An intelligent vacuum pouring processing system suitable for a lightweight transformer

Abstract

The present application belongs to the technical field of transformer manufacturing, and particularly relates to an intelligent vacuum pouring processing system suitable for lightweight transformers, which executes the following method: preheating the winding and vacuum degassing at a first pressure threshold; online monitoring the dynamic viscosity of the epoxy resin combination and adjusting the temperature to maintain the viscosity in a target interval; after constant flow injection in the pouring initial stage, switching to a coordinated feedback mode based on the filling pressure gradient and the cavity cross-sectional area of the mold, and maintaining the linear rising speed of the liquid surface constant by controlling the average cross-sectional flow rate; performing controlled pressure alternating driving micro-bubble removal in the vacuum chamber, and using infrared thermal imaging for temperature field compensation. The present application realizes full-process closed-loop control and dynamic viscosity adjustment, significantly enhances filling uniformity, reduces solidification residual stress, and is particularly suitable for high-quality vacuum pouring of the gap between the micro-fine windings of lightweight transformers.

Description

Technical Field

[0001] This invention belongs to the field of transformer manufacturing technology, specifically relating to an intelligent vacuum casting processing system suitable for lightweight transformers. Background Technology

[0002] With the lightweighting of power equipment becoming a core trend in the industry, lightweight transformers, with their high power density, compact size, and ease of installation in complex environments, have been widely used in urban rail transit, aerospace, ocean-going vessels, and new energy grid connection. For dry-type lightweight transformers, vacuum casting is a core production process that determines their insulation strength and mechanical properties. Because lightweight transformers typically employ more precise structural designs and high-performance composite materials, their internal winding gaps are smaller and their structures are more complex. This places extremely high demands on the permeability, filling uniformity, and air bubble residue control of the epoxy resin composite material during the vacuum casting process.

[0003] Existing transformer casting technologies employ multi-tank, independently configured mixing systems. These systems achieve precise control of raw material ratios through physical isolation and programmed feeding of resin storage tanks, curing agent storage tanks, and final mixing tanks. Another approach focuses on physical intervention during the casting process, such as installing mechanical supports and flow rate regulating components within a vacuum chamber to improve the density of the cast component through physical bubble-breaking mechanisms. However, while these existing technologies have played a positive role in improving production efficiency and reducing raw material losses in their specific application scenarios, they still exhibit the following limitations when dealing with the structural characteristics of lightweight transformer windings, which involve highly compressed winding space and reduced material wall thickness.

[0004] The technical challenges of vacuum casting for lightweight transformers can be summarized into three fundamental physical problems. First, there is the issue of fluid permeation kinetics in porous microchannels. The micron-scale of the winding gaps strictly governs the filling process of the epoxy resin composite material according to Darcy's law: flow rate is directly proportional to the pressure gradient and inversely proportional to dynamic viscosity. Viscosity increases exponentially with curing time, constituting a core technical constraint. Second, there is the issue of the nucleation, growth, and escape mechanisms of bubbles in viscous liquids. The root cause of microbubble residue lies in capillary resistance preventing the epoxy resin composite material from entering dead zones and blind pores. In a static vacuum environment, bubble removal relies solely on buoyancy, resulting in insufficient driving force. The physical essence is the competition between gas-liquid interfacial tension and pressure difference. Third, there is the issue of chemical shrinkage and stress accumulation during the curing process of thermosetting materials. The epoxy resin composite material shrinks by approximately 3% to 6% during curing. Stress concentration in thin-walled lightweight structures is the root cause of microcracks, essentially due to the mismatch between the curing reaction kinetics and the thermal conduction rate.

[0005] Existing technologies primarily focus on the mechanized combination of the front-end feeding system or the static execution of preset parameters, essentially falling under the category of open-loop control. Current mechanical adjustment methods struggle to perceive the complex dynamic environment within the vacuum chamber in real time, especially when filling pressure fluctuates or the liquid level experiences minute drifts, lacking a precise closed-loop adaptive adjustment mechanism. This lack of monitoring allows the epoxy resin composite to easily form micro-bubbles that are difficult to detect with the naked eye during the infiltration process, or to create filling dead zones at complex corners. This, in turn, can trigger partial discharge phenomena in the high-field-strength environment of lightweight transformers, becoming a key hidden danger restricting insulation reliability.

[0006] Furthermore, existing casting equipment typically treats parameters such as temperature, pressure, and flow rate as independent control variables in its process design, failing to establish a logical connection for multi-physics field synergistic optimization. In actual processing, if the current vacuum rate and injection pressure difference cannot be dynamically matched based on the rheological properties of the epoxy resin composition's viscosity changing with temperature, internal stress concentration can easily occur, leading to microcracks in the lightweight structure after curing. Summary of the Invention

[0007] The purpose of this invention is to provide an intelligent vacuum casting processing system suitable for lightweight transformers. This system solves the technical problems that, during the vacuum casting process of lightweight transformers, the lack of real-time coordinated closed-loop control of multiple physical field parameters leads to uneven penetration of epoxy resin composite material in the micro-winding gaps, deep micro-bubble residues, and concentrated residual stress during curing, which in turn limits the insulation reliability and mechanical integrity.

[0008] To solve the above-mentioned technical problems, the technical solution adopted by the present invention is as follows: An intelligent vacuum casting process system suitable for lightweight transformers includes: A vacuum sealing system includes a sealed vacuum chamber, a cascaded vacuum pump group, and an electromagnetic vacuum with an inflation valve. The inner wall of the sealed vacuum chamber is lined with a reflector plate, the surface emissivity of which is not less than 0.85, to enhance the uniformity of diffuse reflection of thermal radiation inside the vacuum chamber. The mixing and feeding system includes a resin storage tank, a curing agent storage tank, a servo gear metering pump, a static mixer, and a three-axis servo motion module at the end. The servo gear metering pump is driven by an AC servo motor. A multi-physics field monitoring sensor array includes an infrared thermal imaging sensor, an online rotational viscometer located at the outlet of a static mixer, a pressure sensor array located inside a mold, and a distributed optical fiber sensing network embedded inside a lightweight transformer winding assembly. The control unit integrates a signal acquisition module, a power drive module, a communication interface module, and an edge computing module for running a three-dimensional flow field digital twin model. The control unit is used to execute closed-loop control logic.

[0009] Furthermore, the system is used to perform the following steps: Step 1: Place the lightweight transformer winding assembly on the leveling support platform in the vacuum chamber, start the temperature control module to preheat the lightweight transformer winding assembly, and simultaneously start the vacuum pump group to reduce the absolute pressure in the vacuum chamber to the first pressure threshold and maintain the vacuum state, so as to achieve deep degassing in the micro gaps of the lightweight transformer winding assembly. Step 2: The dynamic viscosity of the mixed epoxy resin mixture is monitored in real time using an online rotational viscometer. The control unit adjusts the heating temperature of the static mixer according to the measured dynamic viscosity to keep the viscosity of the epoxy resin mixture within the target range before it enters the mold. Step 3: Open the injection valve and use a servo gear metering pump to inject the epoxy resin composite material into the mold. In the initial stage of injection, a constant volumetric flow rate is used. When the liquid level sensor detects that the liquid level has reached the bottom edge of the lightweight transformer winding assembly, the control unit switches to the pressure and volumetric flow rate collaborative feedback mode. The filling pressure is obtained in real time by multiple pressure sensors installed on the inner wall of the mold and the filling pressure gradient is calculated. Combined with the cross-sectional area of ​​the mold cavity corresponding to the current liquid level, the speed of the servo gear metering pump is dynamically adjusted so that the average flow velocity of the cross section flowing through the winding gap is maintained at the set macroscopic linear rise target speed of the liquid level, thereby keeping the linear rise speed of the liquid level in the complex cross-sectional flow channel constant. Step 4: During the filling process, the vacuum pump group controls the pressure in the vacuum chamber to alternate between a first pressure threshold and a second pressure threshold. The second pressure threshold is greater than the first pressure threshold. The pressure alternation effect drives the tiny bubbles remaining deep in the lightweight transformer winding assembly to drift to the liquid surface. At the same time, the infrared thermal imaging sensor monitors the temperature distribution on the outer wall of the mold, and the heating device set inside the vacuum chamber is adjusted to compensate the temperature field of the filling area in real time. Step 5: After the infusion is completed, the stepped curing stage begins. The first curing temperature is maintained for the first curing time, and then the temperature is increased to the second curing temperature at a controlled rate. During the curing process, the internal stress is monitored in real time by a distributed optical fiber sensor network embedded in the lightweight transformer winding assembly. If the stress time-varying rate is detected to exceed the set threshold, the control unit uses a local induction heating coil embedded in the outer wall of the mold to inductively heat the mold wall to adjust the temperature gradient of the mold wall. In turn, the local temperature gradient inside the epoxy resin composite is adjusted through heat conduction to reduce residual stress. After curing is completed, a controlled cooling procedure is executed.

[0010] Furthermore, the preheating temperature in step one is set to... to The preheating time shall not be less than 4 hours; the first pressure threshold shall not exceed [a certain value]. .

[0011] Furthermore, in step two, the target viscosity range of the epoxy resin composite material before it enters the mold is... to .

[0012] Furthermore, in step two, the control unit analyzes the flow activation energy of the epoxy resin composition using a rheological compensation model based on the measured dynamic viscosity. The mathematical expression of the rheological compensation model is as follows: in, Dynamic viscosity, unit: , Frequency factor, unit: , The activation energy of the flow is expressed in units of 1000 kJ / m². , Here is the gas constant, with a value of [value missing]. , Absolute temperature, unit: When solving for the activation energy of the flow, the control unit retrieves temporarily stored historical temperature-viscosity data pairs and performs least-squares regression according to the rheological compensation model to obtain the activation energy of this batch of materials. and Based on this, the target static mixer heating temperature required to maintain the viscosity within the target range is calculated.

[0013] Furthermore, in step two, the control unit has a built-in viscosity prediction model based on a multilayer sensor. The viscosity prediction model uses the ratio of epoxy resin composition, real-time temperature, and pressure difference between the inlet and outlet of the static mixer as input feature vectors to predict the viscosity change trend at future moments. When the predicted viscosity will exceed the target range, the control unit uses the predicted viscosity deviation as input and controls the heating unit of the static mixer through a PID closed-loop control loop to adjust the temperature of the heat transfer oil.

[0014] Furthermore, in step three, under the pressure and volumetric flow rate coordinated feedback mode, the model upon which the control unit dynamically adjusts the servo gear metering pump speed is based is: in, For volumetric flow rate that needs to be output in real time, the unit is... , For the set target velocity of macroscopic linear rise of the liquid level, unit , The current liquid level is obtained based on a three-dimensional flow field digital twin model or a pre-calibrated height-area function. The corresponding instantaneous cross-sectional area of ​​the mold cavity, in units , The dynamic permeability coefficient of the current level winding structure, in units of The flow is identified in real time by the three-dimensional seepage digital twin model built into the control unit. The dynamic viscosity of the epoxy resin composition, in units. , The average value of the filling pressure gradient across the flow channel cross-section, obtained through three-dimensional differential calculation using a pressure sensor array, is expressed in units of... The control unit adjusts... Make the average flow velocity of the cross section in the winding gap reach This allows the liquid level to rise steadily at a constant linear velocity.

[0015] Furthermore, in step four, the alternation frequency of the vacuum chamber pressure is controlled at... to .

[0016] Furthermore, in step four, the alternation frequency of the pressure alternation is dynamically adjusted based on the real-time dynamic viscosity of the epoxy resin composition. The optimal selection of the alternation frequency is based on the following correlation: in, The optimized pressure alternation frequency, in units of , The empirical damping coefficient is related to the mold structure and bubble morphology, and the unit is 1. , This represents the density difference between the gas phase inside the bubble and the liquid phase of the surrounding epoxy resin composite, expressed in units of... , Let gravitational acceleration be the acceleration due to gravity, and its value be [value]. , The defined characteristic diameter for targeted removal of microbubbles, in units of , The real-time dynamic viscosity of the epoxy resin composition is used in the calculation. Substitute the units.

[0017] Furthermore, in step four, the waveform of the pressure alternation is a modified triangular wave, and the pressure rise rate during the pressurization phase is not equal to the pressure fall rate during the depressurization phase.

[0018] Furthermore, the heating rate in step five, during the stepped curing stage, is controlled at... to The cooling rate of the controlled cooling process does not exceed .

[0019] Furthermore, in step five, the relationship between the center wavelength offset of the fiber optic sensor in the distributed fiber optic sensing network and stress and temperature changes follows the formula: in This is the initial center wavelength of the fiber optic sensor. This is the center wavelength offset. The effective photoelastic coefficient of the optical fiber. To eliminate the true microscopic axial strain after thermal deformation, The thermo-optic coefficient of the optical fiber material. The coefficient of thermal expansion of the optical fiber material. The local temperature change acquired by the temperature sensing elements arranged in the same location, in units of .

[0020] Furthermore, in step five, the threshold is set to... The threshold is the stress time-varying rate threshold.

[0021] Furthermore, the control unit is internally configured with a three-dimensional flow field digital twin model corresponding to the geometry of the lightweight transformer winding assembly; In step three, the control unit compares and calculates the filling pressure distribution acquired by multiple pressure sensors with the theoretical predicted pressure of the three-dimensional flow field digital twin model in real time. When the deviation rate between the measured filling pressure of a certain pressure sensor and the corresponding theoretical predicted pressure is calculated... Greater than the preset deviation threshold When the control unit determines that the internal flow channel of the lightweight transformer winding assembly is partially blocked, it instructs the heating device to perform pulsed instantaneous heating on the outer wall of the mold corresponding to the area of ​​partial blockage.

[0022] Furthermore, in step four, the controlled alternating pressure waveform is an asymmetric sawtooth wave, wherein the rise time of the asymmetric sawtooth wave is shorter than the fall time to form an uneven slope difference; the control unit dynamically adjusts the alternating frequency of the asymmetric sawtooth wave according to the real-time polymerization reaction process of the epoxy resin composition.

[0023] Furthermore, during the pressure alternation process in step four, the control unit acquires the absolute pressure of the sealed vacuum chamber and the characteristic temperature of the epoxy resin composite liquid surface captured by the infrared thermal imaging sensor in real time. When the absolute pressure of the sealed vacuum chamber drops to the corresponding saturated vapor pressure of the low-boiling-point component in the epoxy resin composition, and the characteristic temperature drop rate of the liquid surface exceeds the preset flash evaporation trigger threshold, it is determined that a surge in surface viscosity caused by the vacuum flash evaporation effect has occurred. At this time, the control unit activates the pulsed, specific wavelength infrared radiation source located at the top of the sealed vacuum chamber, with the output center wavelength located at... to A pulsed light beam within a certain range, the wavelength of which is selected to match the hydroxyl absorption band in the epoxy resin composition, performs targeted heating on the liquid surface with an extremely narrow penetration depth, thus confining the radiant energy below the liquid surface to no more than [amount missing]. Complete absorption within the shallow region, the timescale of a single pulse heating is forced to be less than the natural convection initiation time constant of the surface fluid, thereby establishing an inverted viscosity funnel in the vertical spatial dimension where the transient dynamic viscosity of the surface is lower than that of the deep bulk without inducing macroscopic convection mixing.

[0024] Furthermore, in the stepped curing stage of step five, the control unit solves the spatial curing gradient amplitude of the epoxy resin composite material inside the lightweight transformer winding assembly in real time by integrating the reaction kinetic equation coupled within the three-dimensional flow field digital twin model. The control unit incorporates an extended Kalman filter data assimilation loop. The system state vector of the extended Kalman filter data assimilation loop includes at least the curing degree and temperature of each voxel unit. The observation vector is composed of real stress growth data acquired in real time by the distributed optical fiber sensor network. Using the observation vector as the observation variable, the control unit periodically corrects the activation energy parameter in the reaction kinetic equation online to eliminate the calculation drift accumulated by the model over time. When the calibrated spatial curing gradient amplitude crosses the preset safety boundary limit, the control unit uses a model predictive control algorithm based on a multi-input multi-output system. It uses the heat transfer coupling matrix of adjacent thermal fields as a pre-estimated constraint, reverse-calculates and outputs the decoupled independent power pulse width modulation command to the multi-zone induction heating matrix arranged around the outer wall of the mold. The characterization equation for the spatial curing gradient amplitude is: in This represents the spatial curing degree gradient amplitude, in units of... , The current transient solidification degree of the mesh voxel element is a dimensionless parameter. , , The coordinate components of the corresponding mesh voxel element in the 3D physical coordinate system are given in units of 1. The control unit minimizes the global... To achieve optimal performance, the epoxy resin composite material is driven to maintain a constant degree of curing within the overall physical structure.

[0025] Compared with the prior art, the present invention has the following beneficial effects: This invention solves the problems of uneven penetration, deep bubble retention, and thermal stress imbalance caused by the fine gaps and compact structure of lightweight transformer windings through a closed-loop integration of five processes: coupled control of preheating and vacuum degassing, online dynamic viscosity adjustment of epoxy resin composites, gradient feedback injection of filling pressure, vacuum pressure alternating assisted degassing, and real-time reduction of curing stress. A multi-physics field monitoring sensor array and control unit work together to dynamically optimize the parameters of each process within the closed-loop circuit. The process no longer relies on fixed preset values, adapting to batch differences in materials and environmental disturbances, thus ensuring the consistency and stability of the processing quality of lightweight transformers in large-scale production.

[0026] In step two, the control unit uses a rheological compensation model to identify the flow activation energy of the epoxy resin composition in real time, locking the casting window along the main physical control path of temperature on viscosity. This prevents premature crosslinking reactions that could cause viscosity to rise and hinder the penetration endpoint. When the control unit incorporates a multilayer sensor viscosity prediction model, the control method is upgraded from passive feedback to active prediction. The static mixer heating temperature is adjusted in advance based on the viscosity trend, reducing overshoot and oscillation caused by control lag and enhancing the smoothness of viscosity locking throughout the process.

[0027] In step three, the control unit dynamically adjusts the speed of the servo gear metering pump based on the filling pressure gradient and Darcy's law correction model. At the same time, the change in the cross-sectional area of ​​the mold cavity is incorporated into the volumetric flow control loop, and the cross-sectional average flow velocity is calculated based on the cross-sectional average pressure gradient. This allows the epoxy resin composite liquid level to rise steadily at a constant linear speed. By pushing and filling the micro gaps at a constant propulsion speed, the interlayer turbulence and air entrapment induced by flow velocity fluctuations in conventional flow rate injection are suppressed from the flow mechanism perspective, reducing the probability of dead zones in the filling process.

[0028] In step four, the pressure alternation frequency is adjusted to match the real-time dynamic viscosity of the epoxy resin composition. This ensures that the bubbles receive effective volume expansion and contraction driving force within each pressure cycle, avoiding the problem of missed detachment windows due to delayed bubble response caused by increased viscosity at a fixed frequency. The alternating waveform uses a modified triangular wave or asymmetric sawtooth wave, employing unequal rise and fall rates to suppress secondary turbulent backflow and enhance unidirectional buoyancy impulse, thereby improving the bubble migration efficiency from deep layers to the liquid surface.

[0029] In step five, the control unit monitors the curing stress online through a distributed optical fiber sensor network. The stress time-varying rate exceeding the set threshold is used as an early warning criterion to trigger the local induction heating coil to adjust the temperature of the mold wall locally. The thermal relaxation effect of the polymer chain segments is used to transfer the internal friction of the constrained molecular chains, thereby reducing the risk of microcracks caused by uneven curing shrinkage in thin-walled lightweight structures from the source.

[0030] When the three-dimensional flow field digital twin model is used during the filling process to compare the deviation between the measured filling pressure and the theoretically predicted pressure and to determine blockages, the system gains the ability to diagnose and instantly recover from local abnormal states within the winding's micro-channels, further reducing underfill defects caused by sudden increases in local penetration resistance. The control unit performs online correction of the reaction kinetic equations through an extended Kalman filter data assimilation loop and employs a model predictive control algorithm to decouple adjacent thermal fields to smooth the global curing gradient. This proactively balances the reaction process misalignment caused by differences in heat dissipation in different regions of the lightweight structure, elevating the curing quality from passive acceptance to active balancing. Attached Figure Description

[0031] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of the present invention and should not be regarded as a limitation of the scope. For those skilled in the art, other related drawings can be obtained from these drawings without creative effort.

[0032] Figure 1 This is a flowchart of the intelligent vacuum casting process of the present invention.

[0033] Figure 2 This is a flowchart of the dynamic viscosity control sub-process of the present invention.

[0034] Figure 3 This is a flowchart of the local obstruction diagnosis and intervention branch of the present invention.

[0035] Figure 4 This is one of the operation interface diagrams of the system described in this invention.

[0036] Figure 5 This is the second diagram of the user interface of the system described in this invention.

[0037] Figure 6 This is the third diagram of the user interface of the system described in this invention. Detailed Implementation

[0038] In the following description, only certain exemplary embodiments are briefly described. As those skilled in the art will recognize, the described embodiments can be modified in various ways without departing from the spirit or scope of the embodiments of the invention. Therefore, the drawings and description are considered to be exemplary in nature and not restrictive.

[0039] The following is in conjunction with the appendix Figures 1-6 The embodiments of the present invention will be described in detail below.

[0040] Example 1: This example provides an intelligent vacuum casting processing system and method suitable for lightweight transformers. The system includes a vacuum sealing system, a mixing and feeding system, a multi-physics field monitoring sensor array, and a control unit.

[0041] The vacuum sealing system includes a sealed vacuum chamber, a cascaded vacuum pump assembly, and an electromagnetic vacuum system with an inflation valve. The effective volume of the sealed vacuum chamber meets the requirements for accommodating the lightweight transformer winding assembly to be cast, and the casing is made of corrosion-resistant metal material.

[0042] The inner wall of the sealed vacuum chamber is fully covered with a metal reflector plate with a roughened surface. The reflector plate has a surface emissivity of not less than 0.85, which is used to enhance the uniformity of diffuse reflection of thermal radiation inside the vacuum chamber, while protecting the sealed vacuum chamber body. This effectively distributes the radiant heat energy emitted by the heating device installed inside the sealed vacuum chamber evenly inside the cavity in the form of diffuse reflection.

[0043] The sealed vacuum chamber door employs a double-layer sealing structure, with an evacuation channel between the two sealing rings, forming a two-stage vacuum sealing structure. The cascaded vacuum pump unit consists of a first-stage rotary vane vacuum pump and a second-stage Roots pump connected in series. The first-stage two-stage rotary vane pump is responsible for rapidly traversing the molecular flow stage from atmospheric pressure to low vacuum, while the second-stage Roots pump... to The transition zone provides pumping conduction capability, and the pumping speed configuration ensures that... The pressure inside the sealed vacuum chamber is reduced from atmospheric pressure to... the following.

[0044] An electromagnetic vacuum valve with inflation is installed on the vacuum exhaust pipe. Its flow resistance changes with high linearity according to the input drive voltage, and the step response time from opening to closing is in the millisecond range, enabling precise pressure alternation control. The sealed vacuum chamber is equipped with a multi-degree-of-freedom leveling platform. This platform has three-axis leveling capabilities and a three-axis servo drive actuator array at its bottom. Combined with dual-axis tilt sensors arranged diagonally on the platform, it forms a closed-loop spatial attitude correction system with a positioning accuracy of [insert accuracy here]. This ensures that the angle deviation between the geometric center axis of the lightweight transformer winding assembly and the normal to the horizontal plane approaches zero.

[0045] The temperature control module is located inside the sealed vacuum chamber and is used for preheating and temperature regulation of the lightweight transformer winding assembly. Far-infrared heating plates are installed on the inner side wall of the sealed vacuum chamber to compensate for heat radiation in the injection area.

[0046] The mixing and feeding system includes a resin storage tank, a curing agent storage tank, a servo gear metering pump, a static mixer, and a three-axis servo motion module at the end. Both the resin and curing agent storage tanks are equipped with low-speed stirring devices to maintain material homogeneity within the tanks. The servo gear metering pump is driven by a high-resolution AC servo motor with an embedded encoder to eliminate hysteresis volumetric errors caused by transmission backlash, and the rated flow error is controlled within [specific parameters]. Within [a certain range]. The static mixer is internally equipped with multi-stage spiral mixing elements arranged in opposite directions, achieving a mixing uniformity variation coefficient of less than 1% under low-speed laminar flow conditions at low Reynolds numbers. A heat transfer oil circulation coil is tightly wound around the outer circumference of the static mixer's tube wall. The coil's flow path employs a counter-current heat exchange design, and the heat transfer oil circulates through an external high-precision thermostatic bath, with a temperature control range from room temperature to [a certain value]. The effective stroke of the end effector three-axis servo motion module covers the entire mold operating space, with a positioning accuracy of [missing information]. Its end is equipped with a filling nozzle, which has a quick-opening and closing control valve to achieve... Opening and closing control within the specified range.

[0047] The multiphysics monitoring sensor array includes an infrared thermal imaging sensor, an online rotational viscometer, a pressure sensor array, a distributed fiber optic sensor network, and a liquid level sensor. The infrared thermal imaging sensor is located at the top of the sealed vacuum chamber, and its temperature measurement range covers [missing information]. to It is used to monitor the temperature distribution on the outer wall of the mold in real time, compare it with the preset safety isotherm model, and perform thermal radiation compensation positioning for corner areas where heat dissipates quickly. The measuring head of the online rotary viscometer is immersed at the outlet of the static mixer, and the measurement accuracy deviation is no greater than [value missing]. The sampling frequency is .

[0048] The pressure sensor array consists of piezoelectric thin-film pressure sensors, which are laid flat and cured onto the stress surfaces of the mold's inner wall using a high-temperature adhesive. These sensors are distributed in an array on the inner side of the mold to acquire the filling pressure and its spatial gradient. The distributed fiber optic sensing network includes sensing fibers and a fiber optic demodulator pre-embedded within the lightweight transformer winding assembly. The sensing fibers are pre-embedded during the interlayer insulation paper binding stage of the winding assembly in the end insulation blocks and inter-turn regions where the electric field distribution is dense and mechanical stress is concentrated. The fiber optic demodulator has a wavelength resolution superior to... The sampling frequency is not less than A liquid level sensor is installed at the bottom of the mold to detect the liquid level of the epoxy resin mixture.

[0049] The control unit employs an industrial-grade real-time control kernel, integrating a signal acquisition module, a power drive module, a communication interface module, and an edge computing module for running a three-dimensional flow field digital twin model. The control unit's software architecture is divided into three layers: a bottom-layer signal conditioning layer, used for high-precision analog-to-digital conversion and filtering of pressure, temperature, and viscosity analog signals, employing hardware-level low-pass filtering and synchronous analog-to-digital conversion to suppress electromagnetic interference in industrial environments; a middle-layer logic control layer, responsible for executing closed-loop regulation, pressure alternation logic switching, and safety interlock protection; and a top-layer expert decision-making layer, which establishes a virtual simulation model synchronized with the physical casting process based on digital twin technology, using intelligent optimization algorithms to perform real-time optimization searches for process parameter combinations, while simultaneously running a flow field deduction model based on fluid dynamics equations.

[0050] This embodiment also provides an intelligent vacuum casting process using the above system, which includes the following steps.

[0051] Step 1: Pre-treatment of the component to be cast and initial construction of the vacuum chamber environment.

[0052] The lightweight transformer winding assembly is precisely installed on a multi-degree-of-freedom leveling platform within a sealed vacuum chamber. The three-axis servo-driven actuator array at the bottom of the platform independently extends and retracts based on tilt sensor feedback, performing three-dimensional attitude reverse compensation on the support plane to eliminate the pouring height difference caused by installation. The temperature control module is then activated to preheat the lightweight transformer winding assembly; the preheating temperature is set to [temperature value missing]. Preheating time remains The infrared radiation and convective heat conduction effect are used to remove adsorbed water from the surface and inside of the insulation material of the lightweight transformer winding components. The water molecules adsorbed in the fiber capillaries inside the insulation paper are violently vaporized and broken free from hydrogen bond binding under the dual action of thermodynamics and vacuum fluid dynamics.

[0053] The cascaded vacuum pump unit is started simultaneously to reduce the absolute pressure in the sealed vacuum chamber to the first pressure threshold. As the pressure drops Then, maintain the vacuum state for a certain duration. This is to achieve deep degassing within the fine gaps of lightweight transformer winding components.

[0054] Step 2: Monitoring and control of dynamic rheological properties of epoxy resin composites.

[0055] The epoxy resin mixture is precisely metered by a servo gear metering pump and then enters a static mixer for mixing. Upon mixing, the cross-linking reaction begins immediately, with the molecular chain length continuously increasing, leading to a surge in intermolecular physical entanglement and frictional resistance. This is macroscopically manifested as a non-linear increase in viscosity over time. An online rotational viscometer collects the dynamic viscosity of the epoxy resin mixture at the static mixer outlet in real time. Based on the measured dynamic viscosity, the control unit analyzes the flow activation energy of the current batch of epoxy resin mixture using a rheological compensation model based on the Arrhenius equation. The Arrhenius equation is in the form of: in, Dynamic viscosity, unit: , Frequency factor, unit: , The activation energy of the flow is expressed in units of 1000 kJ / m². , Here is the gas constant, with a value of [value missing]. , Absolute temperature, unit: .

[0056] When calculating the activation energy of the flow, the control unit retrieves temporarily stored historical temperature-viscosity data pairs and performs online least squares regression according to the rheological compensation model to obtain the activation energy of the current batch of materials. and Based on this, the target static mixer heating temperature required to maintain the viscosity within the target range is calculated.

[0057] The control unit incorporates a viscosity prediction model based on a multilayer sensor. This model uses the epoxy resin composition ratio, real-time temperature, and static mixer inlet and outlet pressure difference as input feature vectors to predict future viscosity. The viscosity variation trend within the multilayer sensor. The mathematical expression of the viscosity prediction model is: in, For predicted future moments Viscosity value, in units of , For the current moment The input feature vector includes the epoxy resin composition ratio, real-time temperature, and static mixer inlet and outlet pressure difference. and These are the weight matrix and bias vector of the first hidden layer, respectively. and These are the weight matrix and bias vector of the output layer, respectively. Nonlinear activation function The output layer is a linear mapping function. The multilayer perceptron consists of one input layer, two hidden layers, and one output layer. Each hidden layer has 64 neurons, and the training loss function is mean squared error.

[0058] The multilayer sensor viscosity prediction model was trained with no fewer than 200 sets of actual casting historical data, and the coefficient of determination between the predicted and measured values ​​was no less than 0.95. When the model predicts viscosity exceeding... to When the target viscosity range is reached, the control unit uses the predicted viscosity deviation as input and controls the heating unit outside the static mixer through a PID closed-loop control loop. This adjusts the temperature of the heat transfer oil to change the temperature of the epoxy resin mixture, reducing the internal energy of the system by enhancing heat exchange. Physical cooling effectively slows down the chemical reaction rate, thus dynamically locking the viscosity of the epoxy resin mixture within the target range before it enters the mold. Temperature control accuracy reaches [a certain level]. Response time is less than .

[0059] Step 3: Adaptive segmented injection control based on flow feedback.

[0060] The filling valve is opened, and the end-effector triaxial servo motion module drives the filling nozzle to move to the preset filling position. During the initial filling stage, to prevent the high-energy epoxy resin composite liquid column from impacting and damaging the fragile underlying insulation, the servo gear metering pump outputs a constant small flow rate. When the liquid level sensor detects that the liquid level completely covers the bottom support structure of the lightweight transformer winding assembly, the control unit switches to the pressure and volumetric flow rate coordinated feedback mode.

[0061] At this point, the pressure sensor array begins to capture the filling pressure data, which is the composite of the hydrostatic pressure at the liquid surface and the flow resistance. This data is processed by a charge amplifier and then transmitted to the control unit. The control unit performs a gridded finite difference operation on the pressure field in discrete space to solve for the transient pressure gradient tensor distributed in three-dimensional space at the flow front, and calculates the average pressure gradient on the flow channel cross-section. The modified control model based on Darcy's law is as follows: in, For volumetric flow rate that needs to be output in real time, In this embodiment, the target velocity for the macroscopic linear rise of the liquid level is set. Set as , To determine based on the current liquid level The instantaneous cross-sectional area of ​​the cavity inside the mold is obtained from a pre-defined height-area function. The dynamic permeation coefficient of the current layer winding structure is determined in real time by the three-dimensional permeation digital twin model built into the control unit. This refers to the dynamic viscosity of the epoxy resin composition. The average pressure gradient across the cross section.

[0062] The difference in winding tightness between different layers of the lightweight transformer winding assembly causes spatial variations in the local permeability coefficient. The control unit solves the above model in real time, utilizing the equivalent relationship between the cross-sectional average flow velocity and the target linear velocity, to dynamically adjust the speed of the servo gear metering pump, ensuring that the liquid level rise velocity is consistently maintained at a certain level. The set value, the fluctuation error is controlled within Within this range. The steady, constant-speed upward movement creates a wavefront effect that effectively eliminates air bubbles trapped between layers, ensuring uniform impregnation of each micropore by the epoxy resin composite and preventing turbulent air entrapment.

[0063] Step 4: Cooperative optimization of vacuum pressure alternation under multi-field coupling.

[0064] The height of the grouting fluid level reaches the overall height of the lightweight transformer winding assembly. , At the three points of completion of infusion, the control unit triggers the pressure pulsation program. High-purity, dry nitrogen is intermittently injected into the sealed vacuum chamber using an electromagnetic vacuum valve; the nitrogen purity is not lower than [a certain level]. This causes the pressure inside the sealed vacuum chamber to reach the first pressure threshold. With the second pressure threshold The pressure alternates in a controlled, modified triangular wave pattern. The pressure alternation frequency is set to... .

[0065] The optimal selection of the pressure alternation frequency is based on the following correlation: in, In this embodiment, the empirical damping coefficient is related to the mold structure and bubble morphology. Pick , Pick , Set as , This represents the real-time dynamic viscosity of the epoxy resin composition.

[0066] The tiny air bubbles hidden deep within the epoxy resin composite are initially held in a static state by the viscous resistance of the surrounding high-viscosity fluid and the capillary binding force of the microscopic gaps. When the external environmental pressure rapidly decreases along the descending edge of a triangular wave, the gas pressure inside the bubble is much greater than the external pressure, causing the bubble to expand. This dramatic increase in volume leads to a significant increase in the buoyancy of the bubble, breaking through the constraints of the fluid's viscous damping. Conversely, when the environmental pressure rises again, the bubble is forcibly compressed, allowing it to pass through extremely small structural gaps that were previously impassable. This microscopic breathing effect, based on physical expansion and contraction, gives the bubbles the kinetic energy to escape.

[0067] Infrared thermal imaging sensors monitor the heat distribution on the outer wall of the mold in real time and generate a thermal field map of the mold's outer wall. The control unit adjusts the radiation intensity of the far-infrared heating plate based on the temperature distribution data, and provides real-time thermal compensation for any localized temperature drops that may occur due to nitrogen injection, ensuring that temperature fluctuations on the mold surface do not exceed [a certain threshold]. .

[0068] Step 5: Gradient temperature-controlled curing and online stress monitoring.

[0069] After the infusion is completed, the stepped curing stage begins. This embodiment uses a three-stage stepped curing curve: the first stage raises the ambient temperature to the first curing temperature. and maintain the first curing time. At this stage, the epoxy resin composite enters the initial gelation phase, and the chemical shrinkage rate begins to increase. The second stage... The heating rate increased to ,Keep Entering the deep cross-linking period. The third stage is... The heating rate is increased to the second curing temperature. ,Keep High-temperature setting treatment is performed to eliminate residual stress within the molecular chains.

[0070] Throughout the curing process, the fiber optic sensors in the distributed fiber optic sensing network transmit the center wavelength shift signal back in real time via a fiber optic demodulator. In this embodiment, the distributed fiber optic sensing network specifically employs a fiber Bragg grating (FBG) sensor array. The fiber optic sensors convert locally occurring micron-level tensile or compressive deformation into a frequency domain shift of the backscattered light wavelength. The relationship between the center wavelength shift of the fiber optic sensor and stress and temperature changes follows the following formula: in, The initial center wavelength of the fiber optic sensor is used in this embodiment. Pick , This is the center wavelength offset. Let be the effective photoelastic coefficient of the optical fiber, which is 0.22 for silica optical fiber. To eliminate the true microscopic axial strain after thermal deformation, Let be the thermo-optic coefficient of the optical fiber material, and take . , Let be the coefficient of thermal expansion of the optical fiber material, and take . , The local temperature change acquired by the temperature sensing elements arranged in the same location, in units of .

[0071] The control unit eliminates temperature interference using the aforementioned decoupling formula, accurately extracts the micro-strain caused by curing shrinkage, and uses Hooke's law to convert the actual micro-axial strain into the internal curing stress distribution. If the fiber optic demodulator detects that the stress time-varying rate in a local area exceeds a set threshold... This indicates a risk of excessively rapid curing or severely uneven shrinkage in the area. At this point, the control unit immediately activates the local induction heating coil embedded in the outer wall of the mold to induction heat the mold wall. Through heat conduction within the mold wall, the local temperature gradient is adjusted, reducing the relaxation time of the polymer chain segments and inducing relative slippage of the constrained polymer chains, thereby achieving active stress reduction.

[0072] After curing, the system executes a controlled cooling program. Using a circulating fan and temperature control module, the cooling rate is locked at a specific level. Until the residual stress detected by the fiber optic sensor stabilizes and the ambient temperature drops to... Only then can the sealed vacuum chamber door be opened to remove the product.

[0073] Example 2: Example 2 is a further optimization based on Example 1, specifically: the parameters in the pressure alternation synergistic optimization process are further limited and optimized.

[0074] In this embodiment, the pressure alternation frequency is not fixed. Instead, it is dynamically adjusted based on the real-time dynamic viscosity of the epoxy resin blend.

[0075] In practice, after the viscosity monitored in real time by the online rotational viscometer is input into the control unit, the control unit operates according to the following correlation formula: Real-time calculation of optimized pressure alternation frequency .

[0076] in, Pick , Pick , Pick .

[0077] During the injection process, when the viscosity is... to When within the interval, the corresponding for to , The calculation results are in to Between; when viscosity increases to to When within the interval, the corresponding for to , The calculation results are in to Between. The control unit calculates in real time... The pressure alternation frequency is dynamically adjusted so that the pressure pulsation is always within the optimal defoaming efficiency range.

[0078] Furthermore, this embodiment optimizes the parameters of the corrected triangular wave waveform for pressure alternation. The pressure rise rate during the pressurization phase is controlled as follows: The rate of pressure decrease during the depressurization phase is controlled as follows: The asymmetry of the waveform causes the bubbles to exhibit a slow energy accumulation followed by an instantaneous burst in a single cycle. The controlled rate pressurization process during the pressurization phase prevents secondary turbulent backflow caused by large-span pressure surges in the fluid within the complex micro-gap. The intense negative pressure tearing effect during the depressurization phase instantly breaks the balance of surface tension on the bubble wall, providing an extremely strong unidirectional buoyancy impulse, enabling the overflow efficiency of the microbubbles to reach the physical limit.

[0079] The other steps and system configurations in this embodiment are the same as in Embodiment 1, and will not be repeated here.

[0080] Example 3: Example 3 provides an alternative stress monitoring and control scheme, demonstrating an alternative front-end material thermal control strategy, as detailed below: In selecting the stress monitoring terminal, this embodiment uses a resistance strain gauge array instead of a fiber optic sensor system. The resistance strain gauge array comprises multiple high-precision foil resistance strain gauges, using polyimide as the substrate, and is tightly bonded in a high-density matrix to key stress points in the lightweight transformer winding assembly, such as the shear-bearing area at the corner of the interlayer support insulation cylinder. The resistance strain gauges convert mechanical strain into voltage signals via a Wheatstone bridge circuit. These signals are amplified and filtered by a dynamic strain gauge before being transmitted to the control unit.

[0081] The relationship between strain and stress satisfies the extended model of Hooke's Law: in, Stress, unit: , The elastic modulus of the epoxy resin composition at the corresponding curing stage is determined by a pre-calibrated curing kinetic model. In response to the situation.

[0082] The strain data acquired by the resistance strain gauge array is also used to trigger the intervention action of the local induction heating coil. When the monitored stress time-varying rate exceeds... When the threshold is reached, the control unit activates the corresponding local induction heating coil to induction heat the mold wall and adjust the local temperature gradient to reduce stress concentration.

[0083] Alternatively, this embodiment can also employ a Brillouin optical time-domain reflectometry (BDR) system in the distributed optical fiber sensing network. When using a Brillouin BDR system as the distributed optical fiber sensing network, the frequency shift signal of the Brillouin scattered light is captured by injecting pulsed pump light into the sensing fiber, and the axial micro-strain distortion points caused by the local volume change of the cured epoxy resin composite are extracted. This scheme differs from the fiber Bragg grating (FBG) sensing scheme used in Embodiment 1 above in terms of sensing principle, but both can achieve online monitoring of stress state.

[0084] Regarding the front-end material thermal control strategy, this embodiment involves directly and tightly winding an induction coil around the outer wall of the static mixer's flow tube. When the control unit outputs a high-frequency pulse width modulation command to the high-frequency switching inverter power supply, the alternating magnetic field directly penetrates the mixer tube wall. Based on the principles of electromagnetic induction and skin effect, Joule heating is instantly induced on the extremely thin surface layer of the inner wall of the metal tube. The heat exchange distance between the fluid and the heating surface is shortened to the level of direct contact. This energy infusion method based on a direct electric field significantly reduces the control response delay time for local heating and cooling. Combined with a pre-set viscosity prediction model, precise locking of the rheological properties of the epoxy resin composite material entering the mold cavity is achieved.

[0085] The other steps and system configurations in this embodiment are the same as in Embodiment 1, and will not be repeated here.

[0086] Comparative Example 1: To verify the beneficial effects of the technical solution of the present invention, a comparative experiment was conducted between Comparative Example 1 and the above embodiments. Comparative Example 1 adopted a conventional vacuum casting process, and the specific technical solution is as follows.

[0087] The epoxy resin compound formulation used in Comparative Example 1 was the same as that in Example 1, and the lightweight transformer winding assembly model was the same as that in Example 1. The preheating temperature during the pretreatment stage was also set to [temperature value missing]. Preheating The pressure inside the sealed vacuum chamber drops to and maintain .

[0088] During the infusion stage, comparative example 1 did not undergo online viscosity closed-loop control of the epoxy resin composition. The jacket temperature of the static mixer was fixedly set via a constant temperature bath. The viscosity of the epoxy resin composite increases naturally with reaction time, without any dynamic adjustment throughout the process. The injection process employs a constant flow mode, with the gear metering pump output flow rate fixed at [value missing]. No pressure feedback mode switching. Pressure is maintained within the sealed vacuum chamber throughout the entire infusion process. Constant high vacuum, no pressure alternation steps.

[0089] During the curing stage, Comparative Example 1 was executed according to a conventional stepped temperature control curve: Heat up to ,Keep ;by Heat up to ,Keep Throughout the entire curing process, there is no online stress monitoring or localized induction heating intervention. After curing, [the product / process is as follows]. It cools naturally to room temperature at a rate that allows it to pass through naturally.

[0090] After the experiment, internal defects were quantitatively detected by X-ray tomography, and partial discharge and thermal cycling shock tests were performed. Ten identical samples were used for parallel testing in each embodiment and comparative example. Specific performance comparison data are shown in Table 1.

[0091] Table 1. Performance Comparison of Examples and Comparative Examples; As can be clearly observed from the quantitative data in Table 1, Embodiments 1 and 2 of the present invention exhibit significant advantages in the processing quality of lightweight transformers. Embodiment 1, through the implementation of rheological property control based on a multilayer sensor viscosity prediction model, achieves a resin filling rate of [missing information]. Example 2 further increases the filling rate by dynamically optimizing the pressure alternation frequency. In contrast, Comparative Example 1, lacking dynamic control, experienced a gradual increase in viscosity over time, leading to a decrease in penetration capacity and a filling rate of only [missing information]. .

[0092] Regarding bubble control, the microbubble density in Example 1 was [insert value here]. Example 2 further reduced to This almost completely eliminates harmful cavities. This is directly reflected in the partial discharge level; the partial discharge level of the product described in this invention is controlled at... Within this range, it far surpasses industry standards, while Comparative Example 1, due to the presence of deep residual bubbles, has a partial discharge level that reaches [missing value]. There is a risk of breakdown during long-term operation.

[0093] Regarding stress control, Examples 1 and 2 utilize online monitoring via a distributed fiber optic sensor network and coordinated regulation with local induction heating to reduce the maximum residual curing stress from that of Comparative Example 1. Reduced to and This is crucial for lightweight structures with relatively low mechanical strength and thin walls. The results of thermal shock testing further validate the outstanding effect of this invention in improving the long-term mechanical reliability of products.

[0094] Furthermore, the solution of this invention reduces the production cycle from shortened to approximately Efficiency improvement exceeds This also ensures higher product quality consistency.

[0095] Example 4: This example, based on Example 2, focuses on the extremely narrow gap (less than) required for ultra-high power density. The vacuum casting process for lightweight transformer winding assemblies is explained in detail below: Preparation stage. A three-dimensional digital twin model of the flow field of the lightweight transformer winding assembly is pre-built within the control unit. The system imports the CAD geometric data of the lightweight transformer winding assembly and generates a node mesh containing the fluid domain using hexahedral mesh generation technology.

[0096] The three-dimensional flow field digital twin model treats the winding region as an anisotropic porous medium model. The local permeability tensor is discretized and assigned values ​​based on the arrangement density of the winding conductors and the layer distribution of the insulating paper. The non-Newtonian rheological constitutive equations of the epoxy resin composition at different temperatures and curing degrees are input into the three-dimensional flow field digital twin model.

[0097] In step one, the pre-dried lightweight transformer winding assembly is hoisted and fixed onto the leveling support platform within the vacuum chamber. Each of the four support corners of the leveling support platform is independently connected to a piezoelectric force sensor and a precision servo hydraulic cylinder. The control unit dynamically calculates the support force data and manipulates the precision servo hydraulic cylinders to control the perpendicularity error of the lightweight transformer winding assembly within a specified range. Within [a certain range]. The temperature control module starts, performing radiative heat transfer to the lightweight transformer winding assembly. The preheating rate is set to [a certain value]. The target preheating temperature is set to The vacuum pump unit starts synchronously; it consists of a Roots pump and a rotary vane pump connected in series. The absolute pressure inside the vacuum chamber is... Internal drop The first pressure threshold state.

[0098] Vacuum chamber in First pressure threshold state maintenance During this period, the dielectric barrier discharge plasma generator installed at the bottom of the leveling platform is activated. A trace amount of argon-oxygen mixed gas is introduced into the vacuum chamber, with the partial pressure controlled at... A dielectric barrier discharge plasma generator excites and produces low-temperature plasma, which continuously saturates the surface of the lightweight transformer winding assembly. In-situ plasma interface activation treatment. The water droplet contact angle on the surface of the winding insulation material is reduced from that before treatment. Reduce to .

[0099] In step two, bisphenol A epoxy resin, anhydride curing agent, and accelerator are dispensed via their respective precision gear pumps according to... The mass ratio is continuously fed to the static mixer. The static mixer is internally equipped with... A staggered spiral mixing unit. An online rotational viscometer is connected in series on the thermostatic pipeline between the static mixer outlet and the filling valve. The control unit uses a PID algorithm with reaction kinetics feedforward compensation for dynamic adjustment. The dynamic viscosity measured by the online rotational viscometer is... Data is fed back to the control unit in real time. The control unit calculates the target heating temperature of the circulating heat transfer oil in the static mixer jacket. The control unit incorporates a Kamal autocatalytic kinetic model to calculate the instantaneous exothermic rate. Based on the principle of thermal balance, it converts the calculated instantaneous exothermic rate into an equivalent temperature rise rate, which is then used as a negative temperature compensation factor and directly added to the PID control command. This allows for earlier reduction of heating power to offset the impact of the exothermic reaction on the material temperature. Through temperature compensation control, the dynamic viscosity of the epoxy resin composite material before entering the mold is limited to [specific value missing]. Within the target range.

[0100] Step three is executed, and the injection valve is opened. During the initial injection phase, the control unit instructs the servo gear metering pump to... A constant flow rate injects the epoxy resin composite material from the bottom interface of the mold. A level sensor continuously monitors the liquid level inside the mold. When the liquid level reaches the lowest edge of the lightweight transformer winding assembly, the control unit switches from the constant flow mode to a pressure and volumetric flow rate coordinated feedback mode. The array is arranged at different elevations on the inner wall of the mold. A thin-film pressure sensor, with The sampling frequency is used to send the measured filling pressure data back to the control unit.

[0101] The control unit calculates the theoretically predicted pressure at each coordinate node under the current liquid level condition in real time. The control unit then compares the measured filling pressure fed back by each pressure sensor with the theoretically predicted pressure output by the three-dimensional flow field digital twin model. The filling process continues until... At that time, the pressure sensor located on the left side of the middle section of the mold reported that the measured filling pressure had increased to [a certain value]. The theoretical predicted pressure at this point in the three-dimensional flow field digital twin model is: The control unit performs calculations to obtain the deviation rate between the two. In this embodiment, a preset deviation threshold is used. Set as Due to deviation rate The control unit determines that the flow channel in the corresponding area is partially blocked.

[0102] The control unit then instructs the heating device inside the vacuum chamber to activate, activating a short-wave infrared spotlight at a specific location to focus the infrared beam onto the outer wall of the mold corresponding to the area of ​​localized blockage. The surface temperature of this localized area of ​​the mold's outer wall is... Internally Rise to Simultaneously, the control unit adjusts the propulsion parameters of the servo gear metering pump, adding an amplitude of [missing value] to the original base speed. , frequency is High-frequency micro-flow pulses. The measured filling pressure fed back by the pressure sensor is applied during the micro-flow pulse. Later it fell back to The infusion process continues.

[0103] In step four, the injection fluid level flows upward through the lightweight transformer winding assembly. The control unit manipulates the proportional servo leakage valve on the vacuum pump assembly to create a controlled pressure alternating field with an asymmetric sawtooth waveform within the vacuum chamber. Each pressure alternating cycle lasts... Within a single alternating cycle, the absolute pressure in the vacuum chamber changes from... The first pressure threshold rises to The second pressure threshold, the time required for pressure boosting This constitutes a short, rapid rise edge. The absolute pressure of the vacuum chamber is... Descent Voltage reduction time A slow, prolonged falling edge is formed. A rapid rising edge forcibly compresses the volume of the tiny bubble to sever its surface tension anchor point with the insulating paper wall, while a slow falling edge causes the bubble to expand slowly and drift towards the liquid surface due to the buoyancy generated by the increased volume. The control unit receives integrated data from the liquid level sensor, every [time period missing]. Increase the alternating frequency of the asymmetric sawtooth waveform. .

[0104] Proceeding to step five, the resin level completely covers the top of the lightweight transformer winding assembly, and the control unit closes the injection valve. A spirally distributed fiber Bragg grating sensor array is pre-embedded between the insulation layers of the high-voltage and low-voltage coils of the lightweight transformer winding assembly. The control unit receives the center wavelength drift data of the fiber Bragg grating sensor array in real time, calculates and outputs the internal three-dimensional stress time-varying rate. The curing temperature control curve is set as follows: constant temperature Subsequently The rate climbed to and maintain constant temperature .

[0105] At temperature Towards During the ascent, the fiber Bragg grating sensor array detected that the local stress time-varying rate at the coil end reached [value missing]. Because this value exceeds the set safety stress time-varying rate threshold. The control unit interrupts the heating command, and the temperature control module enters an adaptive heat preservation state at the current instantaneous temperature node. The isothermal relaxation and the control unit monitored the time-varying rate of stress. The value dropped to The control unit releases the heat preservation restriction and resumes heating at the original rate, executing the remaining heating and temperature control program.

[0106] After the curing and adaptive insulation processes are completed, the control unit initiates the controlled cooling process to... The rate at which the temperature drops to Only then can the sealed vacuum chamber door be opened to remove the product. This completes the curing cycle of the lightweight transformer.

[0107] Example 5: Based on Example 4, this example provides a deep-level intervention strategy that focuses on reshaping the spatial viscosity distribution and digital twin closed-loop calibration to address the interface closure caused by surface flash evaporation induced by high vacuum suction and the runaway curing reaction caused by irregular and complex windings.

[0108] In practice, during the controlled pressure alternation phase of step four, the absolute pressure inside the vacuum chamber drops in a sawtooth pattern. When the control unit detects the absolute pressure dropping to... Furthermore, the infrared thermal imaging sensor reported that the temperature drop rate of the liquid surface suddenly reached [a certain value]. At that time, the system determined that the current environment had exceeded the saturated vapor pressure limit of the reactive diluent in the epoxy resin mixture. Intense vacuum flash evaporation caused the latent heat on the surface to be rapidly removed, leaving the surface layer... The dynamic viscosity of the fluid within the thickness jumps to approximately [value missing] in a very short time. This forms a rigid, gel-like shell that prevents internal bubbles from escaping. If conventional continuous thermal radiation is used, Rayleigh-Benard convection within the fluid will inevitably occur, rendering attempts to reshape the vertical gradient ineffective.

[0109] The control unit instantaneously triggers the top-mounted pulsed infrared radiation source of a specific wavelength at the rising edge of the pressure alternation (during the vacuum breaking phase, when flash evaporation is suppressed). The center wavelength of this radiation source is precisely locked to... The mid-wave infrared pulse, the wavelength of which is located within the hydroxyl absorption band of the epoxy resin composition ( to ).

[0110] High-frequency pulsed light beams are directed at the liquid surface with millisecond-level pulse widths, and their physical penetration depth is strictly attenuated and limited to below the liquid surface. Within the very shallow layer.

[0111] Due to the heat energy infusion time of a single pulse (approximately) The time constant for initiation of natural convection due to density difference in a fluid is much smaller than the initiation time constant (usually on the order of seconds or more), forcibly locking the heat to the surface. The surface temperature of the liquid rises instantly, and the dynamic viscosity drops sharply. The deeper substance remains in At the physical level, a stable system that can be maintained for approximately [time period missing] was successfully constructed. A transient inverted viscosity funnel. Within this extremely narrow time window, the clusters of tiny bubbles rupture and overflow densely, as if piercing a sieve, under the influence of pressure fluctuations.

[0112] With the pouring completed, the processing flow enters the step-by-step curing stage in step five.

[0113] To address the problem of extremely uneven heat dissipation conditions inside the winding assembly of lightweight transformers, the control unit relies on a three-dimensional flow field digital twin model for feedforward extrapolation.

[0114] The system activated the Extended Kalman Filter (EKF) data assimilation loop. During the temperature ramp-up period, the distributed fiber optic sensor network... The internal axial stress data is acquired in real time at a specific frequency. The control unit uses the measured stress growth rate as a physical observation variable and injects it into the digital twin model. By minimizing the residual between the observed values ​​and the theoretical predictions of the model, the system performs this acquisition every [period]. An iteration of the Kalman gain matrix is ​​performed to inversely update the apparent activation energy and pre-exponential factor in the reaction kinetic equations. This mechanism, which strongly binds computational power to the real physical state, ensures the transient solidification of the mesh voxel cells. It is no longer a theoretical value detached from reality.

[0115] When the calibrated three-dimensional computational field feedbacks the spatial solidification gradient amplitude between the thin-walled edge region and the thick core region... Approaching the setting When the safety boundary is reached, the control unit activates the multi-zone induction heating matrix around the outer wall of the mold.

[0116] Faced with the extremely strong thermal diffusion coupling interference between adjacent coils, the system abandons the traditional single-loop PID control and instead adopts a model predictive control (MPC) algorithm. Within the prediction horizon, the control unit transforms the three-dimensional Laplace heat conduction equation into a multivariable system state-space model, substituting the cross-heat transfer matrix of adjacent thermal fields as a hard constraint into the quadratic programming solver. Within a few hundred milliseconds, the system reverse-engineers a globally optimal set of control vectors and issues completely decoupled independent power pulse width modulation (PWM) commands to the induction coils in different zones. This decoupling strategy of coordinated oil extraction and water injection forcibly bridges the reaction process misalignment caused by volume differences, ultimately ensuring the global spatial solidification gradient amplitude... Suppressed steadily fluctuation.

[0117] After production using the special process described in this embodiment, dissection and section analysis were performed. Subsurface The number of microbubbles remaining in the blind zone is reduced by that of the conventional alternating scheme. Complete zeroing; more importantly, because the curing process achieves true iso-wavefront propagation, the peak residual shear stress at the interface between the ultra-thin insulation layer and the thick conductor is reduced. The structural thermal shock peeling cycle life has been improved to [the original value]. It is twice as strong, fully meeting the stringent insulation physical specifications for aerospace-grade lightweight transformers.

[0118] Although preferred embodiments of the invention have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including both the preferred embodiments and all changes and modifications falling within the scope of the invention.

[0119] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. It should be noted that any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. An intelligent vacuum casting processing system suitable for lightweight transformers, characterized in that, include: The vacuum sealing system includes a sealed vacuum chamber, a cascaded vacuum pump assembly, and an electromagnetic vacuum system with an inflation valve. The inner wall of the sealed vacuum chamber is lined with a reflector, the surface emissivity of which is not less than [value missing]. It is used to enhance the uniformity of diffuse reflection of thermal radiation inside the vacuum chamber; The mixing and feeding system includes a resin storage tank, a curing agent storage tank, a servo gear metering pump, a static mixer, and a three-axis servo motion module at the end. The servo gear metering pump is driven by an AC servo motor. A multi-physics field monitoring sensor array includes an infrared thermal imaging sensor, an online rotational viscometer located at the outlet of a static mixer, a pressure sensor array located inside a mold, and a distributed optical fiber sensing network embedded inside a lightweight transformer winding assembly. The control unit integrates a signal acquisition module, a power drive module, a communication interface module, and an edge computing module for running a three-dimensional flow field digital twin model. The control unit is used to execute closed-loop control logic.

2. The intelligent vacuum casting processing system for lightweight transformers according to claim 1, characterized in that, The system is used to perform the following steps: Step 1: Place the lightweight transformer winding assembly on the leveling support platform in the vacuum chamber, start the temperature control module to preheat the lightweight transformer winding assembly, and simultaneously start the vacuum pump group to reduce the absolute pressure in the vacuum chamber to the first pressure threshold and maintain the vacuum state, so as to achieve deep degassing in the micro gaps of the lightweight transformer winding assembly. Step 2: The dynamic viscosity of the mixed epoxy resin mixture is monitored in real time using an online rotational viscometer. The control unit adjusts the heating temperature of the static mixer according to the measured dynamic viscosity to keep the viscosity of the epoxy resin mixture within the target range before it enters the mold. Step 3: Open the injection valve and use a servo gear metering pump to inject the epoxy resin composite material into the mold. In the initial stage of injection, a constant volumetric flow rate is used. When the liquid level sensor detects that the liquid level has reached the bottom edge of the lightweight transformer winding assembly, the control unit switches to the pressure and volumetric flow rate collaborative feedback mode. The filling pressure is obtained in real time by multiple pressure sensors installed on the inner wall of the mold and the filling pressure gradient is calculated. Combined with the cross-sectional area of ​​the mold cavity corresponding to the current liquid level, the speed of the servo gear metering pump is dynamically adjusted so that the average flow velocity of the cross section flowing through the winding gap is maintained at the set macroscopic linear rise target speed of the liquid level, thereby keeping the linear rise speed of the liquid level in the complex cross-sectional flow channel constant. Step 4: During the filling process, the vacuum pump group controls the pressure in the vacuum chamber to alternate between a first pressure threshold and a second pressure threshold. The second pressure threshold is greater than the first pressure threshold. The pressure alternation effect drives the tiny bubbles remaining deep in the lightweight transformer winding assembly to drift to the liquid surface. At the same time, the infrared thermal imaging sensor monitors the temperature distribution on the outer wall of the mold, and the heating device set inside the vacuum chamber is adjusted to compensate the temperature field of the filling area in real time. Step 5: After the infusion is completed, the stepped curing stage begins. The first curing temperature is maintained for the first curing time, and then the temperature is increased to the second curing temperature at a controlled rate. During the curing process, the internal stress is monitored in real time by a distributed optical fiber sensor network embedded in the lightweight transformer winding assembly. If the stress time-varying rate is detected to exceed the set threshold, the control unit uses a local induction heating coil embedded in the outer wall of the mold to inductively heat the mold wall to adjust the temperature gradient of the mold wall. In turn, the local temperature gradient inside the epoxy resin composite is adjusted through heat conduction to reduce residual stress. After curing is completed, a controlled cooling procedure is executed.

3. The intelligent vacuum casting processing system for lightweight transformers according to claim 2, characterized in that, The preheating temperature in step one is set to [temperature value]. to The preheating time shall not be less than Hours; First pressure threshold not greater than .

4. The intelligent vacuum casting processing system for lightweight transformers according to claim 2, characterized in that, In step two, the target viscosity range of the epoxy resin composite material before it enters the mold is: to .

5. The intelligent vacuum casting processing system for lightweight transformers according to claim 2, characterized in that, In step two, the control unit analyzes the flow activation energy of the epoxy resin composition using a rheological compensation model based on the measured dynamic viscosity. The mathematical expression of the rheological compensation model is as follows: in, Dynamic viscosity, unit: , Frequency factor, unit: , The activation energy of the flow is expressed in units of 1000 kJ / m². , Here is the gas constant, with a value of [value missing]. , Absolute temperature, unit: When solving for the activation energy of the flow, the control unit retrieves temporarily stored historical temperature-viscosity data pairs and performs least-squares regression according to the rheological compensation model to obtain the activation energy of this batch of materials. and Based on this, the target static mixer heating temperature required to maintain the viscosity within the target range is calculated.

6. The intelligent vacuum casting processing system for lightweight transformers according to claim 2, characterized in that, In step two, the control unit has a built-in viscosity prediction model based on a multilayer sensor. The viscosity prediction model uses the proportion of epoxy resin composition, real-time temperature, and pressure difference between the inlet and outlet of the static mixer as input feature vectors to predict the viscosity change trend at future moments. When the predicted viscosity will exceed the target range, the control unit uses the predicted viscosity deviation as input and controls the heating unit of the static mixer through a PID closed-loop control loop to adjust the temperature of the heat transfer oil.

7. The intelligent vacuum casting processing system for lightweight transformers according to claim 2, characterized in that, In step three, under the pressure and volumetric flow rate coordinated feedback mode, the model upon which the control unit dynamically adjusts the speed of the servo gear metering pump is based is: in, For volumetric flow rate that needs to be output in real time, the unit is... , For the set target velocity of macroscopic linear rise of the liquid level, unit , The current liquid level is obtained based on a three-dimensional flow field digital twin model or a pre-calibrated height-area function. The corresponding instantaneous cross-sectional area of ​​the mold cavity, in units , The dynamic permeability coefficient of the current level winding structure, in units of The flow is identified in real time by the three-dimensional seepage digital twin model built into the control unit. The dynamic viscosity of the epoxy resin composition, in units. , The average value of the filling pressure gradient across the flow channel cross-section, obtained through three-dimensional differential calculation using a pressure sensor array, is expressed in units of... The control unit adjusts... Make the average flow velocity of the cross section in the winding gap reach This allows the liquid level to rise steadily at a constant linear velocity.

8. The intelligent vacuum casting processing system for lightweight transformers according to claim 2, characterized in that, In step four, the alternation frequency of the vacuum chamber pressure is controlled at... to .

9. The intelligent vacuum casting processing system for lightweight transformers according to claim 8, characterized in that, In step four, the alternation frequency of the pressure alternation is dynamically adjusted based on the real-time dynamic viscosity of the epoxy resin composition. The optimal selection of the alternation frequency is based on the following correlation: in, The optimized pressure alternation frequency, in units of , The empirical damping coefficient is related to the mold structure and bubble morphology, and the unit is 1. , This represents the density difference between the gas phase inside the bubble and the liquid phase of the surrounding epoxy resin composite, expressed in units of... , Let gravitational acceleration be the acceleration due to gravity, and its value be [value]. , The defined characteristic diameter for targeted removal of microbubbles, in units of , The real-time dynamic viscosity of the epoxy resin composition is used in the calculation. Substitute the units.

10. The intelligent vacuum casting processing system for lightweight transformers according to claim 2, characterized in that, In step four, the waveform of pressure alternation is a modified triangular wave, and the pressure rise rate during the pressurization phase is not equal to the pressure fall rate during the depressurization phase.

Images