Gas turbine data visualization and intelligent interaction system

Abstract

The invention relates to the technical field of complex industrial equipment monitoring and digital twinning, and discloses a gas turbine data visualization and intelligent interaction system which comprises a sensor array, a computing unit, a force feedback device, an eye tracker and a display terminal. The system constructs a three-dimensional physical field and corrects a velocity vector by using Kriging interpolation and Euler equation momentum balance, and generates a variable radius flow tube reflecting the compressibility of fluid; calculating mass, momentum and energy conservation residual errors to construct a generalized potential energy field, and generating gradient guide force and self-adaptive damping force by driving force feedback equipment; and dynamically adjusting solving precision and rendering parameters through sight tracking. According to the method, the flow field is reconstructed through physical constraint, multi-modal feedback is introduced, the problem that discrete data feature distortion and single visual interaction are difficult to perceive physical consistency is solved, and high-fidelity reduction of the gas turbine flow field and visual touch synchronous accurate diagnosis of fault types are achieved.

Description

Technical Field

[0001] This invention relates to the field of monitoring and digital twin technology for complex industrial equipment, specifically a data visualization and intelligent interaction system for gas turbines. Background Technology

[0002] The operation of gas turbines involves extreme temperatures, high pressures, and complex unsteady fluid dynamics. To ensure the safe and stable operation of the unit, various sensors are typically placed in the casing and key flow channel sections to monitor physical parameters such as temperature, pressure, and vibration. With the development of industrial internet and digital twin technologies, transforming this discrete sensor data into intuitive and interactive three-dimensional flow field information has become crucial for improving operation and maintenance efficiency and fault diagnosis capabilities.

[0003] Existing gas turbine monitoring systems primarily rely on two-dimensional charts or three-dimensional cloud maps based on geometric interpolation for visualization. However, due to the complex internal flow channel structure and sparse distribution of measurement points within a gas turbine, traditional linear or spline interpolation methods often estimate values ​​for unmeasured areas only from a geometric distance perspective, neglecting the laws of conservation of mass, momentum, and energy that fluids must follow during high-speed flow. This data reconstruction method, lacking physical constraints, easily leads to significant deviations between the generated velocity vector field and the actual thermodynamic state, failing to accurately reflect density changes and pressure gradient deflection characteristics during fluid compression or expansion, thus misleading operators' judgment of the flow field morphology.

[0004] Furthermore, relying solely on visual channels to display massive amounts of 3D flow field data has limitations. In complex internal flow channel structures, multiple layers of flow tubes obscure each other, making it difficult for operators to clearly see the flow details in core areas. Simultaneously, existing interaction methods are mostly mouse clicks or touch operations, representing passive information retrieval, where operators cannot perceive the physical consistency behind the data. When abnormal monitoring data occurs, the system typically only issues threshold alarms, unable to intuitively distinguish whether the anomaly stems from a real fluid physics fault (such as surge or vortex breaking) or from sensor drift or data link noise. The lack of physical conservation criteria for assisted interaction makes fault tracing rely on human experience, increasing cognitive load and failing to meet the demands of real-time accurate diagnosis. Moreover, high-precision numerical solutions and real-time rendering of the entire flow field consume enormous computational resources; without attention mechanisms such as eye tracking, it is difficult to balance the smoothness of interaction with the accuracy of the physical model solution within limited computing power. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention provides a gas turbine data visualization and intelligent interaction system, which solves the problems in existing technologies such as the lack of physical conservation constraints in the reconstruction of discrete sensor data leading to distortion of flow field characteristics, the difficulty of intuitively perceiving physical consistency and distinguishing fault types with a single visual modality, and the difficulty in balancing high-precision numerical calculation of the entire flow field with real-time interactive rendering efficiency.

[0006] This invention provides a gas turbine data visualization and intelligent interaction system, which mainly consists of hardware components and software processing modules running on a high-performance computing unit. The hardware components include sensor arrays arranged at key sections and casing locations of the gas turbine unit, a six-degree-of-freedom force feedback interaction device, an eye-tracking device, and a display terminal.

[0007] The high-performance computing unit is equipped with multiple collaborative processing modules. Among them, the data acquisition and preprocessing module receives time-series physical parameters transmitted from the sensor array and performs timestamp alignment and filtering on the data to construct a standardized sensor measurement set. The thermodynamic state space mapping module utilizes this sensor measurement set to construct a continuous three-dimensional physical scalar field based on the geometric domain of the internal flow channel space of the gas turbine unit through spatial interpolation algorithms, and derives the flow field velocity vector by combining it with the gas equation of state. The flow tube topology reconstruction engine generates a visualized flow tube geometric model based on the flow field velocity vector and transmits it to the display terminal for rendering.

[0008] To achieve tactile perception of physical states, the system is equipped with a residual potential energy field generation module and a tactile impedance control module. The residual potential energy field generation module calculates the local residuals of the infinitesimal control volume within the geometric domain under the mass, momentum, and energy conservation equations, and transforms these residuals into a generalized potential energy field. The tactile impedance control module tracks the position of the virtual probe of the six-DOF force feedback interaction device within the generalized potential energy field, generating force feedback control signals that include damping and restoring forces. Furthermore, the gaze-driven synchronization module dynamically adjusts the solution accuracy of the residual potential energy field generation module and the rendering parameters of the flow tube topology reconstruction engine based on the gaze coordinates captured by the eye-tracking device, enabling on-demand allocation of computational and rendering resources.

[0009] In the specific implementation of thermodynamic state space mapping, the system employs the Kriging space interpolation algorithm to construct a three-dimensional physical scalar field. This algorithm utilizes a Gaussian variogram to characterize the spatial correlation of fluid parameters and obtains weighting coefficients that satisfy the conditions of unbiased estimation and minimum estimation variance by solving the Kriging equations. By linearly weighting and combining the physical parameter values ​​of known measurement points, the temperature field, static pressure field, and density field of any grid node within the defined domain are calculated. For the construction of the flow field velocity vector, this module first calculates the average velocity of the channel cross-section, which is obtained by dividing the instantaneous inlet mass flow rate by the density surface integral on the channel cross-section. Subsequently, an initial velocity vector field is constructed by combining the tangential unit vector of the channel centerline, and a correction component is calculated based on the momentum balance relationship of the Euler equations. This correction component is proportional to the ratio of the local pressure gradient to the fluid density, and its direction is opposite to the local pressure gradient. Finally, the correction component is superimposed on the initial velocity vector field to obtain the flow field velocity vector reflecting the deflection characteristics of the unsteady flow pressure gradient.

[0010] During the construction of the flow tube geometry model, the flow tube topology reconstruction engine sets seed points at the inlet section of the flow channel based on mass flow rate density flux weighted sampling, and uses the Runge-Kutta iterative algorithm to integrate the velocity vector in the flow field to generate the central streamline trajectory. The cross-sectional radius of the flow tube is not a fixed value, but is dynamically calculated based on the fluid compressibility. Specifically, the cross-sectional radius is proportional to the cube root of the ratio of the product of the fluid density and velocity modulus at the flow channel inlet to the product of the local fluid density and velocity modulus, thus visually representing the physical characteristics of fluid expansion (coarsening) and compression (narrowing).

[0011] For gaze-driven interactive control, the gaze-driven synchronization module constructs a 3D spatial weight field that moves with the gaze point. At the computational solution level, for regions where the spatial weight value is higher than a high-precision threshold, the system calls the 3D Reynolds-averaged Navier-Stokes equations, which include turbulent viscosity terms, to solve for local residuals; for regions below this threshold, a one-dimensional isentropic flow model or potential flow model that ignores viscosity terms is used to reduce computational load. At the visual rendering level, the flow tube topology reconstruction engine dynamically adjusts the opacity of the flow tube geometry model according to the spatial weight function, ensuring that the gaze point region remains solidly rendered, while non-gaze point regions appear semi-transparent or blurred, eliminating visual occlusion.

[0012] The system drives tactile feedback through the residuals of physical conservation laws. The residual potential energy field generation module calculates the mass conservation residual, momentum conservation residual, and energy conservation residual, respectively. Specifically, the mass conservation residual is related to the fluid density time-varying rate of change and the mass flux divergence; the momentum conservation residual is related to the vector difference between the fluid inertial force and the sum of the pressure gradient and viscous force; and the energy conservation residual is related to the difference between the internal energy change rate and the sum of pressure work, heat conduction, and viscous dissipation. These residuals are weighted and combined to form a total residual scalar field, which is then transformed into a generalized potential energy field through a nonlinear exponential mapping in the form of a power function. This creates a flat potential energy surface in the low residual region and a steep potential barrier in the high residual region.

[0013] The tactile feedback signal includes a gradient guiding force and an adaptive residual damping force. The gradient guiding force corresponds to the negative gradient direction of the generalized potential energy field. The adaptive residual damping force is opposite in direction to the virtual probe velocity vector, and its magnitude is determined by a variable damping coefficient, which increases with the increase of the total residual scalar field value at the probe position. Based on this, the system performs a physical consistency check: if the gradient force points towards the potential well and the damping force is stable, it is determined to be a real physical fault; if the force feedback signal exceeds the energy integral limit in the high-frequency band and the magnitude and direction of the force randomly jump, it is determined to be a sensor or data link fault. When a sensor fault is determined, the gaze-driven synchronization module triggers reverse feedback, introducing random displacement perturbations to the vertices of the flow tube mesh in the corresponding area, causing the visual image to present a noisy blurring effect synchronized with the tactile vibration.

[0014] This invention provides a data visualization and intelligent interaction system for gas turbines. It has the following beneficial effects: 1. This invention superimposes a correction component based on the momentum balance of the Euler equation onto the initial velocity field through a thermodynamic state space mapping module, and dynamically calculates the flow tube radius based on the cube root relationship of the product of fluid density and velocity modulus using a flow tube topology reconstruction engine. This enables accurate reconstruction of the pressure gradient deflection characteristics and fluid compressibility changes in unsteady flow inside a gas turbine. This construction method allows the visualized flow tube model to directly reflect the actual expansion or compression physical state of the gas in terms of geometry, avoiding the flow field distortion caused by traditional linear interpolation methods under sparse measurement point conditions, and improving the characterization accuracy of the spatial distribution of thermodynamic parameters.

[0015] 2. This invention utilizes a residual potential energy field generation module to map the local residuals of the mass, momentum, and energy conservation equations into a generalized potential energy field. Furthermore, a tactile impedance control module generates adaptive damping forces and gradient guiding forces that increase with the residual value, thus realizing the tactile embodiment of thermodynamic state deviations. Operators can directly perceive abnormal regions within the flow channel that violate physical conservation laws, such as shock wave or vortex breakage locations, through force feedback devices. This transforms abstract numerical calculation errors into an intuitive physical impedance experience, enabling rapid location of thermal fault sources in complex flow fields without relying on purely visual interpretation.

[0016] 3. This invention constructs a spatial weight field that moves with the gaze point through a gaze-driven synchronization module, realizing a logical closed loop of on-demand allocation of computing resources and multimodal interaction. The system only calls the three-dimensional Reynolds-averaged Navier-Stokes equations for high-precision solution in the high-weight gaze region, balancing the real-time interaction frame rate and numerical simulation accuracy. At the same time, combined with physical consistency verification criteria, it synchronously triggers high-frequency tactile oscillations and random perturbations of the visual grid when judging sensor failure, enabling operators to intuitively distinguish between real physical failures and data link failures through the noise-generating characteristics of the gaze-touch synchronization, reducing the misjudgment rate in the monitoring process. Attached Figure Description

[0017] Figure 1 This is a diagram of the system software functional module architecture of the present invention; Figure 2 This is a schematic diagram of a typical application process of an embodiment of the present invention.

[0018] The components include: 1. Gas turbine unit; 2. Sensor array; 3. High-performance computing unit; 4. Six-degree-of-freedom force feedback interaction device; 5. Eye-tracking device; 6. Display terminal; 10. Data acquisition and preprocessing module; 20. Thermodynamic state space mapping module; 30. Flow tube topology reconstruction engine; 40. Residual potential energy field generation module; 50. Tactile impedance control module; and 60. Eye-tracking drive synchronization module. Detailed Implementation

[0019] The technical solutions in 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.

[0020] This invention provides a gas turbine data visualization and intelligent interaction system, which includes: a gas turbine unit 1, a sensor array 2, a high-performance computing unit 3, a six-degree-of-freedom force feedback interaction device 4, an eye-tracking device 5, and a display terminal 6.

[0021] The gas turbine unit 1, as the monitored physical object, contains a compressor, combustion chamber, and turbine components. Sensor array 2 is physically arranged at various key sections and casing locations of the gas turbine unit 1 to collect physical parameters during its operation. Sensor array 2 is connected to the data input interface of the high-performance computing unit 3 via an industrial fieldbus or fiber optic network, converting the collected analog signals into digital signals and transmitting them to the high-performance computing unit 3 in real time. The high-performance computing unit 3, as the data processing and control core of the system, is equipped with a graphics processor and a central processing unit (CPU) for performing parallel computing tasks. A six-DOF force feedback interaction device 4 is connected to the high-performance computing unit 3 via a universal serial bus or dedicated I / O interface. This device is equipped with a spatial position sensor and a torque motor, used to input the spatial coordinates and velocity information of the operating probe to the high-performance computing unit 3, and to receive force feedback control signals output by the high-performance computing unit 3 to generate damping or restoring forces. The eye-tracking device 5 is rigidly fixed to the frame of the display terminal 6 or integrated into the head-mounted display. It is connected to the high-performance computing unit 3 via a data cable to capture the operator's gaze coordinates in real time. The display terminal 6 is connected to the video output interface of the high-performance computing unit 3 to present three-dimensional visualization rendering images.

[0022] See attached document Figure 1 , Figure 1 This is a software functional module architecture diagram of a system according to an embodiment of the present invention. The gas turbine data visualization and intelligent interaction system includes: a data acquisition and preprocessing module 10, a thermodynamic state space mapping module 20, a flow tube topology reconstruction engine 30, a residual potential energy field generation module 40, a tactile impedance control module 50, and a line-of-sight driven synchronization module 60. All of the above modules run in the memory of the high-performance computing unit 3 and are executed by the processor.

[0023] The data acquisition and preprocessing module 10 receives time-series data transmitted by the sensor array 2. This module performs timestamp alignment and filtering on sensor data from different sampling frequencies to construct the current time. Sensor measurement set Measurement set Including but not limited to temperature ,pressure mass flow rate Rotation speed and vibration amplitude The data acquisition and preprocessing module 10 transmits the processed standardized data to the thermodynamic state space mapping module 20.

[0024] Thermodynamic state-space mapping module 20 is configured to map discrete sensor measurement sets The mapping is performed as a continuous three-dimensional physical scalar field. The thermodynamic state-space mapping module 20 is based on the geometric domain of the internal flow channel space of gas turbine unit 1. The interpolation algorithm is used to calculate the vector of any spatial location within the domain. fluid density at static pressure and temperature The thermodynamic state space mapping module 20 further combines the gas state equation and flow channel cross-sectional parameters to calculate the flow field velocity vector. The generated scalar field and vector field data are then transmitted to the flow tube topology reconstruction engine 30 and the residual potential energy field generation module 40.

[0025] The flow tube topology reconstruction engine 30 is configured to be based on the flow field velocity vector. Construct a visual flow tube geometry model. The flow tube topology reconstruction engine 30 initiates streamline integration calculations at preset seed point locations to generate the center streamline trajectory. The flow tube topology reconstruction engine 30 reconstructs the flow based on the local fluid density along the streamline. and flow rate Calculate the cross-sectional radius of the flow tube The system generates three-dimensional tubular surface data that encloses the central streamline. This three-dimensional tubular surface data is transmitted to the display terminal 6 for rendering, and at the same time, the spatial coordinate information of the streamline is transmitted to the tactile impedance control module 50 as the basis for the interaction boundary.

[0026] The residual potential energy field generation module 40 is configured to calculate the local residuals of the physical conservation equations and generate a generalized potential energy field. The residual potential energy field generation module 40 incorporates mathematical models of the mass conservation equation, momentum conservation equation, and energy conservation equation. The residual potential energy field generation module 40 receives continuous physical field data output from the thermodynamic state-space mapping module 20, and for the infinitesimal control volume within the flow channel space, calculates the unbalance quantity, i.e., the residual value, after substituting the measured data into the aforementioned conservation equations. The residual potential energy field generation module 40 then converts the mass conservation residuals... Momentum conservation residual and energy conservation residual The weighted combination yields the total residual scalar field. And a generalized potential energy field is generated based on the total residual scalar field. Generalized potential energy field The data is transmitted to the tactile impedance control module 50.

[0027] The tactile impedance control module 50 is configured to operate based on the generalized potential field. Generate force feedback control signals. This module reads the virtual probe position transmitted in real time from the six-degree-of-freedom force feedback interaction device 4. and speed The tactile impedance control module 50 calculates the negative gradient of the generalized potential field at the probe position. As a restoring force component, and based on the total residual value at that location Calculate the variable damping coefficient This generates a damping force component. The module then synthesizes the total feedback force. The signal is sent to the torque motor of the six-degree-of-freedom force feedback interaction device 4 for execution.

[0028] The gaze-driven synchronization module 60 is configured to adjust the allocation of system computing resources based on the operator's gaze point. This module receives the gaze point coordinates input from the eye-tracking device 5. Calculate the spatial weight function The gaze-driven synchronization module 60 operates according to the spatial weight function. The value of the control command is sent to the residual potential energy field generation module 40 to control the accuracy of the conservation equation solution used in different regions. Specifically, the three-dimensional Reynolds-averaged Navier-Stokes equations are used to solve the problem in the gaze-point region, while the one-dimensional isentropic flow model is used in the non-gaze-point region. Simultaneously, when the gaze-driven synchronization module 60 detects a high-frequency oscillation signal output by the tactile impedance control module 50, it sends a visual rendering correction command to the display terminal 6 to perform mesh jittering or blurring processing on the flow tube model in the corresponding region.

[0029] The data acquisition and preprocessing module 10 first acquires the raw time-series dataset from the sensor array 2. Let... The raw dataset collected from gas turbine unit 1 at any given time is ,in Representing the The data acquisition and preprocessing module 10 collects readings from each sensor. Due to high-frequency electromagnetic interference introduced by the industrial environment, the data acquisition and preprocessing module 10 processes the raw dataset. Each signal channel in the process undergoes low-pass filtering. In one specific embodiment, a cutoff frequency of [frequency missing] is used. The Butterworth filter removes noise components above half the Nyquist frequency, resulting in a denoised measurement set. For sensors with different sampling rates (such as high-frequency dynamic pressure sensors and low-frequency thermocouples), the data acquisition and preprocessing module 10 uses linear interpolation to align them to a unified time axis. .

[0030] Thermodynamic state-space mapping module 20 receives measurement sets And based on the geometric model of the gas turbine flow channel. A continuous three-dimensional scalar field is constructed. The flow field inside the gas turbine exhibits a strongly nonlinear distribution, and sensor measurement points are sparsely distributed on the casing wall or cross-section. This is to reconstruct an arbitrary location within the flow channel. For the thermodynamic parameters, this embodiment uses the Kriging space interpolation algorithm. Taking temperature field reconstruction as an example, at any point... Temperature estimate at [location] This can be represented as a linear weighted combination of the temperatures at known measuring points: ; in, For the first Temperature values ​​at known measuring points, The total number of measurement points participating in the interpolation. For the corresponding to the first Weighting coefficients for each measurement point. This is obtained by solving the following Kriging equations, which satisfy the conditions of unbiased estimation and minimum estimation variance: ; In the formula, The variogram is used to describe spatial correlation as a function of distance. The variation law. In this embodiment, a Gaussian variogram model is selected to characterize the continuous and gradual spatial variation of fluid parameters. Indicates the first The measuring point and the first Euclidean distance between the measuring points Indicates the first One measuring point and one point to be estimated The Euclidean distance between them These are Lagrange multipliers. By solving this system of linear equations, the thermodynamic state-space mapping module 20 obtains the weighting coefficients. Then the domain is calculated. Temperature field of all grid nodes Similarly, static pressure field and density field It is also reconstructed using the aforementioned Kriging interpolation method. For the density field... If direct measurement methods are lacking, the thermodynamic state space mapping module 20 relies on the ideal gas law. Calculation, where is the gas constant.

[0031] Based on the obtained continuous scalar field, the thermodynamic state space mapping module 20 further derives the velocity vector field of the flow field. Given the difficulty of directly measuring the velocity vector of the entire flow field, this embodiment uses analytical derivation based on the principle of mass conservation and the geometric characteristics of the flow channel. Let the centerline of the flow channel be at position... The tangential unit vector at that point is This vector is pre-calculated and stored from the gas turbine CAD geometric model. (Define location) The corresponding flow channel cross section Its effective circulation area is According to the mass continuity equation, the magnitude of the average flow velocity at this cross section... The calculation is as follows: ; in, The instantaneous mass flow rate is measured at the compressor inlet, with the denominator being the surface integral of the density at that cross-section. After obtaining the average flow velocity, the flow channel geometric guidance effect is superimposed to construct the initial velocity vector field: ; To more realistically reflect unsteady flows (such as swirling or secondary flows) under actual operating conditions, the thermodynamic state space mapping module 20 introduces a correction term based on the pressure gradient. According to the momentum balance relationship of the Euler equations, the correction component of the velocity vector... With local pressure gradient Proportional. The corrected final velocity vector field. Represented as: ; In the formula, The pressure gradient correction coefficient is an empirical constant used to adjust the degree to which the pressure gradient deflects the mainstream velocity direction. The final generated three-dimensional velocity vector field... Includes spatial coordinates The corresponding velocity component The data is stored in the data buffer and called by the flow pipe topology reconstruction engine 30.

[0032] The flow tube topology reconstruction engine 30 is based on the velocity vector field output by the thermodynamic state space mapping module 20. A visualized flow tube geometric model is constructed using a streamline integral algorithm based on the Lagrange perspective. This process is not a simple geometric stretching, but a numerical solution process that strictly follows the equations of motion in fluid mechanics.

[0033] First, the flow tube topology reconstruction engine 30 defines an initial seed point set at the inlet section of the gas turbine flow channel (such as the compressor inlet or the combustion chamber fuel nozzle outlet). .set up ,in For the first The starting coordinates of each flow tube. To ensure that the flow tube distribution can cover the critical flow channel area, the spatial distribution of seed points is weighted based on the mass flow rate density flux at the inlet section, that is, the seed points are more densely distributed in areas with high flow rate density.

[0034] For any given seed point The Flow Tube Topology Reconstruction Engine 30 generates the central streamline trajectory by solving the initial value problem of the ordinary differential equation. The position vector of any point on the streamline. With virtual integral time The evolution follows the following equation: ; in, For the current physical time The instantaneous velocity vector field is shown. To ensure the numerical stability and accuracy of the integration path, this embodiment uses a fourth-order Runge-Kutta iterative algorithm for position updates. The position is updated based on the number of integration steps. arrive The position update formula is as follows: ; In the formula, This represents the integration time step. Four intermediate slope vectors. The calculations are as follows: ; ; ; ; Through the above iterative calculations, a series of discrete spatial point sequences are generated. Connecting these points constitutes the first... The central skeleton line of the flow tube.

[0035] After obtaining the central streamline, the flow tube topology reconstruction engine 30 further constructs a volumetric tubular surface to visually represent the changes in the physical properties of the working fluid. The flow tube can be positioned at any point along the streamline. Cross-sectional radius at It is not a fixed value, but rather dynamically related to the local fluid density. and flow rate Therefore, the flow tube radius The calculation formula is defined as follows: ; in, Set the base radius of the flow tube at the inlet. and These represent the fluid density and velocity modulus at the inlet, respectively. This formula causes the flow tube to automatically widen in regions of fluid expansion (density decrease) or deceleration, and automatically narrow in regions of compression (density increase) or acceleration, thus visually directly mapping the changes in the thermodynamic state of the working fluid.

[0036] To address the issue of internal fault paths being invisible due to mutual occlusion among multiple flow tubes in complex flow channels, the flow tube topology reconstruction engine 30 introduces a transparency control mechanism based on line-of-sight weights. The flow tube topology reconstruction engine 30 receives a spatial weight function from the line-of-sight driven synchronization module 60. Regarding spatial location The surface of the flow tube at that location has an opacity. Dynamically adjusted to: ; In the formula, Maximum opacity (fully visible) The minimum opacity (semi-transparent or nearly transparent) is used. This mechanism ensures that the flow tubes in the operator's viewing area remain rendered with high-resolution solidity, while the flow tubes in the non-viewed area are automatically blurred. This preserves the overall flow field context information while achieving visual de-occlusion of critical fault propagation paths. The final generated flow tube mesh data, including geometric vertex positions, normal vectors, and texture coordinates, is transmitted to the display terminal 6 for real-time rendering.

[0037] The core function of the residual potential energy field generation module 40 is to construct a scalar field that can quantify the physical reliability of the data. This module does not rely on simple threshold judgment, but rather performs consistency verification on the currently observed physical field data based on the three conservation laws of fluid mechanics (mass, momentum, and energy).

[0038] First, the residual potential energy field generation module 40 targets the flow channel space. For any infinitesimal control element within the space, establish a system of governing equations. For mass conservation, define the mass residual operator. This operator calculates the imbalance between the rate of change of fluid density over time and the mass flux divergence: ; In the formula, The density field provided for the thermodynamic state-space mapping module 20 For velocity vector field, This is for divergence calculation. Under ideal physical consistency, this value is zero; when sensor data experiences non-physical abrupt changes, this value increases.

[0039] For conservation of momentum, define the momentum residual operator. This operator calculates the vector difference modulus between the fluid inertial force term and the surface forces (pressure gradient and viscous forces) and volume forces acting on the fluid element: ; In the formula, Let be the mass derivative of velocity, and let represent the total acceleration of the fluid element. For pressure gradient; This is the viscous stress tensor. For Newtonian fluids, ,in The dynamic viscosity coefficient is... It is a unit tensor. It reflects whether the flow field motion state conforms to Newton's second law.

[0040] For energy conservation, define the energy residual operator. This operator calculates the difference between the rate of change of internal energy and the work done by pressure, heat conduction, and viscous dissipation: ; In the formula, For specific internal energy, it can be obtained from temperature using the gas law. Calculated; The thermal conductivity coefficient; It is a viscous dissipation function, representing the rate at which mechanical energy is converted into heat energy.

[0041] To unify the residuals from the above three dimensions into a single scalar index for driving force feedback, the residual potential energy field generation module 40 constructs a total residual cost function. Considering the different dimensions of the residual terms, normalized weighting coefficients need to be introduced. , , : ; In the formula, the subscript It represents reference characteristic quantities (such as characteristic density, characteristic time, characteristic velocity, and characteristic length at the inlet) to eliminate the influence of dimensions.

[0042] Finally, the residual potential energy field generation module 40 is based on the total residual cost function. Generate a generalized potential energy field This potential field not only reflects the magnitude of the residual but also enhances the gradient characteristics of the anomalous region through nonlinear mapping, thus facilitating tactile perception. The generalized potential field is defined as follows: ; in, It is the potential energy gain constant, used to adjust the overall amplitude range of the potential energy field to match the output capability of the force feedback device; It is a non-linear exponent, taking values ​​of This index setting results in a flat potential energy field in the low residual region (normal physical fluctuation region), while the potential energy rises sharply in the high residual region (non-physical anomaly region), forming a steep potential barrier. The generated generalized potential energy field data... It is stored in the form of three-dimensional mesh data and transmitted in real time to the tactile impedance control module 50 as a direct basis for calculating tactile restoring force and damping force.

[0043] The core task of the tactile impedance control module 50 is to transform the abstract generalized potential energy field. It converts force feedback into a force signal that the operator can intuitively perceive, and provides a mechanism for verifying physical consistency. The module operates at a high refresh rate of 1kHz to ensure the continuity and hardness of the force feedback.

[0044] First, the tactile impedance control module 50 acquires in real time the position vector of the end effector (i.e., virtual probe) of the six-degree-of-freedom force feedback interaction device 4 in the virtual flow channel space. and velocity vector Based on the generalized potential field The tactile impedance control module 50 calculates the total feedback force output to the handle. This feedback force is guided by a gradient. and adaptive residual damping force It consists of two parts: ; Gradient guiding force Its function is to guide the operator to avoid non-physical data anomalies, or to direct the probe to a point of potential energy minima (i.e., the physically most consistent state point). Its calculation is based on the negative gradient of the potential field at the current location: ; Adaptive residual damping force Its function is to simulate the viscous properties of fluids in high-entropy or high-uncertainty regions, forcing operators to slow down their operations in abnormal areas for more precise sensing. Unlike traditional fixed damping, the damping coefficient in this embodiment... Total residual Nonlinear functions: ; Among them, the variable damping coefficient The calculation formula is: ; In the formula, Based on environmental damping, it simulates the viscous feel of normal fluids; This is the maximum damping saturation value to prevent excessive resistance from damaging the equipment. The sensitivity adjustment coefficient controls the rate at which damping increases with the residual. The hyperbolic tangent function tanh ensures that the damping coefficient is within a certain range. Smooth transition within the interval.

[0045] Based on the force feedback model described above, the tactile impedance control module 50 provides two distinct tactile response modes to assist the operator in identifying the nature of the fault (i.e., physical consistency verification criteria): Scenario 1: Verification of Real Physical Faults When a gas turbine experiences a real component failure (such as combustion chamber flameout or compressor surge), although the monitored parameters deviate from the design values, the fluid medium still obeys the laws of physical conservation. At this time, the generalized potential energy field... A stable local potential well will form at the location of the fault. Tactile characteristics: When the operator pushes the probe towards this area, they will feel... It manifests as a restoring force pointing towards the center of the potential well, and a damping force. A steady increase. The probe is attracted and stably locked in a certain position, feeling firm yet elastic. Mathematical criterion: The tactile impedance control module 50 detects within the time window. Inside, location Potential gradient norm at (Approaching zero), and the Hessian matrix of the potential energy field It is a positive definite matrix, meaning it contains all eigenvalues. .

[0046] Scenario 2: Verification of Sensor / Data Link Failure. When the anomaly is caused by sensor drift, open circuit, or data packet loss, there is a serious physical conflict between the measured data (e.g., a sudden pressure increase but constant temperature, violating the state equation). In this case, the residuals of the conservation equation... The magnitude is extremely large and fluctuates dramatically with spatial location, resulting in a generalized potential energy field. The terrain is characterized by its fragmented, steep, and discontinuous features. Tactile characteristics: When an operator attempts to touch the area, they will feel... The direction and magnitude of the tactile impedance control module exhibit high-frequency random fluctuations, producing a strong sense of jitter or graininess. The probe cannot stably remain on the data point, as if pushing a marble on a smooth sphere, experiencing random repulsive forces from all directions. Mathematical criterion: 50 pairs of feedback force signals from the tactile impedance control module. Perform frequency domain analysis. If in the high-frequency band... Energy integral within (e.g., 30Hz–300Hz) Exceeding the preset threshold If so, it is determined to be a non-physical fault: ; Through the above mechanism, the tactile impedance control module 50 transforms the complex physical equation residuals into an intuitive tactile sensation, enabling the operator to quickly distinguish fault types based on proprioception.

[0047] To achieve high-fidelity physics reconstruction and real-time force feedback with limited computing resources, the gaze-driven synchronization module 60 establishes a dynamic computing resource allocation architecture based on human visual attention mechanisms. This module not only optimizes computational efficiency but also coordinates the consistency between visual and tactile responses in fault representation.

[0048] First, the gaze-driven synchronization module 60 receives the coordinates of the operator's gaze point on the display plane from the eye-tracking device 5. And through the inverse transformation of the camera projection matrix, it is mapped to the gaze focus in the three-dimensional flow channel space. To define the spatial distribution of computational accuracy, the gaze-driven synchronization module 60 constructs a three-dimensional spatial weight field that moves with the gaze point. The weight field is expressed in Gaussian kernel form: ; in, From spatial grid point x to the gaze focus Euclidean distance, The radius parameter for the region of interest is typically set to the corresponding viewpoint. Spatial range.

[0049] Based on weight function The residual potential energy field generation module 40 executes a dynamic model switching strategy. For ( For the core gaze region with a high-precision threshold, such as 0.8, the system calls a high-fidelity 3D Reynolds-averaged Navier-Stokes equation solver to calculate the local residuals. At this point, the viscous stress tensor in the momentum equation... Includes complete turbulent viscosity terms It can accurately capture the characteristics of swirling and separated flows: ; In the formula, It is turbulent kinetic energy.

[0050] for In the outer region, the system automatically downgrades to a simplified one-dimensional isentropic flow model or potential flow model for rapid estimation. At this point, the momentum equation simplifies to the Euler equation, neglecting the viscous term. Only the pressure gradient balance along the streamline is calculated. This hybrid solution strategy significantly reduces the overall computational load while ensuring the realism of the physical feedback at the point of view, and ensures that the force feedback loop maintains an update frequency of more than 1 kHz.

[0051] Furthermore, the gaze-driven synchronization module 60 establishes a reverse feedback loop from touch to vision to enhance the perception of non-physical faults. When the tactile impedance control module 50 detects a high-frequency oscillation signal (i.e., satisfying...), When the criteria are met, the gaze-driven synchronization module 60 sends a synchronization rendering command to the display terminal 6. This command directly acts on the vertex shader of the flow pipe topology reconstruction engine 30. For spatial regions in a high-frequency oscillation state... Internal flow tube mesh vertices Introduce a random displacement perturbation synchronized with the frequency of tactile vibration. : ; In the formula, The amplitude coefficient of visual noise. It is a random unit vector that changes over time. This mechanism allows the flow tube to simultaneously appear as a noisy blur with unstable signal when the operator's hand feels unreliable jitter in the data, thereby enhancing the cognitive certainty of sensor failure through both visual and tactile channels.

[0052] See attached document Figure 2 , Figure 2 This is a schematic diagram of a typical application process according to an embodiment of the present invention. This section will describe in detail the overall workflow of the system through a specific scenario of monitoring abnormal gas turbine exhaust temperature dispersion.

[0053] Step S201: Full-field monitoring and flow tube horizon construction The system is in routine inspection mode. High-performance computing unit 3 receives data collected in real time from sensor array 2. The data acquisition frequency is set to 50Hz. The thermodynamic state space mapping module 20 continuously calculates and updates the overall thermodynamic field, and the flow tube topology reconstruction engine 30 generates a flow tube view covering the entire compressor, combustion chamber, and turbine path. At this time, due to stable unit operation, the residual potential energy field generation module 40 calculates the overall residual... At a low level, the display terminal 6 presents a semi-transparent, smooth-flowing, regular tube shape.

[0054] Step S202: Exception Triggering and Visual Guidance At a certain moment, the system detected abnormal fluctuations in the thermocouple readings at the combustion chamber outlet section, causing the exhaust temperature dispersion in a localized area to exceed the warning value. The residual potential energy field generation module 40 calculated and found the energy conservation residual in this region. Rise, and in the potential energy field A high-potential-energy bulge forms in the middle. At this time, the flow tube topology reconstruction engine 30 automatically adjusts the opacity of the flow tube corresponding to this abnormal region. Upgrade to 1.0 (completely opaque) and map its color to a highlight warning color (such as red) to visually guide the operator's attention.

[0055] Step S203: Eye Focusing and High-Precision Reconstruction The operator notices the anomaly and focuses their gaze on it. Focusing on the highlighted flow tube area. The gaze-driven synchronization module 60 captures this gaze behavior and immediately... Center, radius The computational model within the range is switched to a 3D RANS solver. The system begins high-precision physics reconstruction in this local area, preparing data for subsequent fine tactile detection.

[0056] Step S204: Tactile Exploration and Force Feedback Interaction The operator holds the handle of the six-DOF force feedback interactive device 4 to drive the virtual probe. Approaching the abnormal flow tube area.

[0057] Approach Phase: When the probe enters the edge of the high residual region, the tactile impedance control module 50 determines the potential energy gradient based on the approach phase. The output repulsive force makes the operator feel the handle become heavier, as if a probe is cutting into a viscous fluid.

[0058] Contact phase: Operator applies external force Overcoming the repulsive force, the probe is inserted deep into the core of the anomaly for detection.

[0059] Step S205: Physical Consistency Verification and Conclusion Determination The system assists the operator in determining the nature of the fault based on the force feedback characteristics during the contact phase. Scenario A (Combustion chamber nozzle carbon buildup blockage): If the operator feels the probe being captured or attracted by a stable force field, the feel is a steady elastic damping without high-frequency vibration. This indicates that although the temperature field is distorted, the fluid momentum and energy equations remain closed, and the physical logic is self-consistent. Conclusion: Confirmed as a genuine physical fault (such as localized lean-fuel flameout caused by nozzle blockage).

[0060] Scenario B (Poor Thermocouple Contact): If the operator experiences severe, chaotic, granular vibrations from the handle, and the probe cannot remain stable at a certain point, and simultaneously, the flow tube grid on display terminal 6 shows noisy jitter, this indicates that the data at this location severely violates the laws of conservation of mass and energy (e.g., the rate of temperature change exceeds the physical limit of thermal inertia). Conclusion: Confirmed as a sensor or signal transmission link failure; this is a false alarm.

[0061] Step S206: Maintain decision support Based on the above determination, the system automatically generates a maintenance recommendation report. In case of scenario A, the system highlights the faulty nozzle number and recommends stopping the machine for inspection; in case B, the system recommends disabling the faulty sensor channel and continuing operation to avoid economic losses caused by unplanned downtime.

Claims

1. A gas turbine data visualization and intelligent interaction system, characterized in that, The system comprises: a sensor array arranged at key cross sections and shell positions of a gas turbine unit, a high-performance computing unit in communication connection with the sensor array, a six-degree-of-freedom force feedback interaction device connected with the high-performance computing unit, an eye tracking device, and a display terminal; the high-performance computing unit runs: a data acquisition and preprocessing module for timestamp alignment and filtering processing of time series data collected by the sensor array, and constructing a sensor measurement set; a thermodynamic state space mapping module for mapping the sensor measurement set to a continuous three-dimensional physical scalar field through a spatial interpolation algorithm based on the geometric domain of the internal flow passage space of the gas turbine unit, and deducing a flow field velocity vector in combination with a gas state equation; a flow pipe topology reconstruction engine for constructing a visual flow pipe geometric model based on the flow field velocity vector, and transmitting the flow pipe geometric model to the display terminal for rendering; a residual potential field generation module for calculating local residuals of micro-element control bodies in the geometric domain under mass, momentum and energy conservation equations, and constructing a generalized potential field based on the local residuals; a haptic impedance control module for generating a force feedback control signal containing damping force and restoring force according to the position of a virtual probe of the six-degree-of-freedom force feedback interaction device in the generalized potential field; a line-of-sight driving synchronization module for adjusting the solving accuracy of the residual potential field generation module and the rendering parameters of the flow pipe topology reconstruction engine according to the gaze point coordinates captured by the eye tracking device.

2. The gas turbine data visualization and intelligent interaction system of claim 1, wherein, The way the thermodynamic state space mapping module constructs the flow field velocity vector includes: calculating the average flow velocity of the flow passage cross section, which is obtained by dividing the instantaneous mass flow rate at the inlet by the density area integral on the flow passage cross section; constructing an initial velocity vector field in combination with the tangential unit vector of the flow passage center line; calculating a correction component based on the momentum balance relationship of Euler equation, which is proportional to the ratio of local pressure gradient and fluid density and opposite in direction to the local pressure gradient; superimposing the correction component to the initial velocity vector field to obtain the final flow field velocity vector.

3. The gas turbine data visualization and intelligent interaction system of claim 1, wherein, The way the flow pipe topology reconstruction engine constructs the flow pipe geometric model includes: setting a seed point at the flow passage inlet section according to the mass flow density flux weighted sampling; generating a center streamline trajectory by integrating the flow field velocity vector from the seed point using the Runge-Kutta iteration algorithm; calculating the cross-sectional radius of the flow pipe geometric model, which is proportional to the square root of the ratio of the product of fluid density and flow velocity modulus at the flow passage inlet to the product of local fluid density and flow velocity modulus.

4. The gas turbine data visualization and intelligent interaction system of claim 1, wherein, The control way of the line-of-sight driving synchronization module on the flow pipe topology reconstruction engine includes: calculating a spatial weight function based on the gaze point coordinates; the flow pipe topology reconstruction engine dynamically adjusts the opacity of the flow pipe geometric model according to the spatial weight function, so that the flow pipe located in the gaze point area remains solid rendering, and the flow pipe located in the non-gaze point area presents a semi-transparent or virtual state.

5. The gas turbine data visualization and intelligent interaction system of claim 1, wherein, The residual potential field generation module generates the local residual in the following manner: calculating a mass conservation residual, which is the absolute value of the sum of the time rate of change of fluid density and the divergence of mass flux; calculating a momentum conservation residual, which is the vector difference modulus of the sum of the fluid inertial force term and the pressure gradient and viscous force acting on the fluid element; calculating an energy conservation residual, which is the absolute value of the difference between the internal energy change rate and the sum of pressure work, heat conduction and viscous dissipation; combining the mass conservation residual, the momentum conservation residual and the energy conservation residual by introducing a normalized weight coefficient to obtain a total residual scalar field.

6. The gas turbine data visualization and intelligent interaction system of claim 5, wherein, The residual potential field generation module generates the generalized potential field in the following manner: performing nonlinear exponential mapping on the total residual scalar field to generate the generalized potential field, wherein the nonlinear exponential mapping adopts a power function form; The force feedback control signal generated by the haptic impedance control module includes a gradient guide force and an adaptive residual damping force; The gradient guide force is the negative gradient of the generalized potential field at the virtual probe position; The adaptive residual damping force is opposite to the velocity vector direction of the virtual probe, and the modulus of the adaptive residual damping force is determined by a variable damping coefficient, which increases with the increase of the total residual scalar field value at the virtual probe position.

7. The gas turbine data visualization and intelligent interaction system of claim 6, wherein, The haptic impedance control module is further configured to perform physical consistency verification and judgment: When it is monitored that the gradient guide force points to the center of the local potential well, and the adaptive residual damping force increases smoothly, it is determined that it is a real physical fault that follows the law of conservation of physics; When it is monitored that the energy integral of the force feedback control signal in a preset high frequency band exceeds a preset threshold, and the size and direction of the force acting on the virtual probe randomly jump, it is determined that it is a sensor or data link fault that violates the law of conservation of physics.

8. The gas turbine data visualization and intelligent interaction system of claim 7, wherein, The line-of-sight driving synchronization module is further configured with a reverse feedback link: When the haptic impedance control module determines that it is a sensor or data link fault and outputs a high-frequency oscillation signal, the line-of-sight driving synchronization module sends a synchronous rendering instruction to the display terminal; The display terminal introduces random displacement perturbation to the grid vertices of the flow pipe geometric model corresponding to the region in response to the synchronous rendering instruction.

9. The gas turbine data visualization and intelligent interaction system of claim 5, wherein, The line-of-sight driving synchronization module adjusts the solving precision of the residual potential field generation module in the following manner: constructing a three-dimensional space weight field that moves with the gaze point; in the region where the space weight value is higher than the high precision threshold, controlling the residual potential field generation module to solve the local residual by using three-dimensional Reynolds averaged Navier-Stokes equations containing turbulent viscosity terms; in the region where the space weight value is lower than the high precision threshold, controlling the residual potential field generation module to solve the local residual by using a one-dimensional isentropic flow model or a potential flow model that ignores viscous terms.

10. The gas turbine data visualization and intelligent interaction system of claim 1, wherein, The thermodynamic state space mapping module constructs the three-dimensional physical scalar field by using a Kriging spatial interpolation algorithm: a Gaussian type variation function is used to represent the spatial correlation of fluid parameters; The weight coefficients are obtained by solving a Kriging equation set, and the weight coefficients satisfy the conditions of unbiased estimation and minimum estimation variance; The physical parameter values of the known measuring points are linearly combined to obtain the temperature field, static pressure field and density field of any grid node in the geometric domain, and the temperature field, the static pressure field and the density field are transmitted to the flow pipe topology reconstruction engine and the residual potential field generation module as the continuous three-dimensional physical scalar field.

Images