Industrial semiconductor packaging test system based on multi-dimensional intelligent decision

Abstract

The invention belongs to the technical field of information, and particularly relates to an industrial semiconductor packaging test system based on multi-dimensional intelligent decision making. By deploying multiple types of sensors and detection technologies at the tail end of the industrial robot and packaging test equipment, force touch and vision fusion and multi-modal data fusion technologies are applied, and flexible control, deep learning, big data analysis and reinforcement learning algorithms are combined. And flexible high-precision carrying of chips, rapid detection of packaging quality, deep processing of test data and multi-robot collaborative scheduling are realized. Meanwhile, the packaging test process is optimized by means of digital twinning and virtual simulation. According to the technology, the existing technical problems are effectively solved, and the precision, efficiency and quality of semiconductor packaging testing are remarkably improved.

Description

Technical Field

[0001] The present invention belongs to the field of information technology, and relates to an industrial semiconductor packaging and testing system based on multi-dimensional intelligent decision-making. Background Art

[0002] Semiconductor packaging and testing is a key link to encapsulate and protect the manufactured chips and perform performance tests on them to ensure that the chips meet the design requirements and quality standards. With the continuous development of semiconductor technology, the functions of chips are becoming more and more complex, and the requirements for the accuracy and efficiency of packaging and testing are also getting higher and higher. Industrial robots have been widely used in semiconductor packaging and testing due to their advantages such as automation, high precision, and stability. However, there are still many problems in the existing industrial robot technology during the packaging and testing process, which seriously restrict the further development of semiconductor packaging and testing technology.

[0003] The chips are easily damaged and have low positioning accuracy during the handling process: During the semiconductor packaging and testing process, the chips need to be frequently handled between different devices and workstations. The traditional industrial robot chip handling system mainly relies on simple mechanical grasping and preset motion trajectories, and cannot accurately sense the vulnerability and actual position changes of the chips. This results in the chips being easily damaged by mechanical stress during the handling process, and at the same time, the positioning accuracy is difficult to meet the increasingly fine packaging and testing requirements, affecting the efficiency of packaging and testing and the product quality.

[0004] The accuracy and efficiency of package quality inspection are low: Package quality inspection is a key step to ensure the reliability of semiconductor products. The existing industrial robot package quality inspection system mainly relies on single detection means and simple image processing algorithms, and cannot comprehensively and accurately detect various defects that may occur during the packaging process, such as pin deformation, package body cracks, poor soldering, etc. Moreover, the detection speed is slow, it is difficult to meet the needs of large-scale production, and there are easily missed detections and false detections, affecting the product quality and market competitiveness.

[0005] The ability to process and analyze test data is insufficient: During the semiconductor testing process, a large amount of test data is generated, and this data is crucial for evaluating the performance and quality of the chips. The existing industrial robot testing system mainly uses simple data storage and statistical methods, and cannot deeply analyze and mine complex test data, and it is difficult to extract valuable information from it to provide support for the optimization design of the chips and the improvement of the production process. Moreover, the data processing speed is slow, and the test results cannot be timely feedback, affecting the production efficiency and the timeliness of decision-making. Summary of the Invention

[0006] The present invention provides an industrial robot semiconductor packaging and testing system based on multi-dimensional intelligent decision-making, including:

[0007] Chip Flexible High-Precision Handling Algorithm Module Based on Force-Tactile and Visual Fusion: Install a force-tactile sensor and a high-resolution visual sensor at the end effector of the industrial robot. Let the force collected by the force-tactile sensor be F and the torque be M, and the visual sensor obtain the chip position information as (x, y, z), the attitude information as θ, and the surface feature information as S

[0008] ; Utilize force-tactile and visual fusion to construct a real-time state model C of chip handling through the formula C = f(F, M, x, y, z, θ, S), where f is a force-tactile and visual fusion function based on deep learning; Combine with a flexible control algorithm, and according to the state model C, calculate the adjustment amount ΔFgrip of the industrial robot's grasping force through the formula ΔP = g(C), where g is a flexible control adjustment function

[0009] 、the adjustment amount Δ(x, y, z) of the grasping position, and the adjustment amount ΔT of the motion trajectory, to achieve flexible grasping and high-precision handling of the chip;

[0010] Package Quality Rapid Detection Algorithm Module Based on Multimodal Data Fusion and Deep Learning: Integrate multiple technologies such as X-ray detection, ultrasonic detection, and optical detection in the package quality detection device. Let the X-ray detection obtain the internal structure information Ix, the ultrasonic detection obtain the physical property information as Iu, and the optical detection collect the surface defect information as Io; Utilize multimodal data fusion to construct a comprehensive information model E of the package through the formula E = h(Ix, Iu, Io), where h is a multimodal data fusion function based on deep learning; Combine the deep learning algorithms convolutional neural network (CNN) and long short-term memory network (LSTM) to analyze and learn the fused information, establish the mapping relationship between the defect features and defect types of the package, and achieve rapid detection of the package quality;

[0011] Test Data Processing Algorithm Module Based on Big Data Analysis and Intelligent Decision-Making: Collect a large amount of test data such as electrical performance test data De, thermal performance test data Dt, and reliability test data Dr generated during the semiconductor test process. Use big data analysis technology for data preprocessing, feature extraction, etc., and combine with intelligent decision-making algorithms, the combination of support vector machine (SVM) and decision tree algorithm, to evaluate and predict the chip performance and quality, and provide decision support for chip optimization design and production process improvement;

[0012] Multi-robot collaborative packaging test scheduling algorithm module based on reinforcement learning: Multiple industrial robots in a semiconductor packaging and testing workshop are regarded as a multi-agent system. Define the state space Sspace, including the robot position (xr, yr), task progress ptask, working status swork, and workshop environment information E, that is, Sspace = (xr, yr, ptask, swork, E); Define the action space Aaction, including the robot movement action mmove and task assignment decision dassign, that is, Aaction = (mmove, dassign); Design the reward function

[0013] R = q(eefficiency, qquality, uutilization), where q is the reward calculation function, eefficiency is the packaging and testing efficiency, qquality is the product quality, and uutilization is the equipment utilization rate); Use the deep Q-network (DQN) or proximal policy optimization algorithm (PPO) of the reinforcement learning algorithm to let the robot learn the optimal collaborative packaging and testing scheduling strategy in the interaction with the environment;

[0014] Semiconductor packaging and testing process optimization algorithm module based on digital twin and virtual simulation: Establish a digital twin model of the semiconductor packaging and testing system through modeling software and sensor data to achieve real-time interaction and synchronization between the physical entity and the virtual model. Use virtual simulation technology to simulate and optimize the packaging and testing process, and optimize and adjust the process parameters and flow according to the simulation results to achieve the optimization and improvement of the semiconductor packaging and testing process.

[0015] Furthermore, in the chip flexible high-precision handling algorithm module based on the fusion of force touch and vision, the multi-sensor data fusion adopts an algorithm based on deep learning to fuse the data collected by different sensors and construct a real-time state model of chip handling.

[0016] Furthermore, in the packaging quality rapid detection algorithm module based on multi-modal data fusion and deep learning, the multi-modal data fusion technology adopts an algorithm based on deep learning to fuse the data collected by different detection devices and construct a comprehensive information model of the package.

[0017] Furthermore, in the test data processing algorithm module based on big data analysis and intelligent decision-making, the big data analysis technology performs operations such as preprocessing, feature extraction, and classification and clustering on the test data, and combines the intelligent decision-making algorithm to evaluate and predict the chip performance and quality.

[0018] Furthermore, in the multi-robot collaborative packaging test scheduling algorithm module based on reinforcement learning, the state space includes the positions of the robots, task progress, working status, and environmental information of the workshop. The action space includes various motion actions of the robots and task allocation decisions. The reward function takes packaging test efficiency, product quality, and equipment utilization rate as reward indicators.

[0019] Furthermore, in the semiconductor packaging test process optimization algorithm module based on digital twin and virtual simulation, the semiconductor packaging test process is simulated and optimized through virtual simulation technology, and the process parameters and flow are optimized according to the simulation results.

[0020] Furthermore, in the chip flexible high-precision handling algorithm module based on the fusion of force touch and vision, the sensor collects data in real time at a set frequency and transmits it to the data processing unit. The data processing unit uses an algorithm based on deep learning to perform fusion processing and analysis on the data.

[0021] Furthermore, in the packaging quality rapid detection algorithm module based on multi-modal data fusion and deep learning, the detection device collects data in real time at a set frequency and transmits it to the data processing center. The data processing center uses an algorithm based on deep learning to perform fusion processing and analysis on the data.

[0022] Furthermore, in the test data processing algorithm module based on big data analysis and intelligent decision-making, a large amount of test data is collected, and real-time data processing technology is used to achieve rapid processing and analysis of the test data, and the test results are fed back in a timely manner.

[0023] Furthermore, in the multi-robot collaborative packaging test scheduling algorithm module based on reinforcement learning, multiple industrial robots in the semiconductor packaging test workshop are defined as multiple agents, a communication network between multiple agents is established, historical operation data is collected, and the reinforcement learning algorithm is used to train and optimize the reinforcement learning model. The agent selects actions and adjusts strategies according to real-time state information.

[0024] Beneficial effects:

[0025] Improve the safety and positioning accuracy of chip handling: The chip flexible high-precision handling algorithm based on the fusion of force touch and vision can comprehensively utilize the information of force touch and vision sensors, adjust the grasping force, grasping position, and motion trajectory of the industrial robot in real time, realize the flexible grasping and high-precision handling of chips, reduce the chip damage rate, improve the positioning accuracy, and ensure the smooth progress of packaging tests.

[0026] Improve the accuracy and efficiency of package quality inspection: The fast package quality inspection algorithm based on multi-modal data fusion and deep learning can comprehensively utilize the information of various detection technologies, accurately identify various types of package defects, and conduct quantitative analysis on the defects, improving the detection accuracy and efficiency, ensuring product quality, and enhancing market competitiveness.

[0027] Optimize the processing and analysis capabilities of test data: The test data processing algorithm based on big data analysis and intelligent decision-making can deeply analyze and mine a large amount of test data, extract valuable information, provide decision-making support for the optimized design of chips and the improvement of production processes, and at the same time achieve the fast processing and analysis of test data, timely feedback test results, and improve production efficiency and the timeliness of decision-making.

[0028] Optimize the collaborative package testing scheduling of multiple robots: The multi-robot collaborative package testing scheduling algorithm based on reinforcement learning can achieve efficient collaborative work among multiple robots, dynamically adjust task allocation and production plans according to real-time production situations and equipment status, improve the overall efficiency and resource utilization rate of package testing, and reduce production costs.

[0029] Improve the semiconductor package testing process: The semiconductor package testing process optimization algorithm based on digital twin and virtual simulation can predict and analyze in advance the possible problems in the package testing process through virtual simulation technology, optimize and adjust the parameters and processes of the package testing process, formulate the optimal package testing process plan, realize the optimization and improvement of the semiconductor package testing process, and improve the efficiency and quality of package testing. Description of the Drawings

[0030] Figure 1 Schematic diagram of the module flow. Detailed Implementation Manner

[0031] Example 1:

[0032] An industrial robot semiconductor package testing system based on multi-dimensional intelligent decision-making, including:

[0033] Chip flexible high-precision handling algorithm module based on the fusion of force touch and vision: Install a force touch sensor and a high-resolution vision sensor on the end effector of the industrial robot. Let the force collected by the force touch sensor be F and the torque be M, and the chip position information obtained by the vision sensor be (x, y, z), the attitude information be θ, and the surface feature information be S

[0034] ; Utilize the fusion of force - touch and vision, and construct a real - time state model C of chip handling through the formula C = f(F, M, x, y, z, θ, S), where f is a force - touch and vision fusion function based on deep learning; Combine a flexible control algorithm, and according to the state model C, calculate the adjustment amount of the gripping force of the industrial robot ΔFgrip through the formula ΔP = g(C), where g is a flexible control adjustment function

[0035] , the adjustment amount of the gripping position Δ(x, y, z), and the adjustment amount of the motion trajectory ΔT, to achieve flexible gripping and high - precision handling of the chip;

[0036] An algorithm module for rapid detection of packaging quality based on multi - modal data fusion and deep learning: Integrate multiple technologies such as X - ray detection, ultrasonic detection, and optical detection in the packaging quality detection equipment. Let the X - ray detection obtain the internal structure information Ix, the ultrasonic detection obtain the physical property information Iu, and the optical detection collect the surface defect information Io; Utilize multi - modal data fusion, and construct a comprehensive information model E of the package through the formula E = h(Ix, Iu, Io), where h is a multi - modal data fusion function based on deep learning; Combine the deep learning algorithms convolutional neural network (CNN) and long short - term memory network (LSTM), analyze and learn the fused information, establish the mapping relationship between the defect features and defect types of the package, and achieve rapid detection of packaging quality;

[0037] An algorithm module for test data processing based on big data analysis and intelligent decision - making: Collect a large amount of test data such as electrical performance test data De, thermal performance test data Dt, and reliability test data Dr generated during the semiconductor test process. Utilize big data analysis techniques for data pre - processing, feature extraction, etc., and combine intelligent decision - making algorithms, the combination of support vector machine (SVM) and decision tree algorithm, to evaluate and predict the performance and quality of the chip, and provide decision - making support for chip optimization design and production process improvement;

[0038] An algorithm module for multi - robot collaborative packaging test scheduling based on reinforcement learning: Regard multiple industrial robots in the semiconductor packaging test workshop as a multi - agent system. Define the state space Sspace, including the robot position (xr, yr), task progress ptask, working state swork, and workshop environment information E, that is, Sspace=(xr, yr, ptask, swork, E); Define the action space Aaction, including the robot motion action mmove and task assignment decision dassign, that is, Aaction=(mmove, dassign); Design a reward function

[0039] R = q(eefficiency, qquality, uutilization), where q is the reward calculation function, eefficiency is the encapsulation test efficiency, qquality is the product quality, and uutilization is the equipment utilization rate); using the deep Q-network (DQN) or proximal policy optimization algorithm (PPO) of the reinforcement learning algorithm, enabling the robot to learn the optimal collaborative encapsulation test scheduling strategy during the interaction with the environment;

[0040] Semiconductor packaging test process optimization algorithm module based on digital twin and virtual simulation: Establish a digital twin model of the semiconductor packaging test system through modeling software and sensor data to achieve real-time interaction and synchronization between the physical entity and the virtual model. Use virtual simulation technology to simulate and optimize the packaging test process, and optimize and adjust the process parameters and flow according to the simulation results to achieve the optimization and improvement of the semiconductor packaging test process.

[0041] Implementation of the flexible and high-precision chip handling algorithm based on the fusion of force touch and vision

[0042] Sensor deployment and data acquisition: Reasonably deploy force touch sensors and high-resolution vision sensors on the end effector of the industrial robot to ensure that the acting force, torque, and the position, attitude, and surface characteristics of the chip can be accurately obtained during the chip handling process. The sensors collect data in real time according to the set frequency and transmit the data to the data processing unit through the network. Force touch and vision data fusion and state model construction: The data processing unit uses a deep learning-based force touch and vision data fusion algorithm, such as a fusion network based on the attention mechanism, to fuse the data collected by different sensors and construct a real-time state model of the chip handling process. Use deep learning algorithms, such as convolutional neural networks (CNNs), to analyze and process the fused data and extract key feature information.

[0043] Implementation of the flexible control algorithm and handling execution: According to the constructed real-time state model, the flexible control algorithm calculates the grasping force, grasping position, and motion trajectory adjustment amount of the industrial robot. The robot control system adjusts the grasping force, grasping position, and motion trajectory of the robot in real time according to the adjustment amount to complete the chip handling task. During the handling process, continuously monitor the sensor data in real time and dynamically adjust the grasping force, grasping position, and motion trajectory to ensure the safe handling and precise placement of the chip. Implementation of the fast encapsulation quality detection algorithm based on multi-modal data fusion and deep learning Detection equipment deployment and data acquisition: Reasonably deploy X-ray detection equipment, ultrasonic detection equipment, optical detection equipment, etc. on the encapsulation quality detection equipment to ensure that multi-modal data of the package can be comprehensively obtained. The detection equipment collects data in real time according to the set frequency and transmits the data to the data processing center through the network.

[0044] Multimodal data fusion and feature extraction: The data processing center uses a multimodal data fusion algorithm based on deep learning to fuse the data collected by different detection equipment and build a comprehensive information model of the package. Deep learning algorithms, such as convolutional neural networks (CNN), are used to analyze and process the fused data and extract the key feature information of the package.

[0045] Deep learning model training and inspection execution: Collect a large amount of different types of package defect data to build a training data set. Use deep learning algorithms, such as the combination of convolutional neural networks (CNN) and long short-term memory networks (LSTM), to learn and train the training data set and establish a mapping relationship model between package defect features and defect types. In the actual inspection process, the extracted feature information is input into the trained model to quickly and accurately identify various types of package defects and perform quantitative analysis of the defects.

[0046] Implementation of test data processing algorithms based on big data analysis and intelligent decision making

[0047] Data collection and preprocessing: Collect a large amount of test data generated during the semiconductor testing process, and perform preprocessing operations such as cleaning, denoising, and normalization on the data to ensure data quality and consistency.

[0048] Big data analysis and feature extraction: Use big data analysis technology to perform feature extraction and classification and clustering operations on pre-processed test data to explore the potential relationships and patterns between data. Use deep learning algorithms, such as the combination of convolutional neural networks (CNN) and recurrent neural networks (RNN), to analyze and process test data and extract key feature information.

[0049] Intelligent decision-making algorithm implementation and data processing: Based on the analysis results, combined with intelligent decision-making algorithms, such as the combination of support vector machine (SVM) and decision tree algorithm, the performance and quality of the chip are evaluated and predicted, providing decision support for the optimal design of the chip and the improvement of the production process. At the same time, the real-time data processing technology is used to realize the rapid processing and analysis of the test data and timely feedback of the test results.

[0050] Implementation of multi-robot collaborative packaging test scheduling algorithm based on reinforcement learning

[0051] Agent definition and communication mechanism establishment: Multiple industrial robots in the semiconductor packaging and testing workshop are defined as multiple agents, each with independent decision-making capabilities and goals. A communication network between multiple agents is established to achieve real-time information exchange between agents.

[0052] Reinforcement Learning Model Training and Application: Define the state space, action space, and reward function of reinforcement learning. Collect historical operation data from the semiconductor packaging and testing workshop, and use reinforcement learning algorithms such as Deep Q-Network (DQN) or Proximal Policy Optimization (PPO) to train and optimize the reinforcement learning model. In actual production, the agent selects actions based on real-time state information and adjusts its policy according to the reward feedback after executing the actions, gradually optimizing the collaborative packaging and testing scheduling strategy to achieve efficient collaborative work among multiple robots.

[0053] Implementation of Semiconductor Packaging and Testing Process Optimization Algorithm Based on Digital Twin and Virtual Simulation

[0054] Digital Twin Model Construction: Use modeling software and sensor data to establish a digital twin model of the semiconductor packaging and testing system, mapping physical entities such as industrial robots, packaging and testing equipment, and chips in the physical world to the virtual space to achieve real-time interaction and synchronization between physical entities and virtual models.

[0055] Virtual Simulation and Process Optimization: Use virtual simulation technology to simulate and optimize the semiconductor packaging and testing process, predicting and analyzing potential problems that may occur during the packaging and testing process in the virtual environment in advance. According to the results of virtual simulation, optimize and adjust the parameters and processes of the semiconductor packaging and testing process, and formulate the optimal packaging and testing process plan. Then apply the optimized plan to actual production to achieve the optimization and improvement of the semiconductor packaging and testing process.

Claims

1. An industrial semiconductor packaging test system based on multi-dimensional intelligent decision-making, characterized in that: include: Flexible high-precision chip handling algorithm module based on force-tactile and vision fusion: Install force-tactile sensors and high-resolution vision sensors on the end effector of the industrial robot. Assume that the force collected by the force-tactile sensor is F, the torque is M, and the vision sensor obtains the chip position information (x, y, z), posture information is θ, and surface feature information is S; use force-tactile and vision fusion, and build a chip handling real-time state model C through the formula C=f(F, M, x, y, z, θ, S), where f is the force-tactile and vision fusion function based on deep learning; combine with the flexible control algorithm, according to the state model C, calculate the industrial robot's gripping force adjustment ΔFgrip, gripping position adjustment Δ(x, y, z), and motion trajectory adjustment ΔT through the formula ΔP=g(C), where g is the flexible control adjustment function, to achieve flexible gripping and high-precision handling of chips; Packaging quality rapid detection algorithm module based on multimodal data fusion and deep learning: Integrate multiple technologies such as X-ray detection, ultrasonic detection, and optical detection in the packaging quality detection equipment. Assume that X-ray detection obtains internal structure information Ix, ultrasonic detection obtains physical performance information Iu, and optical detection collects surface defect information Io; use multimodal data fusion, and construct a package comprehensive information model E through the formula E=h(Ix, Iu, Io), where h is a multimodal data fusion function based on deep learning; combine the deep learning algorithm convolutional neural network (CNN) and long short-term memory network (LSTM), analyze and learn the fused information, establish a mapping relationship between package defect characteristics and defect types, and realize rapid packaging quality detection; Test data processing algorithm module based on big data analysis and intelligent decision-making: collects a large amount of test data such as electrical performance test data De, thermal performance test data Dt, and reliability test data Dr generated during semiconductor testing. Use big data analysis technology to preprocess data, extract features, etc., combine intelligent decision-making algorithms, support vector machines (SVM) and decision tree algorithms to evaluate and predict chip performance and quality, and provide decision support for chip optimization design and production process improvement; Multi-robot collaborative packaging and testing scheduling algorithm module based on reinforcement learning: multiple industrial robots in the semiconductor packaging and testing workshop are regarded as a multi-agent system. Define the state space Sspace, including the robot position (xr, yr), task progress ptask, work status swork, workshop environment information E, that is, Sspace = (xr, yr, ptask, swork, E); define the action space Aaction, including the robot movement action mmove and the task allocation decision dassign, that is, Aaction = (mmove, dassign); design the reward function R = q (eefficiency, qquality, uutilization), q is the reward calculation function, eefficiency is the packaging and testing efficiency, qquality is the product quality, and uutilization is the equipment utilization rate); using the reinforcement learning algorithm Deep Q Network (DQN) or the Proximal Policy Optimization algorithm (PPO), the robot learns the optimal collaborative packaging and testing scheduling strategy in the interaction with the environment; Semiconductor packaging and testing process optimization algorithm module based on digital twin and virtual simulation: A digital twin model of the semiconductor packaging and testing system is established through modeling software and sensor data to achieve real-time interactive synchronization between the physical entity and the virtual model. Virtual simulation technology is used to simulate and optimize the packaging and testing process, and process parameters and processes are optimized and adjusted according to the simulation results to achieve optimization and improvement of the semiconductor packaging and testing process.

2. The industrial robot semiconductor packaging test system based on multi-dimensional intelligent decision-making according to claim 1 is characterized in that: In the flexible and high-precision chip handling algorithm module based on the fusion of force, touch and vision, multi-sensor data fusion adopts an algorithm based on deep learning to fuse the data collected by different sensors and build a real-time status model for chip handling.

3. The industrial robot semiconductor packaging test system based on multi-dimensional intelligent decision-making according to claim 1 is characterized in that: In the package quality rapid detection algorithm module based on multimodal data fusion and deep learning, the multimodal data fusion technology adopts an algorithm based on deep learning to fuse the data collected by different detection equipment and construct a comprehensive information model of the package.

4. The industrial robot semiconductor packaging test system based on multi-dimensional intelligent decision-making according to claim 1 is characterized in that: In the test data processing algorithm module based on big data analysis and intelligent decision-making, big data analysis technology performs preprocessing, feature extraction, classification and clustering on the test data, and combines the intelligent decision-making algorithm to evaluate and predict chip performance and quality.

5. The industrial robot semiconductor packaging test system based on multi-dimensional intelligent decision-making according to claim 1 is characterized in that: In the multi-robot collaborative packaging test scheduling algorithm module based on reinforcement learning, the state space includes the robot's position, task progress, working status and workshop environment information, the action space includes the robot's various motion actions and task allocation decisions, and the reward function uses packaging test efficiency, product quality and equipment utilization as reward indicators.

6. The industrial robot semiconductor packaging test system based on multi-dimensional intelligent decision-making according to claim 1, characterized in that: In the semiconductor packaging and testing process optimization algorithm module based on digital twins and virtual simulation, the semiconductor packaging and testing process is simulated and optimized through virtual simulation technology, and the process parameters and processes are optimized according to the simulation results.

7. The industrial robot semiconductor packaging test system based on multi-dimensional intelligent decision-making according to claim 1 is characterized in that: In the chip flexible high-precision handling algorithm module based on the fusion of force, touch and vision, the sensor collects data in real time at a set frequency and transmits it to the data processing unit. The data processing unit uses an algorithm based on deep learning to fuse and analyze the data.

8. The industrial robot semiconductor packaging test system based on multi-dimensional intelligent decision-making according to claim 1 is characterized in that: In the packaging quality rapid detection algorithm module based on multimodal data fusion and deep learning, the detection equipment collects data in real time at a set frequency and transmits it to the data processing center, and the data processing center uses an algorithm based on deep learning to fuse and analyze the data.

9. The industrial robot semiconductor packaging test system based on multi-dimensional intelligent decision-making according to claim 1, characterized in that: The test data processing algorithm module based on big data analysis and intelligent decision-making collects a large amount of test data, uses real-time data processing technology to quickly process and analyze the test data, and promptly feeds back the test results.

10. The industrial robot semiconductor packaging test system based on multi-dimensional intelligent decision-making according to claim 1, characterized in that: The reinforcement learning-based multi-robot collaborative packaging and testing scheduling algorithm module defines multiple industrial robots in a semiconductor packaging and testing workshop as multiple intelligent agents, establishes a communication network among the multiple intelligent agents, collects historical operation data, and uses a reinforcement learning algorithm to train and optimize the reinforcement learning model. The intelligent agents select actions and adjust strategies based on real-time status information.

Images