Packaging method of radio frequency SOI chip

Abstract

The invention discloses a packaging method of a radio frequency SOI chip, and relates to the technical field of chip packaging, in the packaging process, through a sensor array and a GPU parallel thread processor, packaging data are monitored and analyzed in real time, an edge detection algorithm and thermal gradient calculation are combined, the warping degree and a stress concentration area are recognized, and the packaging accuracy is improved. The generated deformation model dynamically reflects deformation distribution in the packaging process, multiple compensation strategies are dynamically generated by integrating the deformation model, process parameters, historical packaging data and a compensation algorithm and are simulated and evaluated, an optimal scheme is selected from the compensation strategies, packaging stability and adaptability are remarkably improved, and packaging quality is improved according to the compensation strategies. The position of the injection mold, the flow of cooling liquid and technological parameters are dynamically adjusted, in addition, a sensor continuously collects the compensated packaging state, a deformation model and compensation algorithm parameters are corrected through error evaluation, the packaging process is adjusted, the packaging quality and reliability are improved, and dynamic management of deformation is achieved.

Description

Technical Field

[0001] The present invention relates to the technical field of chip packaging, and particularly to a packaging method for a radio frequency SOI chip. Background Art

[0002] During the packaging process of a radio frequency SOI chip, deformation may be caused by thermal and mechanical stresses. The deformation may lead to warping of the packaging structure, pin misalignment, and solder joint failure, thereby affecting the signal transmission quality and the reliability of the chip. In a high-power working environment, the impact of packaging deformation on radio frequency performance is particularly significant; therefore, compensating for the deformation generated during the packaging process has become a key link in ensuring the performance of radio frequency SOI chips.

[0003] Currently, the commonly used packaging processes mainly address the deformation problem through the following methods: one is to select packaging materials with a low coefficient of thermal expansion, and the other is to enhance symmetry in mold design to balance internal stresses; in addition, some solutions use simulation technology to simulate the deformation during the packaging process and optimize it by adjusting mold design or process parameters. However, these solutions are mostly based on static simulation models and are difficult to fully reflect the dynamic changes during the packaging process; at the same time, deformation detection mainly relies on off-line measurement and cannot achieve real-time adjustment during the packaging process. This limitation makes it difficult to detect abnormal deformation in a timely manner, thereby increasing the defect rate and production cost; therefore, there is an urgent need for a packaging method for radio frequency SOI chips to solve such problems. Summary of the Invention

[0004] In view of the above existing problems, the present invention is proposed.

[0005] The present invention provides a packaging method for a radio frequency SOI chip to solve the problems that static simulation in traditional packaging solutions is difficult to fully reflect the dynamic changes during the packaging process; deformation detection relies on off-line measurement and cannot be adjusted in real time during the packaging process, resulting in the inability to detect abnormal deformation in a timely manner, increasing the defect rate and production cost.

[0006] To solve the above technical problems, the present invention provides the following technical solutions:

[0007] An embodiment of the present invention provides a packaging method for a radio frequency SOI chip, which includes,

[0008] Step S1, during the packaging process, arrange a sensor array on the surface and key positions of the mold to collect packaging data, including temperature, stress distribution, and geometric deformation information;

[0009] Step S2, perform deformation analysis on the packaging data through a parallel thread processor to generate a deformation model;

[0010] Step S3, input the deformation model combined with the preset process parameters and historical packaging data of the packaging process into a compensation algorithm to simulate a compensation strategy;

[0011] Step S4, control the encapsulation device to act according to the determined compensation strategy, and the device actions include adjusting the position of the injection mold, adjusting the flow distribution of the coolant, or dynamically adjusting the process parameters;

[0012] Step S5, the sensor array continuously collects the compensated encapsulation state, and evaluates the compensation effect based on the compensated encapsulation state.

[0013] As a preferred solution of the encapsulation method of a radio frequency SOI chip according to the present invention, wherein: the key positions of the mold include the inner wall of the mold, the chip fixing area, the inlet and outlet of the cooling channel, and the area around the injection point.

[0014] As a preferred solution of the encapsulation method of a radio frequency SOI chip according to the present invention, wherein: the deformation analysis method includes:

[0015] Adopt edge detection algorithm and thermal gradient calculation to extract the encapsulation deformation characteristics in the encapsulation data, including the warping degree and the stress concentration area, and generate a deformation model for describing the dynamic distribution of deformation during the encapsulation process.

[0016] As a preferred solution of the encapsulation method of a radio frequency SOI chip according to the present invention, wherein: the step of performing deformation analysis on the encapsulation data by the parallel thread processor to generate a deformation model is,

[0017] The temperature, stress distribution and geometric deformation information collected during the encapsulation process are input into the parallel thread processor of the GPU to extract deformation characteristics, and the data is segmented into small regions, and each thread is responsible for the calculation of one region.

[0018] Use the edge detection algorithm to extract the key features of geometric deformation, and the extraction formula is:

[0019]

[0020] wherein, E(x, y) is the edge intensity of the pixel point (x, y), and I(x, y) is the input image.

[0021] is the gradient information of the image.

[0022] Use thermal gradient calculation to analyze the temperature change, and the calculation formula is:

[0023]

[0024] wherein, is the temperature gradient, and T(x, y) is the temperature distribution image.

[0025] Integrate the edge and thermal gradient features:

[0026]

[0027] Among them, M(x, y) is the deformation model value, and α, β are adjustment weights.

[0028] As a preferred solution of the packaging method of a radio frequency SOI chip according to the present invention, wherein: in step S3, the compensation algorithm evaluates and compares multiple compensation strategies to determine the compensation strategy.

[0029] As a preferred solution of the packaging method of a radio frequency SOI chip according to the present invention, wherein: the deformation model combines preset process parameters and historical packaging data of the packaging process and is input into the compensation algorithm to simulate the compensation strategy; the steps for the compensation algorithm to evaluate and compare multiple compensation strategies to determine the compensation strategy are as follows.

[0030] Perform compensation strategy simulation, and the compensation input formula is:

[0031] C(x, y) = M(x, y) + γ·P(x, y) + δ·H(x, y),

[0032] Among them, C(x, y) is the comprehensive compensation input, M(x, y) is the deformation model, P(x, y) is the packaging process parameter, H(x, y) is the historical packaging data, and γ, δ are weight coefficients.

[0033] Simulate the effects of different compensation strategies, and the simulation formula is:

[0034] S i = f(C) for i = 1, 2, …, N,

[0035] Among them, S i is the effect of the i-th compensation strategy, and f(C) is the compensation strategy generation function.

[0036] Select the optimal strategy through the performance evaluation function, and the evaluation formula is:

[0037]

[0038] Among them, S opt is the optimal strategy, and Φ(S i ) is the performance index residual function of the strategy.

[0039] As a preferred solution of the packaging method of a radio frequency SOI chip according to the present invention, wherein: the compensation strategy generation function includes:

[0040] When the local deformation amount in the deformation model is significantly higher than the global average value, adopt the gradient descent strategy generation function to reduce the contribution of the local deformation amount and adjust the compensation input.

[0041] When the temperature gradient exceeds a preset threshold, a thermal compensation strategy generation function is adopted. By introducing a temperature distribution correction term, the compensation input is adjusted to balance the thermally uneven area.

[0042] When the deformation characteristics are distributed on one side of the mold, a stress redistribution strategy generation function is adopted to adjust the regional distribution of the deformation model and balance the stress.

[0043] When the historical encapsulation data indicates deformation anomalies under similar working conditions, a prediction strategy generation function based on historical data is adopted to correct the current compensation input in combination with historical deviations.

[0044] When the combined influence of multi-dimensional process parameters causes excessive deformation, a multi-variable coupling strategy generation function is adopted to incorporate the adjustment of the weights of each parameter into the compensation input model to generate a compensation strategy.

[0045] As a preferred embodiment of the encapsulation method for a radio frequency SOI chip according to the present invention, wherein: the step of controlling the actions of the encapsulation equipment according to the determined compensation strategy, and the equipment actions include adjusting the position of the injection mold, adjusting the flow distribution of the coolant, or dynamically adjusting the process parameters is,

[0046] According to the optimal compensation strategy, the equipment parameters are dynamically adjusted, and the equipment adjustment formula is:

[0047] Mold position adjustment:

[0048] Δx mold =k 1 ·(x target -x current )

[0049] where, Δx mold is the adjustment amount of the mold in the x direction, x target , x current are the target position and the current position of the mold, and k 1 is the adjustment proportional coefficient;

[0050] Coolant flow adjustment:

[0051] Q new =Q current +k 2 ·(T target -T current )

[0052] where, Q new is the adjusted coolant flow, T target , T current are the target temperature and the current temperature, and k 2 is the flow adjustment coefficient;

[0053] Process parameter adjustment:

[0054] P new = P current + λ·(M target - M current )

[0055] where P new , P current are the old and new process parameter values, M target , M current are the target and current deformation model values, and λ is the adjustment coefficient.

[0056] As a preferred solution of the packaging method of a radio frequency SOI chip according to the present invention, wherein: in step S5, the compensation effect is compared with the target state, and the comparison result is used to correct the parameters of the deformation model and the compensation algorithm.

[0057] As a preferred solution of the packaging method of a radio frequency SOI chip according to the present invention, wherein: the step of evaluating the compensation effect based on the compensated packaging state, comparing the compensation effect with the target state, and using the comparison result to correct the parameters of the deformation model and the compensation algorithm is

[0058] collecting data again through a sensor, evaluating the compensation effect, and correcting the deformation model and the compensation algorithm:

[0059] The compensation effect evaluation formula is:

[0060]

[0061] where E comp is the compensation effect error, M target,i , M actual,i are the target and actual deformation values, and n is the number of sampling points.

[0062] The model correction formula is:

[0063] M new = M current + η·E comp

[0064] where M new , M current are the old and new deformation models, and η is the correction coefficient;

[0065] The algorithm parameter update formula is:

[0066]

[0067] where Θ new , Θ current are the old and new compensation algorithm parameters, and μ is the learning rate. is the performance gradient.​

[0068] The beneficial effects of the present invention are as follows: Firstly, during the encapsulation process, through the sensor array and the GPU parallel thread processor, the encapsulation data is monitored and analyzed in real time. The edge detection algorithm is combined with the thermal gradient calculation to accurately identify the warping degree and the stress concentration area. The generated deformation model dynamically reflects the deformation distribution during the encapsulation process, making up for the deficiency of static monitoring in the traditional method. Secondly, by integrating the deformation model, process parameters, and historical encapsulation data, the compensation algorithm dynamically generates multiple compensation strategies, simulates and evaluates them, and selects the optimal solution from them, significantly improving the encapsulation stability and adaptability. According to the compensation strategy, the position of the injection mold, the coolant flow rate, and the process parameters are dynamically adjusted. In addition, the sensor continuously collects the encapsulated state after compensation, corrects the deformation model and the compensation algorithm parameters through error evaluation, and adjusts the encapsulation process, not only improving the encapsulation quality and reliability, but also realizing the dynamic management of deformation. BRIEF DESCRIPTION OF THE DRAWINGS

[0069] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following will briefly introduce the drawings required for the description of the embodiments. Obviously, the following drawings are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can be obtained based on these drawings.

[0070] Figure 1 It is a schematic flow chart of the encapsulation method for the radio frequency SOI chip of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0071] In order to make the above objects, features, and advantages of the present invention more obvious and understandable, the following will describe the specific embodiments of the present invention in detail with reference to the drawings of the specification.

[0072] In the following description, many specific details are set forth to fully understand the present invention. However, the present invention can also be implemented in other ways different from those described herein. Those skilled in the art can make similar promotions without departing from the connotation of the present invention. Therefore, the present invention is not limited by the specific embodiments disclosed below.

[0073] Secondly, the so-called "one embodiment" or "embodiment" herein refers to a specific feature, structure, or characteristic that can be included in at least one implementation manner of the present invention. The "in one embodiment" that appears in different places in this specification does not necessarily refer to the same embodiment, nor is it a separate or selectively exclusive embodiment from other embodiments.

[0074] Embodiment 1, referring to Figure 1 This embodiment provides a method for encapsulating a radio frequency SOI chip, including the following steps:

[0075] In step S1, during the encapsulation process, a sensor array is arranged on the surface of the mold and at key positions to collect encapsulation data, including temperature, stress distribution, and geometric deformation information.

[0076] The key positions of the mold include the inner wall of the mold, the chip fixing area, the inlet and outlet of the cooling channel, and the area around the injection point.

[0077] In step S2, a deformation analysis is performed on the encapsulation data through a parallel thread processor to generate a deformation model.

[0078] The deformation analysis methods include:

[0079] An edge detection algorithm and a thermal gradient calculation are used to extract the encapsulation deformation features in the encapsulation data, including the warping degree and the stress concentration area, and a deformation model is generated to describe the dynamic distribution of the deformation during the encapsulation process.

[0080] The steps of performing a deformation analysis on the encapsulation data through a parallel thread processor to generate a deformation model are as follows.

[0081] The temperature, stress distribution, and geometric deformation information collected during the encapsulation process are input into the parallel thread processor of the GPU to extract deformation features. The data is divided into small regions, and each thread is responsible for the calculation of one region.

[0082] An edge detection algorithm is used to extract the key features of the geometric deformation. The extraction formula is:

[0083]

[0084] where E(x, y) is the edge intensity of the pixel point (x, y), and I(x, y) is the input image.

[0085] is the gradient information of the image.

[0086] A thermal gradient calculation is used to analyze the temperature change. The calculation formula is:

[0087]

[0088] where is the temperature gradient, and T(x, y) is the temperature distribution image.

[0089] Combining the edge and thermal gradient features:

[0090]

[0091] where M(x, y) is the deformation model value, and α, β are the adjustment weights.

[0092] Specifically, a parallel calculator is used to quickly process data. Edge detection and thermal gradient calculation are employed for deformation analysis to provide feature information, and the generated deformation model is used to reflect the warping degree and stress concentration area.

[0093] In step S3, the deformation model is combined with the preset process parameters and historical encapsulation data of the encapsulation process and input into a compensation algorithm to simulate compensation strategies.

[0094] In step S3, the compensation algorithm evaluates and compares multiple compensation strategies to determine the compensation strategy.

[0095] The deformation model is combined with the preset process parameters and historical encapsulation data of the encapsulation process and input into a compensation algorithm to simulate compensation strategies. The steps for the compensation algorithm to evaluate and compare multiple compensation strategies to determine the compensation strategy are as follows.

[0096] Perform compensation strategy simulation. The compensation input formula is:

[0097] C(x,y) = M(x,y) + γ·P(x,y) + δ·H(x,y),

[0098] where y(x,y) is the comprehensive compensation input, M(x,y) is the deformation model, P(x,y) is the encapsulation process parameter, H(x,y) is the historical encapsulation data, and γ,δ are weight coefficients.

[0099] Simulate the effects of different compensation strategies. The simulation formula is:

[0100] S i = f(C) for i = 1,2,…,N,

[0101] where S i is the effect of the i-th compensation strategy, and f(C) is the compensation strategy generation function.

[0102] Select the optimal strategy through a performance evaluation function. The evaluation formula is:

[0103]

[0104] where S opt is the optimal strategy, and Φ(S i ) is the performance index residual function of the strategy.

[0105] Specifically, the deformation model integrates process parameters and historical data, evaluates different strategies through simulation and comparison, and finally determines the optimal solution.

[0106] The compensation strategy generation function includes:

[0107] When the local deformation amount in the deformation model is significantly higher than the global average value, a gradient descent strategy is used to generate a function to reduce the contribution of the local deformation amount and adjust and compensate the input;

[0108] When the temperature gradient exceeds the preset threshold, a thermal compensation strategy is used to generate a function, and the input is adjusted and compensated by introducing a temperature distribution correction term to adjust and balance the thermally uneven area;

[0109] When the deformation characteristics are distributed on one side of the mold, a stress redistribution strategy is used to generate a function to adjust the regional distribution of the deformation model and adjust the stress balance;

[0110] When the historical encapsulation data indicates that there are deformation anomalies under similar working conditions, a prediction strategy based on historical data is used to generate a function to correct the current compensation input in combination with historical deviations;

[0111] When the combined influence of multi-dimensional process parameters causes the deformation to exceed the standard, a multi-variable coupling strategy is used to generate a function, and the weight adjustment of each parameter is incorporated into the compensation input model to generate a compensation strategy;

[0112] Step S4, control the actions of the encapsulation equipment according to the determined compensation strategy, and the equipment actions include adjusting the position of the injection mold, adjusting the flow distribution of the coolant, or dynamically adjusting the process parameters;

[0113] The step of controlling the actions of the encapsulation equipment according to the determined compensation strategy, where the equipment actions include adjusting the position of the injection mold, adjusting the flow distribution of the coolant, or dynamically adjusting the process parameters is,

[0114] According to the optimal compensation strategy, dynamically adjust the equipment parameters, and the equipment adjustment formula is:

[0115] Mold position adjustment:

[0116] Δx mold =k 1 ·(x target -x current ),

[0117] Among them, Δx mold is the adjustment amount of the mold in the x direction, x target ,x current is the target position and the current position of the mold, and k 1 is the adjustment proportionality coefficient;

[0118] Coolant flow adjustment:

[0119] Q new =Q current +k 2 ·(T target -T current ),

[0120] Among them, Q new is the adjusted coolant flow rate, T target , T current is the target temperature and the current temperature, k 2 is the flow rate adjustment coefficient;

[0121] Process parameter adjustment:

[0122] P new = P current + λ·(M target - M current ),

[0123] Among them, P new , P current are the old and new process parameter values, M target , M current are the target and current deformation model values, and λ is the adjustment coefficient;

[0124] Specifically, the mold position, cooling flow rate, and process parameters are adjusted in real time, and the operation is adjusted according to the compensation strategy to dynamically respond to the packaging deformation problem.

[0125] Step S5: The sensor array continuously collects the compensated packaging state and evaluates the compensation effect based on the compensated packaging state;

[0126] In step S5, the compensation effect is compared with the target state, and the comparison result is used to correct the parameters of the deformation model and the compensation algorithm;

[0127] The step of evaluating the compensation effect based on the compensated packaging state and comparing the compensation effect with the target state, and using the comparison result to correct the parameters of the deformation model and the compensation algorithm is,

[0128] Collect data again through the sensor, evaluate the compensation effect, and correct the deformation model and the compensation algorithm:

[0129] The compensation effect evaluation formula is:

[0130]

[0131] Among them, E comp is the compensation effect error, M target,i , M actual,i are the target and actual deformation values, and n is the number of sampling points,

[0132] The model correction formula is:

[0133] M new = M current + η·E comp ,

[0134] Among them, M new , Mcurrent is the new and old deformation models, and η is the correction coefficient;

[0135] The algorithm parameter update formula is:

[0136]

[0137] where, Θ new , Θ current are the new and old compensation algorithm parameters, μ is the learning rate, is the performance gradient; specifically, by continuously monitoring the compensation effect, the deformation model and the compensation algorithm parameters are corrected to form a data-driven closed-loop optimization, continuously improving the packaging quality.

[0138] In summary, in the present invention, first, during the packaging process, through the sensor array and the GPU parallel thread processor, the packaging data is monitored and analyzed in real time. The edge detection algorithm is combined with the thermal gradient calculation to accurately identify the warping degree and the stress concentration area. The generated deformation model dynamically reflects the deformation distribution during the packaging process, making up for the deficiency of static monitoring in the traditional method; second, by integrating the deformation model, process parameters, and historical packaging data, the compensation algorithm dynamically generates multiple compensation strategies, simulates and evaluates them, and selects the optimal solution from them, significantly improving the packaging stability and adaptability. According to the compensation strategy, the position of the injection mold, the coolant flow rate, and the process parameters are dynamically adjusted. In addition, the sensor continuously collects the packaged state after compensation, corrects the deformation model and the compensation algorithm parameters through error evaluation, and adjusts the packaging process, not only improving the packaging quality and reliability, but also realizing the dynamic management of deformation.

[0139] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to the preferred embodiments, those of ordinary skill in the art should understand that the technical solutions of the present invention can be modified or equivalently replaced without departing from the spirit and scope of the technical solutions of the present invention, and they should all be covered by the scope of the claims of the present invention.

Claims

1. A method for packaging a radio frequency SOI chip, characterized in that: include, Step S1, during the packaging process, a sensor array is arranged on the mold surface and key positions to collect packaging data, including temperature, stress distribution and geometric deformation information; Step S2, performing deformation analysis on the package data through a parallel thread processor to generate a deformation model; Step S3, the deformation model is combined with preset process parameters and historical packaging data of the packaging process, and input into the compensation algorithm to simulate the compensation strategy; Step S4, controlling the packaging equipment action according to the determined compensation strategy, the equipment action includes adjusting the position of the injection mold, adjusting the flow distribution of the coolant, or dynamically adjusting the process parameters; Step S5: the sensor array continuously collects the compensated packaging state, and evaluates the compensation effect based on the compensated packaging state.

2. A method for packaging a radio frequency SOI chip according to claim 1, characterized in that: The key positions of the mold include the inner wall of the mold, the chip fixing area, the inlet and outlet of the cooling channel, and the area around the injection point.

3. A method for packaging a radio frequency SOI chip as claimed in claim 2, characterized in that: The deformation analysis method includes: The edge detection algorithm and thermal gradient calculation are used to extract the package deformation features in the package data, including the warpage degree and stress concentration area, and generate a deformation model.

4. A method for packaging a radio frequency SOI chip as claimed in claim 3, characterized in that: The step of performing deformation analysis on the package data by using a parallel thread processor to generate a deformation model is as follows: The temperature, stress distribution and geometric deformation information collected during the packaging process are input into the parallel thread processor of the GPU to extract the deformation features. The data is divided into small areas, and each thread is responsible for the calculation of one area. The edge detection algorithm is used to extract the key features of geometric deformation. The extraction formula is: Among them, E(x,y) is the edge strength of the pixel (x,y), I(x,y) is the input image, is the gradient information of the image, The temperature change is analyzed by thermal gradient calculation, and the calculation formula is: in, is the temperature gradient, T(x,y) is the temperature distribution image, Combined edge and thermal gradient features: Among them, M(x,y) is the deformation model value, and α, β are adjustment weights.

5. A method for packaging a radio frequency SOI chip as claimed in claim 4, characterized in that: In step S3, the compensation algorithm evaluates and compares multiple compensation strategies to determine the compensation strategy.

6. A method for packaging a radio frequency SOI chip as claimed in claim 5, characterized in that: The deformation model is combined with preset process parameters and historical packaging data of the packaging process and input into the compensation algorithm to simulate the compensation strategy; The compensation algorithm evaluates and compares various compensation strategies. The steps to determine the compensation strategy are: To simulate the compensation strategy, the compensation input formula is: C(x,y)=M(x,y)+γ·P(x,y)+δ·H(x,y), Among them, C(x,y) is the comprehensive compensation input, M(x,y) is the deformation model, P(x,y) is the packaging process parameter, H(x,y) is the historical packaging data, γ,δ are weight coefficients, To simulate the effects of different compensation strategies, the simulation formula is: S i =f(C) fori=1,2,…,N, Among them, S i is the effect of the i-th compensation strategy, f(C) is the compensation strategy generating function, The optimal strategy is selected through the performance evaluation function. The evaluation formula is: Among them, S opt is the optimal strategy, Φ(S i ) is the performance indicator residual function of the strategy.

7. A method for packaging a radio frequency SOI chip as claimed in claim 6, characterized in that: The compensation strategy generation function includes: When the local deformation variable in the deformation model is significantly higher than the global mean, the gradient descent strategy is used to generate the function, reduce the contribution of the local deformation variable and adjust the compensation input; When the temperature gradient exceeds the preset threshold, a thermal compensation strategy is used to generate a function, and the compensation input is adjusted by introducing a temperature distribution correction term to balance the thermally uneven area; When the deformation feature is distributed on one side of the mold, the stress redistribution strategy is used to generate a function to adjust the regional distribution of the deformation model to adjust the stress balance; When historical packaging data indicates that there is abnormal deformation under similar working conditions, a prediction strategy based on historical data is used to generate a function, and the current compensation input is corrected in combination with historical deviations; When the combined influence of multidimensional process parameters causes deformation to exceed the standard, a multivariable coupling strategy is used to generate a function, and the weight adjustment of each parameter is incorporated into the compensation input model to generate a compensation strategy.

8. The method for packaging a radio frequency SOI chip according to claim 7, characterized in that: The step of controlling the packaging equipment action according to the determined compensation strategy, wherein the equipment action includes adjusting the position of the injection mold, adjusting the flow distribution of the coolant, or dynamically adjusting the process parameters is as follows: According to the optimal compensation strategy, the equipment parameters are adjusted dynamically. The equipment adjustment formula is: Mould position adjustment: Δx mold =k1·(x target -x current ), Where Δx mold is the adjustment amount of the mold in the x direction, x target ,x current are the target position and current position of the mold, and k1 is the adjustment ratio coefficient; Coolant flow adjustment: Q new =Q current +k2·(T target -T current ), Among them, Q new is the adjusted coolant flow rate, T target ,T current are the target temperature and the current temperature, k2 is the flow adjustment coefficient; Process parameter adjustment: P new =P current +λ·(M target -M current ), Among them, P new ,P current are the new and old process parameter values, M target ,M current is the target and current deformation model value, and λ is the adjustment coefficient.

9. A method for packaging a radio frequency SOI chip as claimed in claim 8, characterized in that: In step S5, the compensation effect is compared with the target state, and the comparison result is used to correct the parameters of the deformation model and the compensation algorithm.

10. The method for packaging a radio frequency SOI chip according to claim 9, characterized in that: The steps of evaluating the compensation effect based on the compensated packaging state, comparing the compensation effect with the target state, and using the comparison result to correct the deformation model and the parameters of the compensation algorithm are as follows: Collect data again through sensors, evaluate the compensation effect, and revise the deformation model and compensation algorithm: The compensation effect evaluation formula is: Among them, E comp To compensate for the effect error, M target,i ,M actual,i are the target and actual deformation values, n is the number of sampling points, The model correction formula is: M new =M current +η·E comp , Among them, M new ,M current are the new and old deformation models, η is the correction coefficient; The algorithm parameter update formula is: Among them, Θ new ,Θ current are the parameters of the new and old compensation algorithms, μ is the learning rate, The performance gradient.

Images