A method and system for manufacturing a non-circular section core thermal zoned potted film capacitor

Abstract

This application provides a method and system for manufacturing a non-circular cross-section core thermally partitioned potted film capacitor. The method includes: determining the dielectric film thickness, sheet resistance of the metallized electrode, and winding tension window of the metallized film based on the obtained capacitor operating conditions; continuously winding the metallized film in multiple layers radially along the core axis of the non-circular structure to obtain an initial capacitor core with a non-circular cross-section; performing metal spraying and pressing treatment on the initial capacitor core with a non-circular cross-section to form end-face electrodes, and fixing the lead terminals to obtain a capacitor core with lead-out end-face electrodes; based on the capacitor core, obtaining the geometric characteristic parameters of the non-circular cross-section of the core, the number of layers of the metallized film, and the target operating frequency; performing thermal partitioning using a preset thermal partitioning intelligent discrimination neural network model to determine the partitioned distribution scheme of the potting material, thereby performing partitioned potting and constructing radial heat dissipation paths to obtain the final capacitor. This solution can improve thermal management efficiency.

Description

[0001] This application is a divisional application of the invention application filed on April 2, 2026, with Chinese application number 202610425618.5 and entitled "A method for manufacturing a non-circular cross-section core thermally partitioned potted film capacitor". Technical Field

[0002] This application relates to the field of capacitor manufacturing technology, specifically to a method and system for manufacturing a non-circular cross-section core thermally partitioned potted thin-film capacitor. Background Technology

[0003] When film capacitors operate under high-frequency, high-ripple-current conditions, the internal metallized film generates a large amount of heat due to dielectric and ohmic losses. If heat dissipation is not timely, it can easily lead to localized overheating, performance degradation, or even failure. Traditional circular core capacitors, due to their symmetrical structure and single potting material, have heat dissipation paths mainly concentrated radially, making it impossible to perform differentiated thermal management based on the actual heat distribution inside the core, thus limiting heat dissipation efficiency. Existing potting processes mostly use a single thermally conductive material for overall encapsulation, which can provide some heat dissipation and support, but fails to optimize zoning based on the actual heat generation characteristics of the core. This results in insufficient heat dissipation in high-heat-flux areas, material redundancy in low-temperature areas, and overall low thermal management efficiency. Summary of the Invention

[0004] This application aims to provide a method and system for manufacturing non-circular cross-section core thermally partitioned potted thin-film capacitors, which can improve thermal management efficiency.

[0005] The technical solution of this application is implemented as follows: This application provides a method for manufacturing a non-circular cross-section core thermally partitioned potted film capacitor, the method comprising: Based on the obtained capacitor operating conditions, the dielectric film thickness, sheet resistance of the metallized electrode, and winding tension window of the metallized thin film are determined. The metallized film is wound in multiple layers along the radial direction of the non-circular structure mandrel to obtain a non-circular cross-section metallized film winding; and the mandrel of the non-circular structure is removed to obtain an initial capacitor core with a non-circular cross-section. The initial capacitor core with the non-circular cross-section is subjected to metal spraying and pressing to form end face electrodes. While keeping the shape of the non-circular cross-section intact, the lead terminals are fixed to obtain a capacitor core with non-circular cross-section winding and end face electrode lead-out. Based on the capacitor core, the geometric feature parameters of the non-circular cross-section of the core, the number of layers of the metallized thin film, and the target operating frequency are obtained; Based on the geometric feature parameters of the non-circular cross-section of the core, the number of layers of the metallized film, the dielectric film thickness of the metallized film, and the target operating frequency, thermal zoning is performed through a preset thermal zoning intelligent discrimination neural network model to determine the zoning distribution scheme of the potting material; Based on the aforementioned potting material zoning distribution scheme, zoning potting and radial heat dissipation path construction are performed to obtain the final capacitor.

[0006] Understandably, the metallized film is continuously wound in multiple layers radially along the core of a non-circular structure to obtain a non-circular cross-section metallized film winding. The core of the non-circular structure is then removed, resulting in an initial capacitor core with a non-circular cross-section. Since a non-circular cross-section core with multiple convex contours is designed instead of a traditional circular core, the metallized film is wound to form a non-circular cross-section core with high-curvature convex and concave regions, providing a clear physical basis for subsequent zoned heat dissipation. Based on the obtained geometric characteristic parameters of the core's non-circular cross-section, the number of metallized film layers, the dielectric film thickness of the metallized film, and the target operating frequency, a preset intelligent thermal zoning neural network model is used to perform thermal zoning, determining the potting material zoning distribution scheme. This achieves automatic division of the thermal management functional zones on the outer surface of the core and intelligent output of the potting material distribution scheme, replacing the traditional experience-based material layout method. Based on the potting material zoning distribution scheme, zoned potting and radial heat dissipation path construction can improve overall heat dissipation efficiency and structural reliability, thereby enhancing thermal management performance.

[0007] In the above scheme, the step of determining the potting material distribution scheme by performing thermal zoning based on the geometric feature parameters of the non-circular cross-section of the core, the number of layers of the metallized film, the dielectric film thickness of the metallized film, and the target operating frequency through a preset thermal zoning intelligent discrimination neural network model includes: Multiple features are obtained by extracting features from the geometric feature parameters of the non-circular cross-section of the core, the number of layers of the metallized thin film, the dielectric film thickness of the metallized thin film, and the target operating frequency; The multiple features are fused to obtain the fused features; Thermal zoning is performed using the fusion features to determine the zoning distribution scheme of the potting material.

[0008] Understandably, based on multiple inputs such as the non-circular cross-sectional geometric features of the core, the number of metallized thin film layers, the dielectric film thickness of the metallized thin film, and the target operating frequency, the automatic division of the thermal management functional zones on the outer surface of the core and the intelligent output of the potting material distribution scheme can be achieved through a preset thermal zoning intelligent discrimination neural network model, which is beneficial for subsequent zoning potting.

[0009] In the above scheme, feature extraction is performed on the geometric feature parameters of the non-circular cross-section of the core, the number of layers of the metallized thin film, the dielectric film thickness of the metallized thin film, and the target operating frequency to obtain multiple features, including: The geometric features are obtained by extracting the geometric feature parameters of the non-circular cross-section of the core; The number of layers in the metallized thin film and the thickness of the dielectric film in the metallized thin film are used to extract features to obtain the stacking features; Feature extraction is performed on the target operating frequency to obtain frequency features; The geometric features, the stacked features, and the frequency features are determined as the plurality of features.

[0010] Understandably, by using different extraction subnetworks for different data, the extracted geometric features, layered features, and frequency features become more accurate.

[0011] In the above scheme, the feature fusion of the multiple features to obtain fused features includes: The multiple features are fused to obtain primary fused features; The attention weights are calculated based on the frequency features among the multiple features. Based on the attention weights, the primary fusion features are subjected to a second feature fusion to obtain the fusion features.

[0012] Understandably, multiple features are fused using the first feature to obtain a primary fused feature; attention weights are calculated using frequency features, and finally, the primary fused feature is fused using the attention weights to obtain a secondary fused feature; the addition of attention weights can improve the accuracy of the fused feature.

[0013] In the above scheme, the step of determining the zoning distribution scheme of the potting material by performing thermal zoning through the fusion feature includes: Thermal partitioning is performed using the aforementioned fusion features to determine high heat flux partitions and conventional heat flux partitions; Based on the high heat flux zone and the conventional heat flux zone, potting material is allocated to determine the potting material zone distribution scheme.

[0014] It is understandable that by fusing features to perform thermal partitioning, high heat flux partitioning and conventional heat flux partitioning can be obtained. Since the thermal partitioning is performed using a preset intelligent discrimination neural network model, the accuracy of thermal partitioning can be improved, which facilitates subsequent partitioning and potting, thereby improving the thermal management efficiency.

[0015] In the above scheme, the step of allocating potting materials based on the high heat flux zone and the conventional heat flux zone, and determining the potting material zoning distribution scheme, includes: Based on the high heat flux zone and the conventional heat flux zone, potting material is allocated, and the probability value corresponding to each of the high heat flux zone and the conventional heat flux zone is determined. If the probability value is greater than a preset threshold, then the partition corresponding to the probability value is determined to be the first potting material; If the probability value is less than or equal to the preset threshold, then the partition corresponding to the probability value is determined to be the second potting material; Based on the high heat flux zone, the conventional heat flux zone, the first potting material, and the second potting material, the zoning distribution scheme of the potting material is determined.

[0016] It is understandable that by determining the probability values ​​corresponding to high heat flux zones and conventional heat flux zones, the corresponding potting materials for each zone can be further determined, and then zoned potting can be carried out in the subsequent process, thereby improving the thermal management efficiency.

[0017] In the above scheme, the first potting material has a high thermal conductivity and is used to fill high heat flux zones; wherein, the high thermal conductivity is a thermal conductivity greater than a preset thermal conductivity threshold. The second potting material has a conventional thermal conductivity and is used to fill conventional heat flux zones and provide structural support; wherein, the conventional thermal conductivity is a thermal conductivity less than or equal to a preset thermal conductivity threshold.

[0018] Understandably, providing two different potting materials allows for zoned potting, thereby improving overall heat dissipation efficiency.

[0019] In the above scheme, the step of performing zoned filling and radial heat dissipation path construction based on the zoned distribution scheme of the potting material to obtain the final capacitor includes: Based on the first and second potting materials in the potting material zoning distribution scheme, the high heat flux zone and the conventional heat flux zone are potted; After the potting process is completed, the first potting material and the second potting material are cured to obtain the radial main heat conduction path and the structural support path. Based on the radial main heat conduction path and the structural support path, a radial heat dissipation path is constructed to obtain the final capacitor.

[0020] It is understandable that by using a first potting material with high thermal conductivity and a second potting material with conventional thermal conductivity, and by performing zonal potting according to the zonal distribution scheme of the potting materials, the high thermal conductivity material is concentrated in the high heat flux area to form the radial main heat conduction path, while the conventional material is filled in the recessed area to provide support and auxiliary heat dissipation, thereby improving the overall heat dissipation efficiency.

[0021] In the above scheme, the metallized thin film includes a dielectric thin film and a metallized electrode layer disposed on at least one surface of the dielectric thin film.

[0022] In the above scheme, the dielectric film is a polypropylene film and / or a polyester film; the metallized electrode layer is formed by vacuum evaporation.

[0023] This application provides a manufacturing system for a non-circular cross-section core thermally partitioned potted thin-film capacitor, characterized in that the system is configured as follows: Based on the obtained capacitor operating conditions, the dielectric film thickness, sheet resistance of the metallized electrode, and winding tension window of the metallized thin film are determined. The metallized film is wound in multiple layers along the radial direction of the non-circular structure mandrel to obtain a metallized film winding body with a non-circular cross-section; and the mandrel of the non-circular structure is removed to obtain an initial capacitor core with a non-circular cross-section. The initial capacitor core with a non-circular cross-section is subjected to metal spraying and pressing to form end face electrodes, and the lead-out terminals are fixed to obtain a capacitor core with a non-circular cross-section winding and end face electrode lead-out. Based on the capacitor core, the geometric feature parameters of the non-circular cross-section of the core, the number of layers of the metallized thin film, and the target operating frequency are obtained; Based on the geometric feature parameters of the non-circular cross-section of the core, the number of layers of the metallized film, the dielectric film thickness of the metallized film, and the target operating frequency, thermal zoning is performed through a preset thermal zoning intelligent discrimination neural network model to determine the zoning distribution scheme of the potting material; Based on the aforementioned potting material zoning distribution scheme, zoning potting and radial heat dissipation path construction are performed to obtain the final capacitor.

[0024] This application provides a method for manufacturing a non-circular cross-section core thermally partitioned potted film capacitor. The method includes: determining the dielectric film thickness, sheet resistance of the metallized electrode, and winding tension window of the metallized film based on the obtained capacitor operating conditions; continuously winding the metallized film in multiple layers along the radial direction of the core of the non-circular structure to obtain a non-circular cross-section metallized film winding body; removing the core of the non-circular structure to obtain an initial non-circular cross-section capacitor core; and performing metal spraying and pressing treatment on the initial non-circular cross-section capacitor core to form end-face electrodes, while maintaining the non-circular cross-section shape. After fixing the leads, a capacitor core with a non-circular cross-section and end-face electrodes is obtained. Based on the capacitor core, the geometric characteristic parameters of the non-circular cross-section, the number of layers of the metallized film, and the target operating frequency are obtained. Based on the geometric characteristic parameters of the non-circular cross-section, the number of layers of the metallized film, the dielectric film thickness of the metallized film, and the target operating frequency, thermal partitioning is performed using a preset thermal partitioning intelligent discrimination neural network model to determine the partitioning distribution scheme of the potting material. Based on the partitioning distribution scheme of the potting material, partitioned potting and radial heat dissipation path construction are performed to obtain the final capacitor. In the above scheme, the metallized film is continuously wound in multiple layers radially along the core axis of the non-circular structure to obtain a non-circular cross-section metallized film winding body. The core axis of the non-circular structure is removed to obtain the initial capacitor core with a non-circular cross-section. Since a non-circular cross-section core with multiple convex contours is designed instead of a traditional circular core axis, the metallized film is formed into a non-circular cross-section core with high curvature convex and concave areas through winding, providing a clear physical partitioning basis for subsequent partitioned heat dissipation. Based on the acquired geometric features of the non-circular cross-section of the core, the number of metallized film layers, the dielectric film thickness of the metallized film, and the target operating frequency, a pre-set intelligent thermal zoning neural network model is used to perform thermal zoning and determine the zoning distribution scheme of the potting material. This achieves automatic division of the thermal management functional zones on the outer surface of the core and intelligent output of the potting material distribution scheme, replacing the traditional experience-based material layout method. From film selection, non-circular winding, end-face treatment to zoning potting and co-curing, the process parameters of each step are matched to ensure that the core structure remains stable during manufacturing. After encapsulation, the potting material and heat dissipation path form an integrated, highly efficient heat dissipation system, thereby improving thermal management efficiency. Attached Figure Description

[0025] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the specification, serve to explain the technical solutions of this application. Obviously, the drawings described below are merely some embodiments of this application, and those skilled in the art can obtain other drawings based on these drawings without any inventive effort.

[0026] The flowcharts shown in the accompanying drawings are merely illustrative and do not necessarily include all content and operations / steps, nor do they necessarily have to be performed in the described order. For example, some operations / steps can be broken down, while others can be combined or partially combined; therefore, the actual execution order may change depending on the specific circumstances.

[0027] Figure 1 A schematic flowchart illustrating a method for manufacturing a non-circular cross-section core thermally partitioned potted thin-film capacitor according to an embodiment of this application; Figure 2 This is a schematic diagram of the partitioning of a method for manufacturing a non-circular cross-section core thermally partitioned potted film capacitor, provided in an embodiment of this application. Detailed Implementation

[0028] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the specific technical solutions of this application will be further described in detail below with reference to the accompanying drawings of the embodiments of this application. The following embodiments are used to illustrate this application, but are not intended to limit the scope of this application.

[0029] Unless otherwise defined, all technical and scientific terms used in this application have the same meaning as commonly understood by one of ordinary skill in the art to which this application pertains. The terminology used in this application is for the purpose of describing embodiments of this application only and is not intended to be limiting of this application.

[0030] In the following description, references to "some embodiments," "this embodiment," "this application embodiment," and examples, etc., describe a subset of all possible embodiments. However, it is understood that "some embodiments" may be the same subset or different subset of all possible embodiments and may be combined with each other without conflict.

[0031] If the application documents contain similar descriptions such as "first / second", the following explanation shall be added: In the following description, the terms "first / second / third" are used only to distinguish similar objects and do not represent a specific order of objects. It is understood that "first / second / third" may be interchanged in a specific order or sequence where permitted, so that the embodiments of this application described herein can be implemented in an order other than that illustrated or described herein.

[0032] Based on this, embodiments of this application provide a method for manufacturing a non-circular cross-section core thermally partitioned potted thin-film capacitor. Figure 1 This is a schematic flowchart illustrating a method for manufacturing a non-circular cross-section core thermally partitioned potted film capacitor according to an embodiment of this application, which will be combined with... Figure 1 The steps shown are explained.

[0033] S101. Based on the obtained capacitor operating conditions, determine the dielectric film thickness, sheet resistance of the metallized electrode, and winding tension window of the metallized film.

[0034] In some embodiments of this application, the metallized thin film includes a dielectric thin film and a metallized electrode layer disposed on at least one surface of the dielectric thin film. The dielectric thin film is a polypropylene film and / or a polyester film; the metallized electrode layer is formed by vacuum evaporation.

[0035] In some embodiments of this application, a metallized thin film is provided, the metallized thin film including a dielectric thin film and a metallized electrode layer disposed on at least one surface of the dielectric thin film; wherein the dielectric thin film is a polypropylene film, a polyester film or a combination thereof, and the metallized electrode layer is formed by vacuum evaporation.

[0036] In some embodiments of this application, a method for manufacturing a non-circular cross-section core thermally partitioned potting film capacitor is adapted to capacitor manufacturing scenarios.

[0037] In some embodiments of this application, the operating conditions of the obtained capacitor are used to preset the following key parameters of the metallization film based on the target application of the capacitor in high-frequency, high-ripple-current conditions, so that the heat source distribution inside the core after winding is discriminable: Dielectric film thickness: Select a thickness that helps reduce high-frequency losses while meeting the withstand voltage requirements; Metallized electrode sheet resistance: Optimized configuration to balance conductivity and self-healing ability, suppressing the generation of local hot spots; Winding tension window: determined based on film parameters and non-circular cross-section core structure, ensuring that the film layer does not break or wrinkle during winding, and can form a stable non-circular cross-section stacked structure.

[0038] For example, for film capacitors intended for high-frequency, high-ripple-current applications (operating frequency 50 kHz to 200 kHz), polypropylene (PP) dielectric film is selected as the dielectric layer material, with a single-layer thickness of 6 μm. An aluminum metallized electrode layer is formed on at least one surface of the dielectric film using a vacuum evaporation process. The sheet resistance of the metallized electrode is controlled within the range of 8 to 15 ohms per square meter to balance conductivity and self-healing properties. Simultaneously, based on the geometric profile characteristics of the subsequent non-circular cross-section mandrel and the stability requirements of winding, the winding tension of the metallized film is preset to a tension window range of 2.5 to 4.0 N. This allows the film to form relatively tight stacks in the raised areas and relatively loose stacks in adjacent recessed areas during winding, thus providing basic electrothermal parameters for the stable formation of the subsequent three-dimensional radial heat dissipation network of the core and the determination of thermal zones.

[0039] S102. The metallized film is continuously wound in multiple layers along the radial direction of the non-circular structure mandrel to obtain a non-circular cross-section metallized film winding body; and the non-circular structure mandrel is removed to obtain the initial capacitor core with a non-circular cross-section.

[0040] In some embodiments of this application, a mandrel extending axially and having a non-circular cross-section is provided. The non-circular structure includes at least three circumferentially spaced protrusions that extend continuously axially, forming an axially continuous non-circular guide profile on the outer surface of the mandrel. Under constant winding tension control, the pre-treated metallized film is continuously wound in multiple layers radially along the mandrel. Through the non-circular geometric constraint of the mandrel, the film forms different degrees of lamination tightness between the protrusions and adjacent recesses, thereby obtaining a metallized film winding with a non-circular cross-section. After winding, the mandrel is removed axially. After the internal support is removed, the metallized film winding forms a stable non-circular cross-sectional shape based on its own lamination stress differences and geometric deformation, ultimately obtaining an initial capacitor core with a three-dimensional radial heat dissipation network.

[0041] S103. The initial capacitor core with a non-circular cross-section is metal-sprayed and crimped to form end-face electrodes. While keeping the non-circular cross-section shape intact, the lead-out terminals are fixed to obtain a capacitor core with a non-circular cross-section winding and end-face electrode lead-out.

[0042] In some embodiments of this application, the end face of the core is metal sprayed or pressed to form an end face electrode; while keeping the non-circular cross-section shape intact, the lead-out terminals are fixed to obtain a capacitor core with non-circular cross-section winding and end face electrode lead-out.

[0043] S104. Based on the capacitor core, obtain the geometric characteristic parameters of the non-circular cross-section of the core, the number of metallized thin film layers, and the target operating frequency.

[0044] In some embodiments of this application, based on the capacitor core, the geometric feature parameters of the non-circular cross-section of the core and the number of metallization film layers are extracted. Simultaneously, the target operating frequency is obtained.

[0045] S105. Based on the geometric characteristic parameters of the non-circular cross-section of the core, the number of metallized film layers, the dielectric film thickness of the metallized film, and the target operating frequency, thermal zoning is performed through a preset thermal zoning intelligent discrimination neural network model to determine the zoning distribution scheme of the potting material.

[0046] In some embodiments of this application, the feature extraction module in the preset thermal partitioning intelligent discrimination neural network model extracts features from the geometric feature parameters of the non-circular cross-section of the core, the number of metallized film layers, the dielectric film thickness of the metallized film, and the target operating frequency to obtain multiple features; the feature fusion module in the preset thermal partitioning intelligent discrimination neural network model fuses these multiple features to obtain fused features; and the thermal partitioning discrimination output module in the preset thermal partitioning intelligent discrimination neural network model thermal partitions the fused features to determine the partitioning distribution scheme of the potting material.

[0047] S106. Based on the partitioned distribution scheme of the potting material, partitioned potting and radial heat dissipation path construction are carried out to obtain the final capacitor.

[0048] In some embodiments of this application, high heat flux zones and conventional heat flux zones are potted based on the first and second potting materials in the potting material zoning distribution scheme; after potting is completed, the first and second potting materials are cured to obtain a radial main heat conduction path and a structural support path; based on the radial main heat conduction path and the structural support path, a radial heat dissipation path is constructed to obtain the final capacitor.

[0049] In some embodiments of this application, the first potting material has a high thermal conductivity and is used to fill high heat flux zones; wherein, the high thermal conductivity is a thermal conductivity greater than a preset thermal conductivity threshold; the second potting material has a conventional thermal conductivity and is used to fill conventional heat flux zones and provide structural support; wherein, the conventional thermal conductivity is a thermal conductivity less than or equal to a preset thermal conductivity threshold.

[0050] For example, at least two potting materials with different thermal conductivity are provided, including: First potting material: with high thermal conductivity, ≥ 1.0 W / (m•K), used for subsequent filling of high heat flux zones; The second potting material has a conventional thermal conductivity and is used for subsequent filling of conventional heat flux zones and providing structural support.

[0051] The dynamic viscosity difference between the two materials at the injection temperature is assumed. The first potting material should have a high thixotropic index to ensure that it can maintain its shape after high heat flux zonal injection and does not produce serious circumferential overflow.

[0052] The specific values ​​of the potting material parameters, such as "the preset thermal conductivity threshold is 1.0 W / (m•K), the thermal conductivity of the first potting material is 1.2-2.0 W / (m•K), and the thermal conductivity of the second potting material is 0.3-0.6 W / (m•K)".

[0053] For capacitor cores with completed non-circular cross-section winding and end-face electrode leads, two potting materials with different thermal conductivity and rheological properties are pre-configured based on the non-circular geometric features of their outer surface, which consists of raised and recessed areas. The first potting material is a high thermal conductivity filler-modified epoxy potting material with a thermal conductivity set to be no less than 1.0 W / (m•K) and exhibiting high thixotropic properties at the filling temperature to ensure stable morphology when filling high heat flux zones. The second potting material is a conventional thermal conductivity epoxy or silicon-based potting material with a lower thermal conductivity than the first potting material, used to fill conventional heat flux zones and provide overall structural support. By pre-setting the thermal conductivity parameters and rheological properties of these two potting materials, they can form clear functional zone boundaries within the same encapsulation cavity during subsequent zone filling, and match the corresponding thermal management functional zones on the outer surface of the core after curing. This provides a stable and controllable material basis for subsequent dual-material co-filling encapsulation based on intelligent thermal zone discrimination.

[0054] Based on the zoning distribution scheme of the potting material, the encapsulation gap between the core and the capacitor shell is subjected to zoning potting treatment. Potting materials with different thermal conductivity are filled into different thermal management functional zones, specifically including: The first potting material (high thermal conductivity) is filled into the high heat flux zone, so that after potting and curing, it is closely attached to the raised area formed by the non-circular cross-section raised structure on the outer surface of the core; A second potting material (with conventional thermal conductivity) is filled within the conventional heat flux zone to fill the recessed areas between adjacent raised structures and to provide overall structural support for the core.

[0055] After the filling is completed, the first potting material and the second potting material are cured so that after curing, they respectively maintain the spatial shape of the corresponding thermal management functional area and together form a continuous and stable encapsulation structure.

[0056] Through the partitioned pouring and curing process, multiple radial main heat conduction paths are formed in the circumferentially distributed raised areas on the outer surface of the core, pointing from the inside of the core towards the capacitor shell. Auxiliary heat conduction and structural support paths are formed in the recessed areas, thereby constructing a multi-path parallel radial heat dissipation network.

[0057] The radial heat dissipation network enables the heat generated by the core during operation to be conducted in parallel to the capacitor casing through the corresponding potting material path according to the heat flux difference of each thermal management functional zone, thereby achieving heat diversion and efficient dissipation.

[0058] Understandably, the metallized film is continuously wound in multiple layers radially along the core of a non-circular structure to obtain a non-circular cross-section metallized film winding. The core of the non-circular structure is then removed, resulting in an initial capacitor core with a non-circular cross-section. Since a non-circular cross-section core with multiple convex profiles is designed instead of a traditional circular core, the metallized film is wound to form a non-circular cross-section core with high-curvature convex and concave areas, providing a clear physical basis for subsequent zoned heat dissipation. Based on the obtained geometric characteristic parameters of the core's non-circular cross-section, the number of metallized film layers, the dielectric film thickness of the metallized film, and the target operating frequency, a preset intelligent thermal zoning neural network model is used to perform thermal zoning, determine the potting material distribution scheme, and achieve automatic division of the thermal management functional zones on the outer surface of the core and intelligent output of the potting material distribution scheme, replacing the traditional experience-based material layout method. From film selection, non-circular winding, end-face treatment to zoned potting and co-curing, the process parameters of each step are matched to ensure that the core structure remains stable during manufacturing. The potting material and heat dissipation path form an integrated high-efficiency heat dissipation system after encapsulation, thereby improving thermal management efficiency.

[0059] In some embodiments of this application, S105 can be implemented by S201-S203, as follows: S201. Through the feature extraction module in the preset hot zone intelligent discrimination neural network model, the geometric feature parameters of the non-circular cross section of the core, the number of metallized thin film layers, the dielectric film thickness of the metallized thin film, and the target operating frequency are extracted to obtain multiple features.

[0060] In some embodiments of this application, the feature extraction module includes a geometric feature extraction subnetwork, a stacked feature extraction subnetwork, and a frequency feature extraction subnetwork.

[0061] In some embodiments of this application, geometric features are obtained by extracting geometric feature parameters of the non-circular cross-section of the core through the geometric feature extraction subnetwork in the preset thermal partition intelligent discrimination neural network model; geometric features are obtained by extracting the number of metallized thin film layers and the dielectric film thickness of the metallized thin film through the stacking feature extraction subnetwork in the preset thermal partition intelligent discrimination neural network model; and frequency features are obtained by extracting the target operating frequency through the frequency feature extraction subnetwork in the preset thermal partition intelligent discrimination neural network model; and geometric features, stacking features, and frequency features are determined as multiple features.

[0062] S202. By using the feature fusion module in the preset hot partition intelligent discrimination neural network model, multiple features are fused to obtain fused features.

[0063] In some embodiments of this application, a feature fusion module in a preset hot-partition intelligent discrimination neural network model performs a first feature fusion on multiple features to obtain a primary fused feature; based on the frequency features among the multiple features, an attention weight is calculated; based on the attention weight, a second feature fusion is performed on the primary fused feature to obtain a fused feature.

[0064] S203. Based on the thermal partitioning discrimination output module in the preset thermal partitioning intelligent discrimination neural network model, the fused features are thermally partitioned to determine the partitioning distribution scheme of the potting material.

[0065] In some embodiments of this application, the thermal partitioning discrimination output module in the preset thermal partitioning intelligent discrimination neural network model performs thermal partitioning by fusing features to determine high heat flux partitions and conventional heat flux partitions; based on the high heat flux partitions and conventional heat flux partitions, potting material is allocated to determine the potting material partitioning distribution scheme.

[0066] In some embodiments of this application, potting materials are allocated based on high heat flux zones and conventional heat flux zones, and the probability values ​​corresponding to the high heat flux zones and conventional heat flux zones are determined. If the probability value is greater than a preset threshold, the zone corresponding to the probability value is determined to be the first potting material. If the probability value is less than or equal to the preset threshold, the zone corresponding to the probability value is determined to be the second potting material. Based on the high heat flux zones, conventional heat flux zones, the first potting material, and the second potting material, a potting material zoning distribution scheme is determined.

[0067] In some embodiments of this application, a non-circular cross-section core with completed end-face electrode lead-out processing is installed inside a capacitor casing, forming a circumferential encapsulation gap between the outer surface of the core and the inner wall of the casing. Under known target operating parameters, the heating behavior of the core is modeled and analyzed. Combined with the core's structural features, the relative heat generation distribution characteristics of the core's outer surface at different circumferential and axial positions are obtained. Based on these heat generation distribution characteristics, a thermal zoning intelligent discrimination model is constructed to divide the core's outer surface into thermal management functional zones along the circumferential and axial directions, such as... Figure 2 As shown, thermal management functional zoning refers to thermal zones divided based on relative heat flux differences, including high heat flux zones and conventional heat flux zones (i.e., Figure 2 The shaded area represents the high heat flux zone, and the blank area represents the regular heat flux zone. The thermal management function zones should include at least: 1. High heat flux zone, corresponding to the raised area formed by non-circular cross-section raised structure on the outer surface of the core; 2. Conventional heat flux zoning corresponds to the recessed areas on the outer surface of the core located between adjacent protruding structures; The preset hot zone intelligent discrimination model makes a comprehensive judgment based on at least the following input parameters: 1. Geometric characteristic parameters of the non-circular cross-section of the core; 2. Number of metallized thin films and thickness of each layer; 3. Target operating frequency.

[0068] The geometric features of the non-circular cross-section of the core, the number of metallized film layers, the thickness of a single layer, and the target operating frequency are extracted using a pre-defined intelligent thermal zoning model, resulting in multiple features. These features are then fused to obtain fused features. Thermal zoning is applied to the fused features to obtain intelligent thermal zoning discrimination results. Based on these results, a distribution scheme for the potting material in each thermal management functional zone on the outer surface of the core is generated. This distribution scheme indicates the pre-determined distribution relationship between the first and second potting materials in each thermal management functional zone, where: The first potting material is a high thermal conductivity potting material, used to fill high heat flux zones; The second potting material is a conventional thermal conductivity potting material used to fill conventional heat flux zones; wherein, the conventional thermal conductivity is a thermal conductivity less than or equal to a preset thermal conductivity threshold.

[0069] In some embodiments of this application, the preset hot zone intelligent discrimination neural network model includes an input layer, a feature extraction module, a feature fusion module, and a hot zone discrimination output module. The input layer contains three independent input channels: a geometric feature channel, a stack thickness channel, and an operating frequency channel. The input to the geometric feature channel is the geometric parameters of the non-circular cross-section of the core, with an input size of 360×4. The geometric parameters of the non-circular cross-section of the core include the radius of curvature, tangent angle, radial height, and local curvature. The geometric parameters of the non-circular cross-section of the core are normalized to the range [0,1], and the output size is 1×1440. The input to the stack thickness channel includes input 1 and input 2. Input 1 is the total number of metallized film layers N. The total number of metallized film layers N is divided by the maximum designed number of layers (e.g., 1000 layers) and normalized to [0,1]. The output size is 1×1. Input 2 is the thickness sequence, i.e., the thickness of each film layer (sequence length is min(N, 80)). If N<80, it is filled with 0; if N>80, it is uniformly sampled to 80 points. The thickness sequence is Z-score normalized, and the output size is 1×80. The input to the operating frequency channel is the operating frequency (Hz). The operating frequency is logarithmically normalized, and the output size is 1×1.

[0070] The lightweight feature extraction module includes a geometric feature extraction subnetwork, a stacked feature extraction subnetwork, and a frequency feature extraction subnetwork. The input layer of the geometric feature extraction subnetwork transforms the 1440-dimensional structure into a 30×48 grid (i.e., evenly distributing 360 sampling points across 30 columns, with each column corresponding to 48 sampling points). Through a 3×3 convolution in convolutional layer 1, four filters process the 30×48 grid to obtain a 30×48×4 vector. Through a 3×3 depthwise convolution, eight filters perform depthwise separable convolution to obtain a 30×48×8 vector. Global average pooling is then applied to the 30×48×8 vector to obtain an 8-dimensional feature vector, i.e., the geometric features. The layered feature extraction subnetwork is a two-branch parallel processing network: a total layer count branch and a thickness sequence branch. The total layer count branch takes a 1-dimensional total layer count as input and converts it into a 4-dimensional concatenated layer count feature through a fully connected layer. The thickness sequence branch takes an 80-dimensional thickness sequence as input and converts it into a 76×4 vector through a 1-dimensional convolutional layer 1 with a convolution kernel 5 and 4 filters. The 76×4 vector is then max-pooled with a pooling size of 2 to obtain a 38×4 vector. The 38×4 vector is then converted into a 36×8 vector through a 1-dimensional convolutional layer 2 with a convolution kernel 3 and 8 filters. The 36×8 vector is then subjected to global average pooling to obtain an 8-dimensional thickness feature. Finally, the concatenated layer count feature and the thickness feature are fused together through a fully connected layer to obtain an 8-dimensional layered feature. The frequency feature extraction subnetwork calculates [log10(f), f², 1 / f, sin(2πf / 1e6)] to transform the 1-dimensional target working frequency into a 5-dimensional vector; then, through a fully connected layer, the 5-dimensional vector is transformed into 4-dimensional frequency features.

[0071] The feature fusion module includes first-level fusion and second-level fusion. First-level fusion fuses concatenated geometric features, stacked features, and frequency features to obtain a 20-dimensional vector. A fully connected layer then transforms this 20-dimensional vector into a 12-dimensional primary fused feature. Second-level fusion introduces a frequency-guided attention mechanism, calculating attention weights based on frequency features (primarily converting 4-dimensional frequency features into a 12-dimensional weight vector). Gating function: Attention weights = Sigmoid(W·frequency feature + b); Weighted features = Primary fusion features ⊙ Attention weights; The 12-dimensional primary fusion features are transformed into 10-dimensional fusion features through a fully connected layer.

[0072] The core design of the thermal zoning output module is to directly output the material distribution scheme. The input to the thermal zoning output module is 10-dimensional fused features. The output layers include output layer 1 and output layer 2. Output layer 1 outputs the probability distribution of using high thermal conductivity materials in each circumferential zone, and output layer 2 outputs the probability distribution of using high thermal conductivity materials in each axial segment. The potting material zoning distribution scheme is determined through the outputs of output layer 1 and output layer 2. The potting material zoning distribution scheme can be represented by a material distribution matrix; where the matrix elements represent the probability of using high thermal conductivity materials (0-1); a probability > 0.5 indicates the use of high thermal conductivity materials, otherwise conventional materials are used.

[0073] A Neural Network Application Example of a Dual-Material Co-Pouring Encapsulation Step Based on Intelligent Thermal Partitioning: High-Frequency Automotive Thin-Film Capacitors. Suppose we need to perform thermal partitioning analysis on a thin-film capacitor used in a new energy vehicle motor driver. The specific scenario is as follows: 1. Input data preprocessing.

[0074] Geometric feature input: The core adopts a three-lobed non-circular structure (3 raised areas and 3 recessed areas). The neural network receives curvature and height data from 360 sampling points.

[0075] Stacking parameter input: Total number of layers N=800 layers, single layer thickness of dielectric film is 6μm.

[0076] Operating frequency input: The operating frequency is 150 kHz (which falls within the high-frequency range of 50 kHz to 200 kHz).

[0077] 2. Feature extraction and fusion process.

[0078] Frequency guidance: The model identifies an extremely high frequency of 150 kHz and automatically increases the weight of the "high curvature bulge" feature through the "frequency guidance attention mechanism" because the skin effect and dielectric loss at high frequencies will cause the heat flux of the bulge to increase sharply.

[0079] Feature refinement: Based on the layer thickness, the model calculates that the central region (mid-axis section) has the highest heat dissipation requirement due to the thermal superposition effect.

[0080] 3. Output: The final output of the two-dimensional material distribution matrix model is a 6x3 matrix (6 circumferential sections and 3 axial sections), as shown in Table 1.

[0081] Table 1 In Table 1, a value of P > 0.5 indicates the use of a high thermal conductivity first potting material. This design specifies filling the three raised ridges with a high thermal conductivity material to form three radial main heat dissipation channels, while the recessed areas are filled with conventional material to provide mechanical support.

[0082] Before using the preset hot partition intelligent discrimination neural network model, it is necessary to train the hot partition intelligent discrimination neural network model. The specific training method is as follows: First, training sample construction and label generation.

[0083] A. Sample input design (three-channel alignment).

[0084] As shown in Table 2, each training sample consists of three sets of features, which strictly correspond to the network input channels.

[0085] Table 2 B. Tag generation method.

[0086] Core principle: Deriving material requirements from thermophysical results.

[0087] Input: Thermal simulation temperature field / heat flux distribution map.

[0088] Steps: Extract the heat flux value at each location on the outer surface of the core; Calculate the heat flux threshold: Threshold = Average heat flux × Safety factor (usually 1.2-1.8); Binarize labels: Heat flux > threshold, representing the need for high thermal conductivity materials (label = 1); Heat flux ≤ threshold, representing the need for conventional materials (label = 0); Aggregate to partition matrix: Aggregate continuous labels to an M×3 circumferential × axial grid.

[0089] Label format: Two-dimensional thermal partition label matrix; Number of circumferential partitions M: usually twice the number of protrusions (each protrusion + adjacent depressions); Number of axial segments: fixed at 3 segments (upper, middle, lower); Matrix element values: 0 (conventional materials) or 1 (high thermal conductivity materials).

[0090] Second, design of the training objective function.

[0091] A weighted binary cross-entropy loss is used to penalize high-risk misclassifications. The loss function is set according to the principle that misclassifying a high-hot zone as a regular zone (missed judgment) carries a high risk and incurs a heavy penalty; misclassifying a regular zone as a high-hot zone (overjudgment) carries a low risk and incurs a light penalty. Weight settings: w_high = 2.0, high-hot zone label weight (missed judgment penalty coefficient); w_low = 0.8, regular zone label weight (overjudgment penalty coefficient). Loss calculation: Loss = -[ w_high×y_true×log(y_pred) + w_low×(1-y_true)×log(1-y_pred) ] Where y_true: the true label, taking a value of 0 or 1. In the context of thermal partitioning, 1 indicates that the partition requires high thermal conductivity materials, and 0 indicates that conventional materials are required. y_pred: the probability predicted by the model, representing the probability that the partition uses high thermal conductivity materials, with a value ranging from 0 to 1. For a single partition of a single sample, it can be broken down as follows: If y_true=1, the loss is -log(y_pred), which means that the closer y_pred is to 1, the smaller the loss; conversely, if y_pred is close to 0, the loss will be large.

[0092] If y_true=0, the loss is -log(1-y_pred), which means that the closer y_pred is to 0, the smaller the loss; conversely, if y_pred is close to 1, the loss will be large.

[0093] Third, the training process.

[0094] The training consists of 200-250 epochs, each containing the following steps: Step 1: Forward propagation, input data, network computation, output prediction result (y_pred); Step 2: Calculate the loss; Step 3: Backpropagation, calculate the gradient of each parameter; Step 4: Parameter update, use the AdamW optimizer to update all parameters; Step 5: Monitor progress. After each training round, evaluate the model using the validation set and record the best model.

[0095] If the validation loss does not decrease for 10 consecutive rounds: stop training and load the model with the lowest loss.

[0096] The embodiments of this application have the following beneficial effects: Firstly, through the structural design of the non-circular cross-section core, a three-dimensional radial heat dissipation network with raised and recessed areas is actively formed during the winding process, providing a clear physical basis for subsequent differentiated thermal management and breaking through the technical limitations of the traditional circular core with its single heat dissipation path and crude thermal management.

[0097] Secondly, a thermal zoning intelligent discrimination model is constructed based on the thermal generation distribution characteristics, dividing the outer surface of the core into high heat flux and conventional heat flux zones along the circumferential and axial directions. This achieves a precise mapping between heat generation behavior and heat dissipation path, providing a scientific basis for zoning potting.

[0098] Furthermore, a dual-material synergistic infusion strategy is adopted. By partitioning and filling with high thermal conductivity and conventional thermal conductivity potting materials, multiple radial main heat conduction paths and auxiliary support paths are constructed on the outer surface of the core, forming a multi-path parallel radial heat dissipation network, which significantly improves the heat diversion and conduction efficiency.

[0099] Finally, by using rheological matching and synergistic curing treatment of the partitioned potting materials, the morphological stability and interface bonding quality of different materials in the corresponding partitions are ensured. While achieving efficient heat dissipation, the overall mechanical stability and reliability of the encapsulation structure are guaranteed, thereby improving the service life and performance consistency of the capacitor under high frequency and high ripple current conditions.

[0100] The methods disclosed in the several method embodiments provided in this application can be arbitrarily combined without conflict to obtain new method embodiments.

[0101] The features disclosed in the several product embodiments provided in this application can be arbitrarily combined without conflict to obtain new product embodiments.

[0102] The features disclosed in the several method or device embodiments provided in this application can be arbitrarily combined without conflict to obtain new method or device embodiments.

[0103] The above description is merely an implementation method of the present application, but the protection scope of the present application is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in the present application should be included within the protection scope of the present application.

Claims

1. A method for manufacturing a non-circular cross-section core thermally partitioned potted film capacitor, characterized in that, The method includes: Based on the obtained capacitor operating conditions, the dielectric film thickness, sheet resistance of the metallized electrode, and winding tension window of the metallized thin film are determined. The metallized film is wound in multiple layers along the radial direction of the non-circular structure mandrel to obtain a metallized film winding body with a non-circular cross-section; and the mandrel of the non-circular structure is removed to obtain an initial capacitor core with a non-circular cross-section. The initial capacitor core with a non-circular cross-section is subjected to metal spraying and pressing to form end face electrodes, and the lead-out terminals are fixed to obtain a capacitor core with a non-circular cross-section winding and end face electrode lead-out. Based on the capacitor core, the geometric feature parameters of the non-circular cross-section of the core, the number of layers of the metallized thin film, and the target operating frequency are obtained; Based on the geometric feature parameters of the non-circular cross-section of the core, the number of layers of the metallized film, the dielectric film thickness of the metallized film, and the target operating frequency, thermal zoning is performed through a preset thermal zoning intelligent discrimination neural network model to determine the zoning distribution scheme of the potting material; Based on the aforementioned potting material zoning distribution scheme, zoning potting and radial heat dissipation path construction are performed to obtain the final capacitor.

2. The method according to claim 1, characterized in that, The process involves determining the distribution scheme of the potting material based on the non-circular cross-sectional geometric features of the core, the number of metallized film layers, the dielectric film thickness of the metallized film, and the target operating frequency, using a preset thermal zoning intelligent discrimination neural network model. Multiple features are obtained by extracting features from the geometric feature parameters of the non-circular cross-section of the core, the number of layers of the metallized thin film, the dielectric film thickness of the metallized thin film, and the target operating frequency; The multiple features are fused to obtain the fused features; Thermal zoning is performed using the fusion features to determine the zoning distribution scheme of the potting material.

3. The method according to claim 2, characterized in that, The process involves extracting features from the non-circular cross-sectional geometric parameters of the core, the number of metallized thin film layers, the dielectric film thickness of the metallized thin film, and the target operating frequency to obtain multiple features, including: The geometric features are obtained by extracting the geometric feature parameters of the non-circular cross-section of the core; The number of layers in the metallized thin film and the thickness of the dielectric film in the metallized thin film are used to extract features to obtain the stacking features; Feature extraction is performed on the target operating frequency to obtain frequency features; The geometric features, the stacked features, and the frequency features are determined as the plurality of features.

4. The method according to claim 2, characterized in that, The feature fusion of the multiple features to obtain fused features includes: The multiple features are fused to obtain primary fused features; The attention weights are calculated based on the frequency features among the multiple features. Based on the attention weights, the primary fusion features are subjected to a second feature fusion to obtain the fusion features.

5. The method according to claim 2, characterized in that, The step of determining the zoning distribution scheme of the potting material by performing thermal zoning based on the fusion features includes: Thermal partitioning is performed using the aforementioned fusion features to determine high heat flux partitions and conventional heat flux partitions; Based on the high heat flux zone and the conventional heat flux zone, potting material is allocated to determine the potting material zone distribution scheme.

6. The method according to claim 5, characterized in that, The process of allocating potting materials based on the high heat flux zones and the conventional heat flux zones, and determining the potting material zone distribution scheme, includes: Based on the high heat flux zone and the conventional heat flux zone, potting material is allocated, and the probability values ​​corresponding to the high heat flux zone and the conventional heat flux zone are determined respectively; the high heat flux zone corresponds to the protruding area formed by the non-circular cross-section protruding structure on the outer surface of the core; the conventional heat flux zone corresponds to the recessed area located between adjacent protruding structures on the outer surface of the core. If the probability value is greater than a preset threshold, then the partition corresponding to the probability value is determined to be the first potting material; If the probability value is less than or equal to the preset threshold, then the partition corresponding to the probability value is determined to be the second potting material; Based on the high heat flux zone, the conventional heat flux zone, the first potting material, and the second potting material, the zoning distribution scheme of the potting material is determined.

7. The method according to claim 6, characterized in that, The first potting material has a high thermal conductivity and is used to fill high heat flux zones; wherein, the high thermal conductivity is a thermal conductivity greater than a preset thermal conductivity threshold. The second potting material has a conventional thermal conductivity and is used to fill conventional heat flux zones and provide structural support; wherein, the conventional thermal conductivity is a thermal conductivity less than or equal to a preset thermal conductivity threshold.

8. The method according to claim 1, characterized in that, The process of partitioning and constructing radial heat dissipation paths based on the partitioned distribution scheme of the potting material to obtain the final capacitor includes: Based on the first and second potting materials in the potting material zoning distribution scheme, the high heat flux zone and the conventional heat flux zone are potted; After the potting process is completed, the first potting material and the second potting material are cured to obtain the radial main heat conduction path and the structural support path. Based on the radial main heat conduction path and the structural support path, a radial heat dissipation path is constructed to obtain the final capacitor.

9. The method according to claim 1, characterized in that, The metallized thin film includes a dielectric thin film and a metallized electrode layer disposed on at least one surface of the dielectric thin film.

10. A manufacturing system for a non-circular cross-section core thermally partitioned potted thin-film capacitor, characterized in that, The system is configured as follows: Based on the obtained capacitor operating conditions, the dielectric film thickness, sheet resistance of the metallized electrode, and winding tension window of the metallized thin film are determined. The metallized film is wound in multiple layers along the radial direction of the non-circular structure mandrel to obtain a metallized film winding body with a non-circular cross-section; and the mandrel of the non-circular structure is removed to obtain an initial capacitor core with a non-circular cross-section. The initial capacitor core with a non-circular cross-section is subjected to metal spraying and pressing to form end face electrodes, and the lead-out terminals are fixed to obtain a capacitor core with a non-circular cross-section winding and end face electrode lead-out. Based on the capacitor core, the geometric feature parameters of the non-circular cross-section of the core, the number of layers of the metallized thin film, and the target operating frequency are obtained; Based on the geometric feature parameters of the non-circular cross-section of the core, the number of layers of the metallized film, the dielectric film thickness of the metallized film, and the target operating frequency, thermal zoning is performed through a preset thermal zoning intelligent discrimination neural network model to determine the zoning distribution scheme of the potting material; Based on the aforementioned potting material zoning distribution scheme, zoning potting and radial heat dissipation path construction are performed to obtain the final capacitor.

Images