy quantity of cooling regulation output; The output fuzzy value is defuzzified into the target adjustment value of cooling fan speed or coolant flow rate, and an adjustment command is generated accordingly.

9. The method for temperature monitoring and control of a gas generator set according to claim 1, characterized in that, The specific process for setting multi-level early warning thresholds is as follows: Historical data sequences of temperature gradient change rate, temperature uniformity index, and thermal inertia coefficient of key thermal regions under historical normal operating conditions were obtained and standardized preprocessed. Based on the preprocessed historical data sequence, feature fusion is performed using principal component analysis to extract the first dynamic principal component as a comprehensive early warning indicator. Calculate the statistical mean and standard deviation of the historical data series of the comprehensive early warning indicators, and determine the benchmark range of normal fluctuations by adding or subtracting a certain number of times the standard deviation from the mean; Starting from the upper limit of the baseline interval, the standard deviation is accumulated at different multiples to generate an increasing warning threshold, which is then divided into three levels: attention, warning, and danger.

10. A temperature monitoring and control system for a gas generator set, characterized in that, include: The sensor embedding module is a distributed array of temperature sensors embedded in the key hot areas of the gas generator set. The three-dimensional field reconstruction module reconstructs the three-dimensional temperature field of key thermal areas by periodically collecting temperature data from each monitoring point based on the sensor array and combining it with the three-dimensional structural geometric information of the unit. The feature extraction module extracts multi-dimensional feature parameters such as the highest temperature, lowest temperature, average temperature, and temperature uniformity index from the three-dimensional temperature field, identifies local hotspots, calculates the difference between them and the average temperature of the background neighborhood, and constructs the local hotspot temperature difference. The thermal inertia calculation module fits the time-series response changes of local hot spot temperature and unit load parameters, calculates the time constant of hot spot temperature change lagging behind load change, and obtains the thermal inertia coefficient. The fuzzy control module establishes a real-time mapping relationship between multi-dimensional feature parameters and local hot spot temperature differences, and a real-time mapping relationship between thermal inertia coefficient and unit operating parameters, and adaptively adjusts the operating parameters of the cooling components based on fuzzy control rules. The multi-level early warning module calculates the temporal rate of change of the temperature gradient based on the three-dimensional temperature field, combines historical data of temperature uniformity index and thermal inertia coefficient, and sets multi-level early warning thresholds to execute fault warnings.

Images