Metal composite substrate stress balance regulation method

Abstract

The application discloses a metal composite substrate stress balance regulation method, and relates to the technical field of substrate thermal control, and sequentially comprises the following steps: acquiring geometric topology parameters of a first side metal layer and a second side metal layer and generating an asymmetric polarization tensor; inputting the asymmetric polarization tensor into a thermal-mechanical dynamic coupling feedforward model, solving a target heat exchange difference threshold and generating a safety envelope line; in a thermal treatment cooling process, when the body temperature enters a preset phase change window, spatially different thermodynamic intervention boundaries are applied to both sides; during the application, a cliff phase-locked fuse protection mechanism is triggered according to the safety envelope line, the spatially different thermodynamic intervention boundaries are removed, and synchronous uniform cooling is restored. The method can realize feedforward cancellation of asymmetric thermal bending moments, reduce the warping and instability risks after annealing, and improve the process adaptability, manufacturing consistency and pre-stress control stability of the metal composite substrate, and the service reliability thereof.

Description

Technical Field

[0001] This invention relates to the field of substrate thermal control technology, specifically to a method for stress balance control of metal composite substrates. Background Technology

[0002] Metal-bonded ceramic substrates are typically created by adding a copper or aluminum metal layer to a ceramic substrate such as alumina, aluminum nitride, silicon carbide, or silicon nitride to pattern the circuit area. These substrates are then cut into individual substrates and assembled. During large-scale panelization, automated handling, vacuum fixing, subsequent heat treatment, and assembly, these substrates must withstand two major challenges: internal stress caused by differences in the upper and lower metal patterns, and thermal mismatch during temperature rise and fall. Existing processes employ two main methods: one uses heat treatment equipment to radiate heat and gas cool the individual substrates, aiming for as uniform a lateral temperature distribution as possible; the other uses infrared sensors or thermal imaging to obtain zoned temperatures.

[0003] US Patent document US9839120B2 discloses a substrate anti-warping structure. This document addresses the warping problem that easily occurs in large-scale metal-bonded ceramic substrates. It involves adding an additional metal layer to the outer edge of the substrate with an already patterned metal layer. Multiple additional metal patterns are generated in the area where the additional metal layer is located. The basic idea is as follows: when the additional metal layer is located on the upper and lower surfaces of the substrate, the warping caused by the difference in the total mass of the additional metal layers on the upper and lower surfaces, or the difference in thermal expansion between the metal patterns and the ceramic substrate, is compensated by adjusting the total mass difference of the additional metal layers on the upper and lower surfaces. When the substrate shows a warping tendency along the length or width direction, the warping direction is structurally corrected by adjusting the total mass difference of the additional metal layers in different directions. This solution is applicable to metal-bonded ceramic substrates with metal bonding and pattern formation. The control method involves the placement of the additional metal layer, the number of patterns, and the distribution of their total mass.

[0004] The aforementioned technologies can only partially suppress warping by utilizing the static mass configuration of the additional metal layers, but ultimately rely on the structural compensation relationship after substrate formation. For metal composite substrates with asymmetrical upper and lower metal layer layouts, significant local variations in copper density, and requiring heat treatment for cooling, thermal stress is determined not only by the total metal mass of the upper and lower surfaces but also by the local metal area duty cycle, thickness, connectivity, transient temperature gradient, and the thermal contraction relationship of the material at different temperature zones. Although existing heat treatment equipment emphasizes temperature uniformity and points out that rapid cooling of high-temperature substrates can induce stress, warping, defects, or slippage, the temperature homogenization approach often only focuses on overall temperature control and cannot specifically constrain the bending trend caused by the imbalance of upper and lower metal layers. Therefore, in scenarios where the metal pattern is highly concentrated on one side and sparse on the other, relying on external additional metal layers or overall temperature homogenization cooling will still result in local thermal contraction mismatch and bending stress accumulation during the heat treatment cooling stage, affecting the stability of subsequent transportation, positioning, and assembly.

[0005] Therefore, how to reduce the accumulation of bending stress caused by the transient temperature gradient and thermal shrinkage mismatch during the cooling stage of heat treatment for metal composite substrates with asymmetrical upper and lower metal layers without changing the existing structure of the substrate is an urgent technical problem to be solved. Summary of the Invention

[0006] (a) Technical problems to be solved To address the shortcomings of existing technologies, this invention provides a method for stress balance control of metal composite substrates. The method involves inputting an asymmetric polarization tensor into a dynamically coupled feedforward thermomechanical model to calculate the target heat transfer range threshold and generate a safety envelope. During heat treatment cooling, when the substrate temperature enters a preset phase transition window, spatially differentiated thermodynamic intervention boundaries are applied to both sides. During this application, a cliff-locked phase-locked fuse protection mechanism is triggered based on the safety envelope, removing the spatially differentiated thermodynamic intervention boundaries and restoring synchronous cooling. This improves the process adaptability, manufacturing consistency, and prestress control stability of the metal composite substrate, as well as its service reliability; and solves the technical problems described in the background art.

[0007] (II) Technical Solution To achieve the above objectives, the present invention provides the following technical solution: A method for stress balance control of a metal composite substrate, executed by a control device, includes: acquiring the geometric topological parameters of the first and second metal layers of the composite substrate to be processed, and generating an asymmetric polarization tensor characterizing the imbalance between heat capacity and stiffness on both sides; inputting the asymmetric polarization tensor into a dynamic coupling feedforward model of a heat engine, solving for the target heat transfer range threshold and generating a safety envelope. During the heat treatment cooling process, the body temperature of the composite substrate is continuously collected. When it enters the preset phase change window, a spatially differentiated thermodynamic intervention boundary is applied to the first side metal layer and the second side metal layer according to the target heat transfer difference threshold. During the application of the spatially differentiated thermodynamic intervention boundary, the cliff-locked phase-locked fuse protection mechanism is triggered in real time according to the safety envelope, the spatially differentiated thermodynamic intervention boundary is removed and synchronous homogenization is restored.

[0008] Furthermore, geometric topological parameters are obtained, including: dividing the first side metal layer and the second side metal layer into multiple micro-regions based on the board coordinates of the composite substrate; extracting the metal area duty cycle, metal thickness information and connectivity state of each micro-region; and generating an asymmetric polarization tensor according to the area, thickness and connectivity state of each micro-region.

[0009] Furthermore, the asymmetric polarization tensor is input into the dynamic coupling feedforward model of the heat engine, including: determining the main process direction based on the transport direction, the action direction of the actuator, and the plate direction; projecting the asymmetric polarization tensor along the main process direction; and calculating the target heat transfer range threshold based on the projection results, the target service junction temperature, and the reference cold end temperature.

[0010] Furthermore, when it enters the preset phase transition window, spatially differentiated thermodynamic intervention boundaries are applied, including: mapping the bulk temperature to multiple execution partitions; determining the phase boundary hit state and safety margin state for each execution partition; and applying spatially differentiated thermodynamic intervention boundaries only to execution partitions where the phase boundary hit is achieved and the safety margin is positive.

[0011] Furthermore, after generating the target heat transfer range threshold and the safety envelope, the control device sends a control message to the thermal boundary execution control unit within the control device. The control message includes at least the substrate number, the asymmetric polarization tensor, the target heat transfer range threshold, the segment nodes of the safety envelope, and the index information of the preset phase change window.

[0012] Furthermore, the hot boundary execution control unit generates a first side boundary instruction and a second side boundary instruction for each execution partition based on the control message; each first side boundary instruction and each second side boundary instruction carries the corresponding execution partition number and is written into the partition control registers corresponding to the first side execution mechanism and the second side execution mechanism, respectively.

[0013] Furthermore, during the application of spatially differentiated thermodynamic intervention boundaries, the control device sends a control frame to the phase-locked fuse control unit within the control device; the control frame includes at least the substrate number, body temperature, current equivalent tensile stress, each first side boundary command, each second side boundary command, execution partition number, time sequence number, and verification field.

[0014] Furthermore, the control equipment includes an industrial camera, a high-frequency thermal imager, an acoustic emission probe array, a safety PLC, an FPGA phase-locked loop board, a first-side actuator, and a second-side actuator; the industrial camera acquires geometric topology parameters, the high-frequency thermal imager collects the body temperature, and the safety PLC and FPGA phase-locked loop board jointly execute the cliff-break phase-locked loop fuse protection mechanism.

[0015] Furthermore, when the surface metal of the first and second side metal layers is oxygen-free copper, the preset phase change window corresponds to a cooling range of 380 degrees Celsius to 280 degrees Celsius; the control device maintains the spatially differentiated thermodynamic intervention boundary in an effective state only after the body temperature enters the cooling range from high to low.

[0016] Furthermore, when the control device is in at least one of the following states: insufficient edge computing power, network disconnection, or black boxing of drawings, it calls the one-dimensional equivalent thermal expansion bias lookup table in the local read-only memory, and uses the difference in theoretical copper plating weight between the upper and lower metals and the asymmetric polarization tensor as the lookup table input, and outputs the target heat transfer range threshold and the segment node level of the safety envelope.

[0017] (III) Beneficial Effects This invention provides a method for stress balance control of metal composite substrates, which has the following beneficial effects: By acquiring the geometric topological parameters and asymmetric polarization tensors of the first and second side metal layers, the dispersed area, thickness, and connectivity differences can be transformed into continuously invoked imbalance characterization objects, avoiding control deviations based solely on empirical judgments of the direction of heat engine imbalance. By inputting the asymmetric polarization tensor into the dynamically coupled feedforward model of the heat engine, and simultaneously generating the target heat transfer range threshold and safety envelope, the hot bending moment during service can be pre-converted into an executable boundary control object during the manufacturing stage, providing subsequent interventions with a clear direction and brittle fracture boundary.

[0018] By applying a spatially differentiated thermodynamic intervention boundary when the body temperature enters the preset phase change window, the differentiated cooling can be limited to the stage when the surface metal still has stress release capability, so that the thermal field intervention is adapted to the material state change and the risk of ineffective loading or instability caused by applying pressure throughout the process can be avoided.

[0019] By coupling the target heat transfer range threshold, zoning determination, and spatially differentiated thermodynamic intervention boundary, it is possible to achieve directional control of local copper-plated imbalance areas rather than a homogeneous cooling path for the entire board, thus making it more adaptable to extremely asymmetric metal composite substrates. By triggering the cliff-locked phase-locked fuse protection mechanism based on the safety envelope, the spatially differentiated thermodynamic intervention boundary can be removed in time and synchronous cooling can be restored. This can retain the pre-set reverse prestress in the front section, suppress the continued accumulation of stress, and form a closed-loop control chain between differentiated intervention and tail-end convergence. Attached Figure Description

[0020] Figure 1 This is a diagram showing the overall architecture of the control system corresponding to the stress balance control method for a metal composite substrate according to the present invention. Figure 2 This is a schematic diagram of the topological analysis of the double-sided metal layer and the generation of the asymmetric polarization tensor in step one of the present invention; Figure 3 This is a schematic diagram of the process for solving the target heat transfer range threshold and the safety envelope feedforward in step two of this invention. Figure 4 This is a schematic diagram of the phase change window identification and partitioned differential thermal boundary execution in step three of the present invention; Figure 5This is a timing diagram illustrating the phase transition window hit, safety margin, and boundary switching between steps three and four of the present invention. Figure 6 This is a schematic diagram of the phase-locked fuse and synchronous cooling control in step four of the present invention; Figure 7 This is a comparative schematic diagram of the active fluid implementation method and the passive thermal resistance implementation method of the present invention. Detailed Implementation

[0021] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0022] Please see Figures 1-7 This invention provides a method for controlling the stress balance of a metal composite substrate, comprising: This embodiment addresses the problem of asymmetric endothermic shapes in metal composite substrates that cannot be accurately characterized before subsequent annealing and cooling. Step one involves compressing the discrete differences between the first and second metal layers in terms of layout, thickness, material, and connectivity into an asymmetric polarization tensor that can be directly used later. With macro polarization index The subsequent steps no longer deal with incomparable two-dimensional images and scattered process tables, but with thermodynamic feedforward coordinates that already have directionality, strength, and position. Therefore, it can provide an executable entry point for cross-phase boundary differential cooling without changing the existing material formulation.

[0023] Step 1: Convert the geometric, thickness, and material differences between the metal layers on both sides of the composite substrate, which are difficult to compare directly, into asymmetric polarization tensors that can be directly used in subsequent thermomechanical coupling solutions. .

[0024] In this embodiment, step one is performed by the parsing and control unit located in the production line control cabinet. The parsing and control unit includes an industrial camera, an area array light source, an encoder, an industrial computer, a drawing parsing program, an image segmentation program, a tensor generation program, and a message interface connected to the annealing equipment.

[0025] An industrial camera is mounted directly above the conveyor line, with area array light sources positioned on both sides of the substrate's normal direction to suppress false edges caused by pad reflections. An encoder outputs board displacement pulses, which the industrial control computer stitches together continuous images to create a flattened image consistent with the substrate's shape. For batches with existing design documents, the drawing parsing program directly reads the aperture definitions, region fill commands, polarity switching commands, and trace paths from the GerberRS-274X file (manufacturing drawing file) to generate the first and second side metal layer outlines in the same coordinate system as the camera flattened image. For batches where the drawings are packaged into a non-editable format, the parsing control unit uses a combination of the camera flattened image and the process BOM (Bill of Materials) as input, still generating the asymmetric polarization tensor (described later) according to the same terminology system. .

[0026] If only the total copper area is recorded, the operator can only know which side of the board has more or less copper, but cannot know which area has localized heat retention, nor can they know which narrow metal bridge will pull the strain direction during subsequent cooling. Therefore, step one addresses the issues of location and connectivity first, and then the issues of thickness and material. The former determines how heat travels along the board surface, and the latter determines how shrinkage forces are superimposed along the thickness direction; only when these two types of information coincide in the same grid coordinate system can subsequent steps apply a set of heat transfer boundaries to the correct area.

[0027] The analysis and control unit first establishes the board surface coordinates based on the substrate outline and divides the first and second side metal layers into micro-grids with fixed side lengths. The preferred grid side length is no greater than the narrowest functional trace width and no less than the minimum window length to avoid a single micro-area simultaneously absorbing a complete copper island and a large area of ​​blank substrate, which would lead to distortion of the subsequent symbol direction. For the flattened image from the camera, the image segmentation program first performs grayscale normalization, specular reflection suppression, and edge closure, then uses Otsu's threshold segmentation method to obtain a metal binary image. For the layer data from the GerberRS-274X file, the drawing analysis program directly rasterizes it according to the aperture shape, trajectory width, and area fill command to obtain a metal mask in the same format as the binary image.

[0028] Then, the connected component extraction program calculates whether there is a metal entity in each micro-region that is connected to adjacent micro-regions, classifying isolated islands, closed-loop copper rings, and long bridge connections into different connectivity states. Two metal pieces of the same area will have different effects on subsequent heat escape if one is just a closed island and the other runs through the main heat dissipation structure. The analytical control unit generates geometric polarization nuclei for each micro-region. : In the formula, the area of ​​the metal on the first side : No. The projected area of ​​the first side metal layer within each micro-region is a plate surface area with a value greater than 0; the area of ​​the second side metal layer... : No. The projected area of ​​the second side metal layer within each micro-region is the plate area with a value greater than 0. First side metal thickness : No. The nominal thickness of the first-side metal layer within each micro-region is a value greater than 0, used to convert a simple area difference into a volumetric thermal capacity difference; the thickness of the second-side metal layer... : No. The nominal thickness of the second-side metal layer within each micro-region is a thickness greater than 0, used to simultaneously upgrade the area information of the second-side metal layer to volume-level information. First-side connectivity coefficient : No. The degree of continuity between the first side metal layer and adjacent micro-regions and the main heat dissipation path within each micro-region is denoted by a value range of . Second-side connectivity coefficient : No. The degree of continuity between the second-side metal layer within each micro-region and adjacent micro-regions and the main heat dissipation path, with a value range of [value range missing]. ; Geometric polarization nucleus : No. The geometric bias direction and strength of each micro-region, with a value range of [missing information]. Its symbol is used to distinguish whether it is biased towards the first side metal layer or the second side metal layer, and its absolute value is used to represent the bias strength. Stable small quantity A constant used to prevent the denominator from becoming undefined when both metal layers are close to blank at the same time, with a value greater than 0 and much smaller than the principal numerator; For the first The micro-region, corresponding to the first Side metal layer, connectivity coefficient Generate using the following formula: Among them, the adjacency connectivity ratio The ratio of the number of adjacent micro-regions on the same layer that share a boundary and are interconnected with this micro-region to the theoretical maximum number of adjacent micro-regions; backbone connection marker. This is a binary quantity. It is set to 1 if the micro-region can be reached via a connected component search from the main heat sink copper strip, the main island of the device pad, or the external terminal metal; otherwise, it is set to 0. The heat source reachability is... The normalized inverse vector of the shortest conduction path length from the micro-region to the projection area of ​​the packaged heat source; weighting coefficients It is a preset constant that satisfies the condition that the sum is 1, and is used to balance the three factors of local connectivity, trunk reachability and heat source reachability.

[0029] For example, on a substrate used as a power module carrier, the first metal layer has a large area of ​​copper plating in the middle, while the second metal layer is dominated by narrow return lines. After the camera flattening image is entered into the image segmentation process, the central copper plating is identified as a large connected surface, and the narrow return lines are identified as elongated connected bodies extending longitudinally. Even if their projected areas in a certain local window are similar, the connectivity extraction program will still give different first-side connectivity coefficients. Connectivity with the second side Thus, the geometrically polarized nucleus Moving beyond simply measuring area differences, the system now incorporates information about the direction of heat flow. For older production lines without visual modifications, the analytical control unit can directly read the theoretical copper plating weight, layer thickness, and corresponding board area from the process BOM. It first divides the entire board into a coarse grid, then distributes the theoretical copper plating weight difference proportionally to each micro-area to generate low-resolution geometric polarization nuclei. Although this downgrade path eliminates image acquisition, it retains the same input format as subsequent steps, thus not compromising the single inventive concept of the entire method.

[0030] Furthermore, the analytical control unit obtains a geometric polarization diagram with positive and negative directions. This diagram can be overlaid back onto the original flattened diagram for process engineers to verify, or it can be directly used as the input base map for the next technical point. This compresses the heterogeneous information scattered in drawing files, camera images, and process BOM tables into a micro-area-level offset that can be transferred under a unified coordinate system.

[0031] Only geometrically polarized nuclei This is still insufficient to support subsequent thermomechanical solutions because area and connectivity only explain how heat spreads across the plate surface, not how different material layers convert thermal contraction into bending driving force in the thickness direction. Therefore, the analytical control unit, while keeping the micro-region coordinates of technical point A unchanged, continues to incorporate the metal layer thickness, elastic modulus, coefficient of linear expansion, and substrate layer relationships into the same micro-region to generate a thermomechanical bias core and an asymmetric polarization tensor. The dimensionality reduction here does not mean discarding thickness direction information, but rather extracting the direction of the bending driving force that is actually needed later, so that subsequent solutions do not need to start from the complete three-dimensional structure every time the process changes.

[0032] The parsing control unit first determines the first-side metal layer, second-side metal layer, and dielectric layer sequence corresponding to each micro-region based on the stack-up table, and then reads the material grade and layer thickness from the process BOM. If a material database is provided for this batch, the elastic modulus and coefficient of linear expansion in the material database are directly retrieved; if the material database is not included with the order, the nominal parameters corresponding one-to-one with the grade in the preset material table are read. Subsequently, a thermomechanical bias core is generated. : In the formula, the elastic modulus of the first side : No. The elastic modulus of the first metal layer within each micro-region is a modulus value greater than 0; the elastic modulus of the second layer... : No. The elastic modulus of the second metal layer within each micro-region is a modulus value greater than 0; the coefficient of linear expansion of the first side... : No. The linear expansion coefficient of the first side metal layer in each micro-region is the thermal expansion amount with a value greater than 0. Second lateral expansion coefficient : No. The linear expansion coefficient of the second-side metal layer within each micro-region is a thermal expansion quantity with a value greater than 0; the thermomechanical bias core : No. The direction and intensity of thermomechanical imbalance in the thickness direction of each micro-region can be positive or negative. The sign of the value distinguishes which side dominates the bending driving force, and the absolute value represents the local intensity of that driving force. (Stable small quantity) This is used to ensure that the formula is continuous when the thickness on both sides is extremely small at the same time. Obtaining the thermodynamic bias core Then, the tensor generation procedure generates the geometric polarization kernel. With thermal bias core Superimposed as asymmetric polarization tensor and macro polarization index : The analytical control unit first forms a three-dimensional polarization parameter set in the plate surface coordinate system and the thickness coordinate system, and then performs dimensionality reduction and merging along the principal axes of the plate surface, outputting a second-order asymmetric polarization tensor for use in step two. To ensure that the geometrically polarized nucleus and the thermomechanical bias nucleus have consistent physical dimensions, the thermomechanical bias nucleus is written in a dimensionless form: Among them, the reference linear expansion coefficient The baseline linear expansion coefficient in the material library of this product family is used to perform dimensionless processing on the thermomechanical bias core; after this processing, the geometric polarization core... With thermal bias core Both can be used as dimensionless quantities in subsequent tensor assembly.

[0033] Asymmetric polarization tensor The asymmetric thermomechanical characterization results of the entire plate in a unified coordinate system serve as a compression carrier for information on the coupling direction between the plate surface and thickness; macroscopic polarization index. : The coupling strength between geometric bias and thermomechanical bias within the entire plate area, a scalar with a value greater than or equal to 0, used to provide a single-value criterion for whether the entire plate belongs to a strongly asymmetric batch; Weighting coefficient : No. The contribution weight of each micro-region to the bending driving force of the entire plate is within the range of [value range missing]. The parameters are determined based on the area of ​​the micro-region, the distance of the micro-region to the geometric neutral axis, and whether the micro-region is located within the projection area of ​​the device's heat source; the total number of micro-regions : The number of micro-regions involved in tensor generation, taking values ​​greater than 0, used to limit the summation range; Geometric polarization kernel With thermal bias core The definition remains the same as before, and the meaning of the symbol is unchanged; When the first and second metal layers are made of the same material, the volumetric heat capacity coefficient is considered a constant and canceled out in the geometric polarization nucleus; when the first and second metal layers are made of different materials, the geometric polarization nucleus adopts the following generalized form: Among them, volumetric heat capacity coefficient The volumetric heat capacity coefficient is the product of the density and specific heat capacity of the first side metal layer. The product of the density of the second-side metal layer and its specific heat capacity is used to extend the simple geometric difference into a difference in material heat capacity.

[0034] Volumetric heat capacity coefficient The meaning is the heat storage capacity per unit volume of the first metal layer, which is taken as the product of density and specific heat capacity; volumetric heat capacity coefficient. Similarly, there are two ways to obtain this information: one is to read it directly from the material supplier's data sheet, and the other is for the applicant to establish a material library and bind the material grade with the volumetric heat capacity coefficient.

[0035] Preferably, the tensor generation program runs locally on the industrial control computer, with inputs including a flattened diagram, GerberRS-274X parsing results, a stack-up table, and a process BOM table, and outputs an on-board control message. The on-board control message includes at least the substrate number and the asymmetric polarization tensor. Macro polarization index The polarization diagram coordinate origin and micro-area mapping table are recorded; the message is written to the control buffer of the annealing equipment via fieldbus for direct reading in subsequent steps. If a batch cannot access the complete material database due to network interruption, the parsing control unit switches to a degraded path: retaining the obtained geometric polarization nuclei. Simultaneously, a simplified thermomechanical bias core is constructed using the theoretical copper plating weight difference and layer thickness difference between the upper and lower metals in the process BOM. Then generate the asymmetric polarization tensor using the same formula. .

[0036] In the preferred implementation, the drawing parsing program first creates a layer list based on five object categories: board outline, window, copper pour, via, and island. Then, it aligns the layer list with the overlay table one by one. Finally, the tensor generation program generates the geometric polarization kernel for each micro-region. Thermal bias core and weighting coefficients Write the same mapping record. What the operator sees on the screen is not an abstract matrix, but a polarization overlay diagram distributed along the board surface: areas biased towards the first metal layer are marked in the same direction, and areas biased towards the second metal layer are marked in the opposite direction. The color depth only corresponds to the strength of the bias and does not change the meaning of the symbols. In this way, the field personnel can directly see which device pad area, which via-intensive area, and which blank dielectric area is dominating the subsequent bending driving force, and the output of step one also has a verifiable physical orientation.

[0037] Furthermore, the previously dispersed differences in area, thickness, material, and connectivity between the first and second metal layers are converged into an asymmetric polarization tensor. With macro polarization index Two unified objects. Subsequent steps can then directly determine which side the differential heat transfer boundary should be offset towards. Process engineers can also check whether the polarization direction in the overlay diagram matches the material list, completing the result verification of step one. Corresponding evaluation methods include checking the overlap between the polarization diagram and the original contour, checking the consistency between the overlay table and the material list, and using the asymmetric polarization tensor. Check whether the main direction is consistent with the theoretical copper plating bias.

[0038] Step 2: Based on the asymmetric polarization tensor With macro polarization index Inverse calculation of the target heat transfer range threshold required to offset the service-state bending moment This forms a safety envelope that constrains subsequent differentiated cooling actions. .

[0039] This step is executed by the feedforward solver control unit. The feedforward solver control unit is located within the control cabinet of the annealing equipment and includes an industrial computer, solver program, local material library, local read-only memory, fieldbus interface, electrical isolation board, and cloud forwarding interface. The feedforward solver control unit receives the on-board control message from step one. This on-board control message includes at least the substrate number and the asymmetric polarization tensor. Macro polarization index The stack-up code, material code, and cyclic redundancy check field are used. The feedforward solver control unit first checks the cyclic redundancy check field. If the check passes, it loads the stack-up table and material table. If the check fails, it re-requests the message. If it still fails, it switches to the one-dimensional equivalent thermal expansion bias lookup table path in the local read-only memory, thus ensuring that step two can always output continuous control data to step three.

[0040] The asymmetric polarization tensor obtained in step one The location of the bias towards the first and second metal layers of the entire board has been given, but this result remains at the structural characterization level. Annealing equipment cannot operate directly based on matrix values; the placement of servo-frequency array air knives or sacrificial masks requires a boundary quantity with direction and amplitude. Therefore, in step two, the feedforward solution control unit first establishes a thermomechanical coupling model consistent with the substrate stack, and then applies the asymmetric polarization tensor... Projecting onto the main cooling direction, the target heat transfer range threshold is then calculated in reverse. Then, the upper limit of the fracture tolerance of the ceramic medium during the cooling stage is incorporated into the same solution chain to form a temperature-stress two-dimensional safety envelope. The resulting single-chain output can drive active fluid intervention and guide passive thermal resistance deployment, avoiding the disconnect between the subsequent controlled object and the preceding characterized object.

[0041] The feedforward solver control unit does not directly convert the asymmetric polarization tensor Instead of substituting the airflow distribution table, we first determine the main escape direction of heat on this batch of substrates, and then read the asymmetric polarization tensor along that direction. The effective bias. This is done to prevent the same large copper area from being misidentified as the same cooling object under different connectivity configurations. Only by using the asymmetric polarization tensor... With the process principal direction vector Binding, and subsequently calculated target heat transfer range threshold This will prevent the strong intervention that should be applied to the edge heat release zone from being mistakenly sent to the central heat retention zone.

[0042] The feedforward solution control unit first establishes the main direction vector of the process based on the conveying direction, the air knife opening direction, and the long side direction of the substrate. Then, for the asymmetric polarization tensor Perform directional projection to obtain the directional projection factor. : In the formula: the main direction vector of the process The annealing and cooling stage is characterized by the unified direction of heat escape and force application by the actuator, taking the value of a two-dimensional or three-dimensional non-zero vector, used to align the structural representation coordinates with the equipment action coordinates; asymmetric polarization tensor. The whole-plate asymmetric thermomechanical characterization results generated in step one consist of components derived from micro-area area, thickness, connectivity, and material imbalance, providing the directional bias source required for subsequent inverse calculations; directional projection factor. Asymmetric polarization tensor In the main direction vector of the process The effective bias strength can be positive or negative. Its sign indicates whether the cooling drive of the first side metal layer or the second side metal layer is preferentially enhanced, and its absolute value characterizes the bias strength. Obtaining the directional projection factor Then, the feedforward solution control unit calls the thermo-elastic-plastic solver.

[0043] Preferably, the thermo-elastic-plastic solver employs a nonlinear finite element path, using an eight-node hexahedral element or an equivalent three-dimensional solid element for the mesh type. The temperature and displacement fields are time-stepped simultaneously, and the material constitutive model uses the Anand constitutive relation or an equivalent high-temperature viscoplastic constitutive relation. Time discretization uses a backward difference scheme, and nonlinear iteration uses Newton iteration or an equivalent sparse matrix iteration path. For each optional metallic layer material, the applicant obtains the required parameters for the Anand model through high-temperature constant-stress creep tests or by accessing a publicly available material database from the supplier and writes them into the material library; during step two, the solution is automatically read according to the material code.

[0044] The solver does not directly output wind speed, but first combines it with the target service junction temperature. Reference cold junction temperature and stacked thermal channel coefficient Inverse calculation of the target heat transfer range threshold : The feedforward solution control unit first uses the asymmetric polarization tensor Macro polarization index The target cancelling bending moment is generated by the temperature difference during service with the target, and then the sensitivity of the bending moment to the heat transfer range on both sides is calculated by the thermo-mechanical coupling model. Based on this, the target heat transfer range threshold is calculated in reverse: Among them, the target cancels the bending moment This represents the reverse bending moment that needs to be pre-set during the manufacturing stage to counteract the service-state thermal bending moment; calibration factor. Material library is written by fitting calibration plates from the same product family; reference boundary difference This represents the baseline heat transfer difference under the current symmetrical cooling boundary; the disturbance amount. Small boundary disturbances applied to determine sensitivity; bending moment response For the thermo-mechanical coupling model at boundary differences of The total bending moment of the plate output at that time; sensitivity This represents the incremental response of bending moment to heat transfer range.

[0045] Target heat transfer range threshold The third step in the subsequent process requires establishing a spatially differentiated heat transfer amplitude between the first and second metal layers. This amplitude can be positive or negative, with the sign corresponding to which side is prioritized for cooling and the absolute value corresponding to the intervention intensity. The directional projection factor is also required. Using the previous definition, it is used to define the asymmetric polarization tensor. The direction bias injection threshold is calculated in reverse; Macro polarization index The overall plate coupling bias strength output in step one is a scalar with a value greater than or equal to 0, used to incorporate the total cumulative amount of local micro-region bias into the threshold calculation; target service junction temperature. The target temperature required to counteract hot bending moments during device service is a process setting temperature that is above room temperature but below the metal's melting point; reference cold junction temperature. The reference temperature after annealing before entering the room temperature range is a value lower than the target service junction temperature. Process temperature; thermal channel coefficient of the laminate The through-heat flux capability formed by the first side metal layer, the second side metal layer, and the dielectric layer is a scalar value greater than 0, used to suppress the heat flux solely determined by the macroscopic polarization index. Amplification over-threshold; stable small quantity To prevent the denominator from losing its definition under extremely weak thermal conductivity conditions, a constant is taken, with a value greater than 0 and much smaller than the stacked thermal channel coefficient. A fixed value; For example, when an aluminum nitride composite substrate with a continuous large copper island on the first side and multiple thin-line loops on the second side enters the pre-annealing buffer, the feedforward solver control unit first reads the asymmetric polarization tensor from the on-board control message. With macro polarization index Then, the parameters for oxygen-free copper, ceramic dielectric, and brazing layer are imported from the local material library. The industrial control computer screen first displays the central heat retention zone, edge heat release zone, and through-hole dense zone divided along the conveying direction. Then, the thermo-elastic-plastic solver provides the target heat transfer range threshold on the same plate surface mapping. The direction and magnitude. At this point, step three no longer requires reading the original drawing; only the target heat transfer range threshold needs to be read. By mapping the partitions, the opening of the servo frequency converter array air knife on both sides can be organized, or the placement of the sacrificial mask can be organized.

[0046] Furthermore, the asymmetric polarization tensor Principal direction vector of the process After projection, to avoid the structural characterization results losing their physical orientation at the control end; target heat transfer range threshold. From the macro polarization index Temperature difference lever and stacked thermal channel coefficient The joint generation enables subsequent differentiated cooling actions to have a direct engineering focus.

[0047] Furthermore, achieve the target heat transfer range threshold. After reverse calculation, if it is directly handed over to the execution mechanism, there will still be a risk: when some substrates still have plastic slip margin in the metal layer, the ceramic medium is already close to the starting point of cracking. If the heat transfer difference between the two sides is further amplified at this time, the action originally used to offset the bending moment in service state will be transformed into a source of cracking in manufacturing state.

[0048] Therefore, by incorporating the upper limit of ceramic medium fracture tolerance into the same feedforward solution chain, the output includes not only the target heat transfer range threshold. It also includes the temperature corresponding to each process. Safety envelope In this way, when implementing differentiated cooling in step three, we always work within the allowed boundaries, rather than going back to fix things after execution.

[0049] The feedforward solution control unit first calculates the process temperature based on the nodal temperatures and displacement fields in the finite element mesh. equivalent tensile stress Then, the temperature-dependent fracture strength of the ceramic medium is called. With Weber modulus The fracture risk factor was obtained. : Where: fracture risk factor Current process temperature The risk of the ceramic medium entering the unstable fracture region under the current stress state is considered, with a value range of [value missing]. The closer the value is to 1, the smaller the margin for subsequent differentiated cooling. Equivalent tensile stress : The tensile stress scalar generated by the temperature field and displacement field and projected onto the critical surface of the ceramic medium, with a value greater than or equal to 0; The current equivalent tensile stress The maximum principal tensile stress at the critical node of the ceramic layer is taken; it is directly extracted from the thermomechanical coupling model under the current boundary conditions, or calculated based on the material's elastic constants after measuring the principal strain using strain gauges placed on the back of the ceramic. Temperature-dependent fracture strength. With Weber modulus The results were obtained from ring-on-ring bending tests or four-point bending tests performed on ceramic samples from the same batch at multiple calibration temperatures, and the results were fitted and written into the material library according to the failure sequence; gating risk thresholds were used. The process parameters are fixed at the product family level and do not change with individual boards after the database is created once.

[0050] Temperature-dependent fracture strength Ceramic media at process temperature The ultimate tensile strength before fracture that can be withstood, taking a value greater than 0 as a function of temperature; process temperature. The plate temperature at any sampling moment during the annealing cooling process is taken as the process temperature within the current annealing window; Weber modulus. The discrete sensitivity corresponding to the defect distribution in the ceramic medium is a material parameter with a value greater than 1, used to adjust the fracture risk factor. The amplification rate of stress growth; Subsequently, the feedforward solution control unit is based on the gating risk threshold. Reverse the safety envelope : Where: safety envelope At process temperature The following allows for the application of differentiated cooling actions without crossing the entry control risk threshold. The maximum equivalent tensile stress boundary, with a value greater than 0, is a temperature function used to provide a directly comparable red line for step three; Gating risk threshold : The upper bound of the fracture risk that can be retained during the manufacturing stage, with a value between 0 and 1, used to transform the enterprise's process access requirements into mathematical boundaries; Temperature-dependent fracture strength With Weber modulus Using the previous definition, their functions correspond to the material's temperature-sensitive load-bearing capacity and the steepness of the risk curve, respectively. In the preferred implementation, the feedforward solver control unit generates two types of outputs. The first type of output is a control message that can be directly invoked in step three, containing the substrate number and the target heat transfer range threshold. Safety envelope The first type of output includes segmentation nodes, phase change window index bits, and message verification fields; the second type of output is a process page for field personnel to view, which displays the priority cooling markers for the first and second side metal layers and the safety envelope. The temperature inflection point. If the cloud forwarding interface is unobstructed, the feedforward solver control unit will send the same batch of data to the cloud solver to perform a full-scale model recalculation, and then set the recalculated target heat transfer range threshold. Write back to the local cache; if edge computing power is limited, network is interrupted, or the design drawings only provide a simplified stack-up record, the feedforward solver control unit will instead read the one-dimensional equivalent thermal expansion bias lookup table from the local read-only memory. The row index of the lookup table is the macroscopic polarization index. The segmented values, with column indices representing the segmented values ​​of the theoretical copper plating weight difference, are stored within the cross cells, representing the target heat transfer range threshold. Gear command and safety envelope Conservative piecewise nodes. In this way, even if the solution path for partial differential equations in continuous media is bypassed, step three can still receive control inputs with consistent sign and object.

[0051] Furthermore, fracture risk factors The upper limit of ceramic media embrittlement and the target heat transfer range threshold. As a solution chain, to prevent subsequent actuators from continuing to apply pressure in areas where the structure allows but the materials do not; safety envelope After the segmented node output, the actuator in step three can directly determine the process temperature. By comparing boundaries without needing to recalculate material fracture, the executability of the entire chain at the device end is maintained.

[0052] Step 3: When the surface metal transitions from viscoplastic rheology to elastic brittleness, based on the target heat transfer range threshold... and safety envelope Spatially differentiated thermodynamic intervention boundaries are set for the first and second side metal layers respectively, and the intervention boundaries are limited to the temperature-stress margin that can be managed in step four.

[0053] This step is executed by the thermal boundary execution control unit. The thermal boundary execution control unit is connected to the feedforward solution control unit via a fieldbus, and is also connected to a high-frequency thermal imager, a reference blackbody, a support fixture encoder, a first-side air knife cavity, a second-side air knife cavity, a proportional valve group, a frequency converter group, a sacrificial mask coating station, and a furnace door interlock. The high-frequency thermal imager is installed between the annealing chamber outlet and the differential cooling chamber inlet, with its lens field of view covering the entire plate. The reference blackbody is fixed to the edge of the support fixture for emissivity correction. The first-side and second-side air knife cavities correspond to the first and second side metal layers, respectively. If a passive thermal resistance implementation is adopted, the sacrificial mask coating station is located in the pre-furnace loading section and shares the same substrate number with the partition control message in step two. The reason for this setup is that step three is not simply about lowering the plate temperature, but rather about using differentiated heat transfer within an extremely narrow temperature range to pre-replace the bending moment that would otherwise be released during service into a controllable reverse prestress during manufacturing. Therefore, image acquisition, window judgment, command calculation, and boundary application should be in a continuous signal chain.

[0054] During the annealing and cooling process, the surface metal is not always suitable for withstanding the temperature difference between the two sides. If the temperature is too high, the metal is too soft overall, and differential heat transfer will only cause heat to be drawn off from the surface, making it difficult to form directional prestress. If the temperature is too low, the metal's slippage ability is significantly narrowed, and the temperature difference between the two sides will directly increase the tensile stress of the ceramic medium. Therefore, the thermal boundary execution control unit first uses a high-frequency thermal imager to read the plate surface temperature field frame by frame, and then maps the plate surface temperature field back to the micro-area grid in step one to confirm whether each zone has entered the phase transition window pre-stored in the material library.

[0055] For example, when OFHC oxygen-free copper is used as the surface metal, the phase transformation window is set between 380 and 280°C; when the surface metal is changed to a copper-molybdenum composite layer, the thermal boundary execution control unit calls the corresponding window based on the material code. Simultaneously, the thermal boundary execution control unit also adjusts the current equivalent tensile stress... With safety envelope Synchronous comparison: Only when the phase change window is hit and the safety margin still exists will the subsequent differentiated thermal boundary be issued to the actuator.

[0056] The thermal boundary execution control unit first performs three steps of processing on the continuous images sent by the high-frequency thermal imager: the first step is to perform emissivity correction according to the reference blackbody, so that the bright reflective areas of the pads and the dark areas of the dielectric are brought back to the same temperature scale; the second step is to call the micro-area mapping table retained in the first step and resample the current temperature field to match the asymmetric polarization tensor. The same partition coordinates; the third step is to calculate the surface cooling slope of each partition using three consecutive frames of temperature values.

[0057] The use of three frames instead of a single frame differential is to suppress isolated jumps caused by furnace door opening, fixture edges, and lens thermal drift, and to avoid misidentifying non-physical noise as phase boundary crossing signals. After processing, the thermal boundary execution control unit generates phase boundary hit scores for each zone. : Where: process temperature The current zone's surface temperature after high-frequency thermal imager correction is taken as the process temperature during annealing and cooling, indicating whether the zone has entered the phase transition window; the lower limit temperature of the window. The lower limit temperature at which this material system transitions from viscoplastic rheology to elastic brittleness is a value lower than the upper limit temperature of the window. Material parameters; Upper limit temperature of the window The upper limit temperature at which this material system enters the phase transition window is a value higher than the lower limit temperature of the window. Material parameters; stable small quantities : To prevent the denominator from losing its defined constant when the window width is extremely narrow, a constant with a value greater than 0 and much smaller than the square of the window width; cooling slope The surface cooling rate of the current partition is obtained from the temperature values ​​of three consecutive frames, and the value can be negative; Slope benchmark value : The slope reference value written into the local parameter library during the equipment calibration stage, with a value greater than 0; Phase boundary hit rate The current partition satisfies both the window position and the phase boundary hit intensity after the cooling activity level. The value is a scalar greater than 0 or less than or equal to 0. When the value is greater than 0, it indicates that the partition has the conditions to enter the subsequent differential thermal boundary solution. Obtain the accuracy of the phase boundary. Subsequently, the thermal boundary execution control unit continues to compare the current equivalent tensile stress. With safety envelope The safety margin factor is obtained. : Where: safety envelope The temperature-dependent stress allowable boundary generated in step two is a temperature function with a value greater than 0, used to provide the current process temperature. The upper limit of the acceptable equivalent tensile stress; the current equivalent tensile stress The instantaneous tensile stress obtained by back-calculation from the thermo-mechanical coupling model in step two, combined with the current temperature field, is taken as a stress value greater than or equal to 0; stable small quantity The previous definition is used to maintain the continuity of the denominator; safety margin factor. Current partition distance from the safe envelope The remaining margin, when the value is greater than 0, indicates that the differential thermal boundary can still be applied, and when the value is less than or equal to 0, it indicates that the enhanced cooling of this zone should be stopped; For example, when a substrate with a large-area device pad in the center of the first side metal layer and reflow copper strips at the edge of the second side metal layer enters the differentiated cooling cavity, the carrier fixture encoder first indicates the substrate's entry position, and the thermal boundary execution control unit then locks the corresponding partition mapping. A high-frequency thermal imager continuously sends full-board temperature images, and the control interface first displays the temperature trajectories of the central area, edge area, and via dense area; when the process temperature of the central area... Enter window upper limit temperature With the lower limit temperature of the window Between, and the phase boundary hit rate of that area Greater than 0, safety margin factor The system will only allow the area to be sent to subsequent steps if the value is still positive.

[0058] The process temperature Instead of directly taking the pixel values ​​from the thermal imager surface, it uses the equivalent plate temperature calculated from the surface temperature using a thickness correction table. For the first... Execution partitions: Among them, surface temperature Obtained from a high-frequency thermal imager after emissivity correction; temperature difference correction amount The thickness correction table for this product family is provided; linear velocity parameters. Obtained from the encoder of the load-bearing fixture; material code and stacked code It is brought in by the control message from step two.

[0059] If the edge heat release zone falls below the lower limit temperature window first. If so, the area is marked as a read-only area on the interface and will no longer accept differentiated strong cold boundaries.

[0060] Furthermore, step three no longer relies on the average temperature of the entire board, but instead uses the hit of the partition window as the starting point to lock the intervention time; at the same time, it uses a safety margin factor. By filtering out areas that are close to the brittle fracture threshold, the intervention starting point is limited to a narrow temperature range that can generate plastic energy absorption without triggering medium collapse.

[0061] Furthermore, after the partition passes the dual determination, the thermal boundary execution control unit does not set the target heat transfer range threshold. Instead of applying a uniform value for the entire board directly, this method combines the partition weights from step one and the direction determination from step two to generate partition boundary instructions. The purpose of this approach is to decompose the directional threshold, which originally only exists in the solver, into first and second side boundary instructions that the actuator can receive, and to ensure that strong and weak cooling actions in different partitions do not cancel each other out.

[0062] For the Each execution partition, the hot boundary execution control unit generates the first side boundary instruction. With the second side boundary command : In the formula: First side boundary command : Applied to the first The target heat transfer boundary value of the first side metal layer of each execution zone is a boundary command that can be converted into flow rate, air velocity, or thermal resistance thickness by the actuator; the boundary command of the second side... : Applied to the first The target heat transfer boundary value of the second-side metal layer of the execution partition is the same as that of the first-side boundary command. Similar to the first side boundary instruction. Formation of controlled reverse difference; reference boundary quantity Before the implementation of differentiated intervention, the first The baseline heat transfer level of each execution zone is taken as the zone baseline value written into the parameter table after the equipment is calibrated. Target heat transfer range threshold Step 2 outputs the bilateral target heat transfer difference, which can be positive or negative. Its sign determines whether the cooling of the first or second metal layer is prioritized; partition coupling coefficient. : From the asymmetric polarization tensor of step one The first one obtained by conversion with the micro-area mapping table Each execution partition weight, with a value range of [value range missing]. Phase boundary accuracy Using the previous definition, this is used to directly write the window hit rate into the boundary amplification chain; safety margin factor. Using the previous definition, it is used to measure the distance of the current partition from the safety envelope. Write the remaining space into the boundary shrink chain; execute the partition number. The partition number after the transverse division along the conveying direction is a positive integer and is used to correspond the boundary command to the specific spraying or coating area.

[0063] In the active fluid implementation, the first side boundary command With the second side boundary command Instead of acting directly on the proportional valve, the values ​​are input into the equipment calibration table. This calibration table uses the zone number, target boundary quantity, and current total flow rate level as indexes to output the corresponding proportional valve opening and fan frequency. The calibration table is generated from calibration plate tests: under known nozzle slot width, cavity pressure, and plate spacing conditions, the correspondence between valve position, fan frequency, and actual heat transfer coefficients on both sides is recorded for each level and written to the local read-only memory.

[0064] In the passive thermal resistance implementation, the thermal boundary execution control unit does not output air knife commands, but instead assigns partition weights. and gating coefficient The mapping is applied to a lookup table of mask thickness level, porosity level, and pre-baking time, and the sacrificial mask is only deployed on the side with the lower copper duty cycle determined in step one. Thus, the active fluid path and the passive thermal resistance path maintain consistency in their controlled objects and form a parallel substitution relationship in their execution methods.

[0065] In the active fluid implementation, the first side boundary command With the second side boundary command The control registers for the first and second side air knife cavities are written separately. Each air knife cavity is divided into multiple independent injection zones along the conveying direction. The nozzle slit width is preferably designed to be no more than one-third of the length of the short side of the corresponding execution zone to prevent cross-zone interference at the zone boundaries. A proportional valve group and a frequency converter group are set behind each independent injection zone. The proportional valve group is responsible for adjusting the instantaneous air volume, and the frequency converter group is responsible for adjusting the fan speed. Together, they transcribe the boundary commands into zoned jets. To prevent the first and second side air knife cavities from competing for heat from the plate surface at the same time, the thermal boundary execution control unit first writes the boundary command of the high polarization side and then the compensation command of the low polarization side within the same control cycle, and both sides call the same zone coupling coefficient. When the central pad area needs enhanced cooling on the first side, the control interface shows that the opening of the spray area corresponding to the first side air knife cavity increases while the opening of the spray area corresponding to the second side air knife cavity simultaneously narrows. After the substrate passes through the cooling cavity, the central area and the edge area maintain a controlled temperature difference rather than uncontrolled cooling.

[0066] In the passive thermal resistance implementation, the thermal boundary execution control unit transmits the first side boundary command. With the second side boundary command Convert to a sacrificial mask layout diagram.

[0067] For example, the sacrificial mask is formulated with Teflon-modified resin, volatile pore-forming agent, and alcohol ether carrier liquid; its coating position is selected on the side with a lower copper duty cycle, and its thickness partition follows the execution partition number. Before entering the furnace, the dispensing head or screen printing squeegee deposits the material in the designated partition according to the sacrificial mask layout, and then pre-baking forms a transient heat-resistant layer with a set porosity and thermal resistance attenuation path. After the substrate enters the conventional cooling furnace, the sacrificial mask does not actively blow air, but instead suppresses the heat escape rate of the low duty cycle partition, causing the metal layer on the other side to release heat first, thereby passively establishing a temperature difference between the two sides under the same furnace conditions. This parallel implementation and the active fluid implementation share the same target heat transfer difference threshold. Coupling coefficient of the same partition and the same safety margin factor Therefore, the overall technical concept of the present invention will not be changed, but the boundary application means will be changed from external fluid to surface thermal resistance.

[0068] Furthermore, the first side boundary command With the second side boundary command Target heat transfer range threshold Clearly decompose into executable two-sided actions, partition coupling coefficient The structural polarization information formed in step one is then fully incorporated into the execution end; simultaneously, the phase boundary hit rate... Safety margin factor By converging boundary strengths, the same set of boundary formulas can support both active fluid implementations and passive thermal resistance implementations, while maintaining a safe boundary consistent with step two within each execution partition.

[0069] Step 4: At the target heat transfer range threshold After being transformed into a partitioned, two-sided differentiated thermal boundary, the phase-locked margin is continuously monitored using the same temperature-stress object. Once the current process temperature... Approaching the lower limit temperature of the window Or the current equivalent tensile stress Approaching the safety envelope It immediately triggers the cliff-locked phase-locked fuse protection mechanism, cancels the asymmetric boundary, restores synchronous cooling, and freezes the macroscopic anti-bending rate and interface micro fingerprint at room temperature.

[0070] This step is executed by the phase-locked fuse control unit. The phase-locked fuse control unit consists of a safety PLC, an FPGA phase-locked board, a hardware relay array, a dual-channel isolated power supply, a first-side flow equalization valve group, a second-side flow equalization valve group, a common-mode bypass air curtain, a temperature equalization buffer chamber door, an acoustic emission probe array, a thermal image feedback interface, and a room temperature output detection interface.

[0071] The safety PLC receives the control frame sent by the thermal boundary execution control unit in step three. The control frame contains at least the board number and the current process temperature. Current equivalent tensile stress First side boundary command Second side boundary command Execution partition number And a cyclic redundancy check field; the FPGA phase-locked loop board directly receives the high-frequency voltage signal output from the acoustic emission probe array and the temperature-stress boundary signal output from the safety PLC. The two signals are parallel inputs rather than serial waiting relationships. The reason for this setting is that step four is not a normal shutdown, but a hard switch to the differentiated strong cooling action in the front end. If all information is still bypassed by the upper industrial control computer, the prestress formed in the front end will be pulled into a dangerous peak by the continued amplification of the temperature difference on both sides within tens of milliseconds when the temperature continues to drop and the material slip window is about to close.

[0072] After the plates are removed from the room temperature tray, the macroscopic inverse curvature is measured using the three-point laser profilometry method; at the reference span Inside, take the chord height of the board And calculate: Among them, macroscopic anti-bending rate Used to characterize the pre-set bending degree of the plate in the cold state.

[0073] For microstructural fingerprints, a cross-section was prepared at the interface between the thick and thin metal residue sides. The average thickness of the intermetallic compound layer was obtained by scanning electron microscopy and energy dispersive spectroscopy. The average thickness of the first side was then recorded. Average thickness of the second side And calculate the thickness ratio: As a quantitative indicator of the inverted characteristics of microstructures after asymmetric quenching.

[0074] Step three, the spatially differentiated thermodynamic intervention boundary, is essentially a short-term leveraging action. It utilizes the slippage capability retained by the surface metal within the phase transition window to preemptively shift the bending moment that might accumulate in the future service state to the manufacturing state for offsetting. Once this action exceeds the lower limit temperature of the window... As the slip margin of the metal layer continues to narrow rapidly, the ceramic dielectric, brazing layer, and brittle intermetallic compound layer begin to bear more direct stress transfer; at this point, the temperature difference driving force originally used for hedging will become the source of crack initiation.

[0075] Therefore, in this step, the phase-locked fuse control unit no longer pursues a larger reverse disturbance, but instead focuses on how much of the already applied prestress can be safely retained. To this end, the phase-locked fuse control unit adjusts the safety envelope of step two. The phase boundary temperature parameters from step three and the bilateral boundary commands being executed in step three are incorporated into the same decision chain to continuously calculate the phase-locked margin. This ensures that subsequent depressurization actions are triggered based on the minimum safety margin, rather than using average temperature or fixed time as a rough stopping condition.

[0076] In each control cycle, the phase-locked fuse control unit first reads the current process temperature written by the safety PLC. and current equivalent tensile stress Then read the upper limit temperature of the window from the local parameter area. Lower limit temperature of the window and safety envelope The segmented nodes. Subsequently, the safety PLC does not assign these quantities to different shutdown logics separately, but instead first compresses the temperature margin and stress margin into a single phase-locked loop (PLL) margin. This approach is necessary because step four deals with a dual-approach scenario: some batches first become dangerous due to temperature exceeding limits, while others become dangerous due to stress exceeding limits. If two independent time-delay relays are still used, the first path to reach the danger may have already rendered the other path irreversible. Phase-locked loop margin. The specific form is as follows: Where: Phase-locked margin The minimum safety margin that the current substrate can retain between the stress boundary and the temperature boundary. This is a scalar value that can be positive or negative; the closer the value is to zero, the closer it is to triggering a hard switch. Safety envelope. The temperature-dependent stress allowable boundary generated in step two is a temperature function with a value greater than 0. Current equivalent tensile stress The transient tensile stress obtained from the model in step two and the on-site temperature calculation in step three is taken as a stress value greater than or equal to 0; process temperature : The current substrate zone or representative point of the entire board temperature, taken as the process temperature during annealing cooling; lower limit temperature of the window. The lower boundary temperature at which the material system transitions from viscoplastic rheology to elastic brittleness is a value lower than the upper limit of the window temperature. The material parameters are used to define the minimum temperature limit at which the asymmetric boundary can be applied; the upper limit temperature window. The upper boundary temperature at which the material system enters the phase transition window, which is higher than the lower boundary temperature of the window. Material parameters used to normalize the current process temperature. The window position it is in; Stable small quantity : A fixed constant used to maintain the continuity of the denominator, whose value is greater than 0 and much smaller than the boundary principal quantities; Minimum operator Choose the smaller of the stress margin and the temperature margin as the phase-locked loop margin. This is used to ensure that triggering decisions are controlled by the weakest boundary. After completing the phase-locked margin After calculation, the FPGA phase-locked loop continues to connect the bandpass energy signal of the acoustic emission probe array in parallel. The acoustic emission probe array is preferably installed on the rigid sound transmission path between the bottom surface of the support fixture and the side arm of the pressure plate. The probe tip uses an alumina wear-resistant pad, and the space between the probe and the fixture is filled with silicon-based coupling grease to suppress spurious pulses caused by loose contact. The bandpass frequency band preferably covers 200 to 500 kHz, but a broadband acoustic probe with equivalent function can also be used.

[0077] The FPGA phase-locked loop (PLL) performs rectification, envelope extraction, and short-window integration on the bandpass signal to generate the acoustic emission transition ratio. This is then combined with the threshold to form the circuit breaker trigger factor. : Where: Circuit breaker trigger factor The final trigger strength obtained by the phase-locked fuse control unit after integrating acoustic precursors and boundary margins is a scalar value greater than 0. When this value reaches the trigger threshold, hardware pass-through fuse is initiated; acoustic emission transition ratio. The ratio of the short-window energy to the stationary baseline energy measured by the acoustic emission probe array within the target frequency band is a scalar with a value greater than or equal to 0, used to characterize the strength of the high-frequency energy surge before microcrack initiation. Threshold The acoustic trigger threshold, calibrated on the test board and written into the FPGA phase-locked loop (PLL) board, is a scalar value greater than 0, used to separate environmental noise from actual crack precursors; phase-locked loop margin. Using the previous definition, this method directly incorporates the degree of temperature-stress boundary approximation into the fuse triggering process; stabilizing small quantities The previous definition is used to prevent mathematical singularities from occurring near minimal boundary values; When the circuit breaker trigger factor Once the preset threshold is reached, the FPGA phase-locked loop (PLL) board no longer performs calculations by the industrial control computer, but instead directly pulls up the emergency stop contacts of the hardware relay array.

[0078] In the active fluid implementation, the first-side flow equalization valve group and the second-side flow equalization valve group are simultaneously switched to the common-mode opening, and the common-mode bypass air curtain is opened. Previously, the flow was controlled by the first-side boundary command. Second side boundary instructions The resulting difference between the two sides is instantly smoothed out; in the passive thermal resistance implementation method, the temperature equalization buffer chamber door is opened, and the carrying fixture enters the low-convection temperature equalization buffer chamber at a predetermined conveying pace. The external heat flux is reduced overall, and the residual heat-insulating effect of the sacrificial mask no longer amplifies the temperature difference between the two sides. If the safety PLC loses two consecutive control frames or the cyclic redundancy check field fails twice consecutively, the phase-locked margin is also directly reduced. It is considered invalid and will be redirected to the same hardware fuse path.

[0079] Furthermore, phase-locked margin Compressing temperature and stress boundaries into a single decision value avoids hysteresis caused by multiple shutdown thresholds operating in parallel; circuit breaker trigger factor. The acoustic precursors are then connected in parallel to the hardware pass-through chain, so that step four can respond to both model predictions and microcrack precursor sounds, thereby stably locking the pressure release action before the phase transition window closes.

[0080] After the fuse trips, the phase-locked fuse control unit does not immediately send the substrate out. Instead, it allows the substrate to continue passing through the homogeneous temperature buffer zone, the low-gradient cooling zone, and the room temperature convergence zone under synchronous homogeneous cooling. Here, synchronous homogeneous cooling is not simply stopping the airflow, but rather unifying the remaining heat transfer conditions of the first and second metal layers to the same common mode state. This prevents the reverse disturbance already introduced from being amplified by the new bilateral difference, nor from being offset by localized reheating.

[0081] In the active fluid implementation, the safety PLC rewrites all execution zones to common mode valve positions in the first control cycle after the fuse is blown, and then reduces the total flow rate at a fixed beat, so that the temperature field inside the plate gradually converges along the thickness direction. In the passive thermal resistance implementation, the support fixture enters the co-flow cooling section after staying in the uniform temperature buffer chamber for a preset time. The airflow in the chamber changes from independent jets on both sides to a slow return flow parallel to the plate surface, so that both the side where the sacrificial mask is located and the exposed side fall into a low heat flux density environment.

[0082] In the preferred implementation, the phase-locked fuse control unit will also display the current process temperature of the previous cycle before the fuse trips. Current equivalent tensile stress Phase-locked margin Compared with acoustic emission transition ratio A board exit record is written and bound to the substrate number before being transmitted to the room temperature board exit detection interface. The room temperature board exit detection interface reads the macroscopic inverse bending rate, board corner warping direction, and cross-sectional sample retention mark. For batches requiring microscopic verification, the support fixture automatically sorts out the retention plates at the cooling endpoint. The retention plates are sent to the cross-sectional sample preparation station, where a cross-section is prepared at the junction of the first side thick metal residual area and the second side thin metal residual area. Subsequently, the thickness gradient and grain boundary morphology of the intermetallic compound layer are observed. Because step four cuts off the continuing increase of the bilateral temperature difference before the lower edge of the phase transition window, the long-range diffusion of interfacial active atoms is interrupted in advance, and the asymmetric structure distribution formed by the rapid cooling in the front section is preserved as a microscopic fingerprint at room temperature. At the same time, the reverse prestress that has been introduced at the macroscopic level is frozen in the board exit state in the form of a stable inverse bending rate. In engineering, the room temperature inverse bending rate, the consistency of the board surface warping direction within the tolerance zone, the thickness gradient of the cross-sectional intermetallic compound layer, and the correspondence between the acoustic emission trigger record and the actual retention position can be used as evaluation indicators.

[0083] For example, when a substrate with a large copper area concentrated in the center of the first metal layer and only retaining thin return lines at the edges of the second metal layer undergoes differentiated strong cooling in step three, the control interface first displays the current process temperature. It has already approached the lower limit of the window temperature. Subsequently, a high-frequency energy surge appeared on an acoustic emission probe located on the side arm of the fixture, and the FPGA phase-locked loop immediately triggered the hardware relay array. On-site observation showed that the first and second side flow equalization valve groups returned to the same opening almost simultaneously, and the common-mode bypass air curtain expanded on both sides of the board. The previously stronger heat extraction action in the central area immediately stopped. After the substrate entered the homogenization buffer chamber, the thermal image of the board surface changed from a bimodal gradient to a single-peak convergence. When exiting the board, the board surface maintained a reverse bending state in the predetermined direction, instead of flipping back at the end. The beneficial effect is that the synchronous homogenization and recovery action binds the sudden stop and retention to the same chain, both preventing the brittle fracture from continuing to propagate and preserving the reverse prestress already established in the previous stage. The room temperature board exit detection interface and the cross-sectional sampling path fix the macroscopic reverse bending rate and microstructural fingerprint into verifiable process results, facilitating subsequent specification and claims to support the product characteristics.

[0084] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0085] Those skilled in the art will understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0086] In the several embodiments provided in this application, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between apparatuses or units may be electrical, mechanical, or other forms.

[0087] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0088] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A method for stress balance control of a metal composite substrate, executed by a control device, characterized in that: include: Obtain the geometric topological parameters of the first and second metal layers of the composite substrate to be processed, and generate an asymmetric polarization tensor characterizing the imbalance between thermal capacity and stiffness on both sides. The asymmetric polarization tensor is input into the dynamic coupling feedforward model of the heat engine to solve the target heat transfer range threshold and generate a safety envelope. During the heat treatment cooling process, the body temperature of the composite substrate is continuously collected. When it enters the preset phase change window, a spatially differentiated thermodynamic intervention boundary is applied to the first side metal layer and the second side metal layer according to the target heat transfer difference threshold. During the application of the spatially differentiated thermodynamic intervention boundary, the cliff-locked phase-locked fuse protection mechanism is triggered in real time according to the safety envelope, the spatially differentiated thermodynamic intervention boundary is removed and synchronous homogenization is restored.

2. The method for stress balance control of metal composite substrates according to claim 1, characterized in that: Obtain geometric topology parameters, including: dividing the first side metal layer and the second side metal layer into multiple micro-regions based on the board coordinates of the composite substrate; extracting the metal area duty cycle, metal thickness information and connectivity state of each micro-region; and generating an asymmetric polarization tensor according to the area, thickness and connectivity state of each micro-region.

3. The method for stress balance control of metal composite substrates according to claim 2, characterized in that: The asymmetric polarization tensor is input into the dynamic coupling feedforward model of the heat engine, including: determining the main process direction based on the transport direction, the action direction of the actuator, and the plate direction; projecting the asymmetric polarization tensor along the main process direction; and calculating the target heat transfer range threshold based on the projection results, the target service junction temperature, and the reference cold end temperature.

4. The method for stress balance control of metal composite substrates according to claim 3, characterized in that: When an entity enters a preset phase transition window, it is identified and a spatially differentiated thermodynamic intervention boundary is applied, including: mapping the bulk temperature to multiple execution partitions; determining the phase boundary hit state and safety margin state for each execution partition; and applying the spatially differentiated thermodynamic intervention boundary only to execution partitions where the phase boundary hit is achieved and the safety margin is positive.

5. The method for stress balance control of metal composite substrates according to claim 1, characterized in that: After generating the target heat transfer range threshold and the safety envelope, the control device sends a control message to the thermal boundary execution control unit within the control device. The control message includes at least the substrate number, the asymmetric polarization tensor, the target heat transfer range threshold, the segment nodes of the safety envelope, and the index information of the preset phase change window.

6. The method for stress balance control of metal composite substrates according to claim 5, characterized in that: Based on the control message, the hot boundary execution control unit generates a first side boundary instruction and a second side boundary instruction for each execution partition. Each first side boundary instruction and each second side boundary instruction carries the corresponding execution partition number and is written into the partition control registers corresponding to the first side execution mechanism and the second side execution mechanism, respectively.

7. The method for stress balance control of metal composite substrates according to claim 6, characterized in that: During the application of spatially differentiated thermodynamic intervention boundaries, the control device sends a control frame to the phase-locked fuse control unit within the control device; the control frame includes at least the substrate number, body temperature, current equivalent tensile stress, each first side boundary command, each second side boundary command, execution partition number, time sequence number, and verification field.

8. The method for stress balance control of metal composite substrates according to claim 1, characterized in that: The control equipment includes an industrial camera, a high-frequency thermal imager, an acoustic emission probe array, a safety PLC, an FPGA phase-locked loop board, a first-side actuator, and a second-side actuator. The industrial camera acquires geometric topology parameters, the high-frequency thermal imager collects the body temperature, and the safety PLC and FPGA phase-locked loop board jointly execute the cliff-break phase-locked fuse protection mechanism.

9. The method for stress balance control of metal composite substrates according to claim 8, characterized in that: When the surface metal of the first and second side metal layers is oxygen-free copper, the preset phase change window corresponds to the cooling range of 380 degrees Celsius to 280 degrees Celsius; the control device maintains the spatially differentiated thermodynamic intervention boundary in an effective state only after the body temperature enters the cooling range from high to low.

10. The method for controlling stress balance in a metal composite substrate according to claim 8, characterized in that: When the control device is in at least one of the following states: insufficient edge computing power, network disconnection, or black boxing of drawings, it calls the one-dimensional equivalent thermal expansion bias lookup table in the local read-only memory, and uses the difference in theoretical copper cladding weight between the upper and lower metals and the asymmetric polarization tensor as the lookup table input, and outputs the target heat transfer range threshold and the segment node level of the safety envelope.

Images