Performance prediction optimization method for sorting machine of semiconductor device test system

Abstract

The invention discloses a performance prediction optimization method for a sorting machine of a semiconductor device test system, and the method comprises the steps: obtaining a real-time data flow from the sorting machine, and determining the respective action duration sequence of a plurality of actuators used for executing the taking and placing actions of semiconductor devices in the sorting machine according to the real-time data flow; comparing the action duration of each actuator with a corresponding preset reference duration mean value to determine an action duration deviation sequence of each actuator; and according to the action duration deviation sequence of each actuator, processing is carried out based on the trained performance prediction model to determine the prediction change condition of the indexing time of the sorting machine, the future performance degradation trend can be pre-judged from the early fine deviation, the potential performance degradation risk of the sorting machine can be recognized in advance, and the sorting efficiency is improved. Therefore, non-planned downtime is reduced, equipment throughput and product yield are guaranteed, dependence on manual real-time monitoring and empirical judgment is reduced, and stability and efficiency of the semiconductor device testing and sorting process are improved.

Description

Technical Field

[0001] This invention relates to the fields of semiconductor automated manufacturing and semiconductor device testing technology, and in particular to a method for performance prediction and optimization of a semiconductor device testing system sorting machine. Background Technology

[0002] In the final semiconductor testing process, the test sorter is a core piece of equipment in MEMS (Micro-Electro-Mechanical Systems) testing systems. It needs to perform high-speed, micron-level precision device pick-up and placement operations, and its indexing time (i.e., the interval between two placement operations) is a crucial indicator determining the equipment's throughput and testing costs. During the new product introduction phase, the sorter is debugged to its optimal state, providing an initial standard for efficient equipment operation.

[0003] Current solutions for the performance degradation of sorting machines are mainly passive responses, including manual debugging, threshold alarms, general predictive maintenance systems, and general data analysis tools. These solutions address problems in equipment operation through manual intervention, abnormal triggering alerts, component failure prediction, and general data processing, respectively.

[0004] However, manual debugging is time-consuming, inconsistent, and relies on expert experience; threshold alarm schemes only trigger after performance degradation severely impacts production; general predictive maintenance schemes can only predict sudden component failures and cannot identify gradual performance degradation; general data analysis tools lack knowledge specific to the sorting machine field, cannot provide effective optimization solutions, and are difficult to intervene before performance degradation affects equipment efficiency, thus reducing the stability and efficiency of the semiconductor device testing and sorting process. Summary of the Invention

[0005] This invention provides a performance prediction and optimization method for a semiconductor device testing system sorter, which solves the problems of the inability to identify progressive performance degradation of the sorter, the high dependence of sorter performance prediction on real-time manual monitoring and experience judgment, and the resulting poor stability and low efficiency of the semiconductor device testing and sorting process.

[0006] According to one aspect of the present invention, a performance prediction and optimization method for a semiconductor device testing system sorter is provided, comprising:

[0007] A real-time data stream is acquired from the sorting machine, and the action duration sequence of each of the multiple actuators in the sorting machine used to perform the picking and placing of semiconductor devices is determined based on the real-time data stream; wherein, the action duration sequence includes the action duration of multiple picking and placing action cycles; the action duration of each actuator is compared with the corresponding preset benchmark duration average to determine the action duration deviation sequence of each actuator; based on the action duration deviation sequence of each actuator, a trained performance prediction model is used to determine the predicted change in the sorting time of the sorting machine.

[0008] According to another aspect of the present invention, a performance prediction and optimization apparatus for a semiconductor device testing system sorter is provided, comprising:

[0009] A sequence determination module is used to acquire real-time data streams from the sorting machine and determine the action duration sequence of each of the multiple actuators in the sorting machine that perform semiconductor device pick-and-place actions based on the real-time data streams; wherein the action duration sequence includes the action duration of multiple pick-and-place action cycles; a benchmark comparison module is used to compare the action duration of each actuator with the corresponding preset benchmark duration average to determine the action duration deviation sequence of each actuator; a prediction change module is used to process the action duration deviation sequence of each actuator based on a trained performance prediction model to determine the predicted change in the sorting time of the sorting machine.

[0010] According to another aspect of the present invention, an electronic device is provided, the electronic device comprising:

[0011] At least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores a computer program executable by the at least one processor, the computer program being executed by the at least one processor to enable the at least one processor to execute the performance prediction and optimization method for the semiconductor device test system sorter according to any embodiment of the present invention.

[0012] According to another aspect of the present invention, a computer-readable storage medium is provided, the computer-readable storage medium storing computer instructions, the computer instructions being configured to cause a processor to execute and implement the performance prediction and optimization method for the sorting machine of the semiconductor device testing system according to any embodiment of the present invention.

[0013] According to another aspect of the present invention, a computer program product is also provided, including a computer program that, when executed by a processor, implements the steps of the method as described in any embodiment of the present invention.

[0014] The technical solution of this invention, by acquiring the action duration sequence of multiple actuators from the sorting machine in real time, can capture the timing performance of each actuator in actual operation, avoiding the influence of randomness in single-cycle data. Furthermore, by comparing the real-time action duration with the average of a preset benchmark duration, the action duration deviation sequence of each actuator is determined. Based on a trained performance prediction model, the deviation sequence is processed to determine the predicted changes in the grading time. This allows for the prediction of future performance degradation trends, such as abnormal events, degradation magnitude, and occurrence cycles, from early subtle deviations, rather than passively responding only after severe performance degradation. It can identify potential progressive performance degradation risks of the sorting machine in advance, providing a basis for subsequent targeted optimization, thereby reducing unplanned downtime, ensuring equipment throughput and product yield, and reducing reliance on manual real-time monitoring and experience-based judgment, thus improving the stability and efficiency of the semiconductor device testing and sorting process.

[0015] It should be understood that the description in this section is not intended to identify key or essential features of the embodiments of the present invention, nor is it intended to limit the scope of the invention. Other features of the invention will become readily apparent from the following description. Attached Figure Description

[0016] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0017] Figure 1 This is a flowchart of a performance prediction and optimization method for a semiconductor device testing system sorter according to Embodiment 1 of the present invention;

[0018] Figure 2 This is a flowchart of another performance prediction and optimization method for a semiconductor device testing system sorter provided in Embodiment 2 of the present invention;

[0019] Figure 3 This is a flowchart of another performance prediction and optimization method for a semiconductor device testing system sorter provided in Embodiment 3 of the present invention;

[0020] Figure 4 This is a schematic diagram of an actuator timing sequence comparison applicable to an embodiment of the present invention;

[0021] Figure 5 This is a schematic diagram of the performance prediction and optimization device for a semiconductor device testing system sorter according to Embodiment 4 of the present invention;

[0022] Figure 6This is a schematic diagram of the structure of an electronic device that implements the performance prediction and optimization method of the semiconductor device testing system sorter according to an embodiment of the present invention. Detailed Implementation

[0023] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.

[0024] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this invention are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of the invention described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0025] Example 1

[0026] Figure 1 This is a flowchart of a performance prediction and optimization method for a semiconductor device testing system sorter provided in Embodiment 1 of the present invention. This embodiment is applicable to the prediction and optimization of the performance of a semiconductor device testing system sorter. The method can be executed by a performance prediction and optimization device for the semiconductor device testing system sorter. This performance prediction and optimization device for the semiconductor device testing system sorter can be implemented in hardware and / or software and is generally configured in electronic devices.

[0027] The sorting machine in the semiconductor device testing system may include: a motion module, a pick-and-place module, a sensing module, and a control module.

[0028] The motion module may include an X-axis actuator, a Y-axis actuator, a Z-axis actuator, and a Theta rotary actuator. The X-axis actuator and the Y-axis actuator are horizontally and vertically distributed. The Z-axis actuator is vertically mounted on the moving end of the X-axis actuator or the Y-axis actuator. The Theta rotary actuator is fixed to the lower end of the Z-axis actuator and is used to drive the load (such as a MEMS device) to achieve translation in the X, Y, and Z directions and angle adjustment around the Z-axis.

[0029] The pick-and-place module may include a nozzle assembly and a vacuum system. The nozzle assembly is detachably connected to the output end of the Theta rotary actuator. The vacuum system is connected to the nozzle assembly through a pipeline and is used to provide vacuum suction for the nozzle assembly to adsorb or release the items to be sorted.

[0030] The sensing module may include a motor current sensor, an optical encoder, a vacuum pressure sensor, and a temperature sensor. The position sensor is electrically connected to the motion module to provide feedback on motion positioning information. The vacuum pressure sensor is connected to the vacuum system to monitor the adsorption state. The temperature sensor is installed at the test base to collect the test environment temperature.

[0031] The control module is electrically connected to the motion module, the pick-and-place module and the sensing module, respectively, and is used to receive feedback signals from each module and output control commands.

[0032] The movement phases of the sorting machine can include the duration of each of the X, Y, Z, and Theta actuators from pick-up trigger to placement completion.

[0033] In this embodiment of the invention, the sorting machine can be specifically understood as: a downstream device in a semiconductor device testing system, used to quickly sort, classify, and package tested semiconductor devices (such as MEMS devices, radio frequency (RF) devices, and high-end logic devices), performing high-speed and micron-level precision pick-up and placement operations. The motion module can be specifically understood as: a module in the sorting machine that adjusts the spatial position of the devices, composed of multiple actuators, used to drive the load to complete translational and rotational movements.

[0034] The pick-and-place module can be understood as: a module used to pick up and release the items to be sorted, achieving stable operation of the devices through vacuum suction. The sensing module can be understood as: a module used to collect operational status data from each module, providing feedback information to the control module to ensure operational accuracy and stability. The control module can be understood as: the control component of the sorting machine, used to receive feedback from the sensing module and output control commands, coordinating the collaborative work of all modules.

[0035] The actuator can be understood as the core component of the motion module, capable of horizontal and vertical translation and rotation around the Z-axis, collectively adjusting the spatial position of the device. The nozzle assembly can be understood as the component in the pick-and-place module that directly contacts and adsorbs the items to be sorted; it is detachable for easy maintenance and replacement. The vacuum system can be understood as the system that provides vacuum suction to the nozzle assembly, serving as the power source for picking up and releasing the device.

[0036] Specifically, the sensor can be understood as the specific components of the sensing module, which may include a motor current sensor, an optical encoder, a vacuum pressure sensor, and a temperature sensor, respectively monitoring the motor's operating status, actuator position, adsorption status, and ambient temperature. The motion phase can be understood as the working cycle of each actuator during a single device pickup and placement operation by the sorting machine; it is the duration from pickup triggering to placement completion.

[0037] Specifically, the sorting machine in the semiconductor device testing system consists of a motion module, a pick-and-place module, a sensing module, and a control module. The motion module includes X-axis and Y-axis actuators that are horizontally and vertically distributed, a Z-axis actuator that is vertically mounted on the moving end of the X-axis or Y-axis, and a Theta rotary actuator fixed at the lower end of the Z-axis. Together, they realize the translation of the load in the X, Y, and Z directions and the adjustment of the angle around the Z-axis. The motion phase specifically refers to the duration from the pick-up trigger to the completion of placement by each actuator, and may include sub-phases such as acceleration, constant speed, deceleration, and settling.

[0038] The pick-and-place module consists of a detachable nozzle assembly connected to the output of the Theta rotary actuator and a connected vacuum system. It uses vacuum suction to pick up and release the parts to be sorted. Event timestamps such as Vacuum_OK (indicating successful vacuum establishment after nozzle contact with the device) are key indicators for measuring the reliability of the nozzle's operation.

[0039] The motor current sensor in the sensing module can monitor the actuator motor load and health status, the optical encoder can provide high-resolution position feedback to support actuator positioning and motion duration calculation, the vacuum pressure sensor is connected to the vacuum system to monitor the adsorption state, and the temperature sensor is installed on the test base to collect the test environment temperature. These sensor data and actuator timing data together constitute the data source for system analysis.

[0040] The control module is electrically connected to the motion module, the pick-and-place module, and the sensing module respectively. It is used to receive feedback signals from each module (such as sensor data and actuator action completion signals) and output control commands (such as actuator motion parameter adjustment or vacuum suction adjustment commands) to ensure that each module of the sorting machine works together to complete high-speed and micron-level precision pick-and-place operations.

[0041] Understandably, semiconductor test sorters can be configured for testing MEMS devices.

[0042] Correspondingly, such as Figure 1 As shown, the method includes:

[0043] S110. Obtain a real-time data stream from the sorting machine, and determine the action duration sequence of each of the multiple actuators in the sorting machine used to perform semiconductor device pick-and-place actions based on the real-time data stream.

[0044] The action duration sequence includes the action duration of multiple pick-up and put-down action cycles.

[0045] In this embodiment of the invention, the action duration sequence can be specifically understood as: the set of durations during which each actuator completes its corresponding action in multiple consecutive pick-and-place action cycles arranged in chronological order. The pick-and-place action cycle can be specifically understood as: the complete operation process of the sorting machine picking up a semiconductor device, adjusting its position, and placing it, with each cycle corresponding to an independent pick-and-place operation.

[0046] The duration of an action can be specifically understood as the time from the start of an action (such as picking up) to the completion of an action (such as placing) within a single pick-up and place-down action cycle.

[0047] Understandably, the duration of each actuator's action can be further divided into various statistical perspectives based on monitoring and analysis needs: it can be the duration of a single motion phase, such as the acceleration time from start-up to reaching rated speed, the constant speed time during uniform operation at rated speed, the deceleration time from start-up to complete stop, and the fine-tuning time from positioning correction to precise placement, used to analyze the actuator's motion control accuracy and mechanical response characteristics; it can also be the duration of a single independent action, such as the one-way motion time of the X or Y axis moving horizontally from the pickup point to above the placement point, the descent time of the Z axis from the standby height to the pickup height, and the angle adjustment time of the rotary actuator rotating from the initial angle to the target angle, suitable for addressing the performance degradation problem of a single action; or it can be the duration of a complete cycle including the reset process, i.e., the entire process duration from pickup triggering to placement completion and actuator reset to the standby position or state before pickup, which is closer to the actual cycle of a single pick-up and placement operation of the sorting machine and can be directly associated with the equipment indexing time optimization needs.

[0048] Specifically, the system acquires real-time data streams from the sorting machine through industry standard protocols (such as SECS (Semiconductor Equipment Communication Standard) / GEM (Generic Equipment Model) or Modbus TCP (Modbus Transmission Control Protocol)). These data streams include timestamps of actuator action events, sensor telemetry data (such as motor current and vacuum pressure), and equipment operating status identifiers. The system parses the real-time data streams, extracts key action events such as action triggering and action completion with timestamps from the data streams of each actuator, and combines the position feedback data provided by the optical encoder to correct the time calculation results. The system then determines the action duration of each actuator from triggering to completion within a single action cycle. According to the time sequence of the action cycle, the multiple action durations of the same actuator are arranged to form the action duration sequence of each actuator.

[0049] Optionally, based on the above embodiments, the multiple actuators of the sorting machine may include: an X-axis actuator and a Y-axis actuator for horizontal movement, a Z-axis actuator for vertical movement, and a rotary actuator for rotation about the Z-axis.

[0050] In this embodiment of the invention, the X-axis actuator can be specifically understood as an actuator used to drive the load to translate horizontally (usually in the front-to-back direction), and is a component for adjusting the horizontal position of the load. The Y-axis actuator can be specifically understood as an actuator used to drive the load to translate in another horizontal direction (usually in the left-to-right direction), and works with the X-axis actuator to move the load to any position on the horizontal plane. The Z-axis actuator can be specifically understood as an actuator used to drive the load to move vertically (up-down direction), and is used to lift and lower the load, and works with the horizontal movement to complete the picking and placement height adjustment of the device. The rotary actuator can be specifically understood as an actuator fixed to the lower end of the Z-axis actuator, which can drive the load to rotate around the Z-axis (vertical axis), and is used to adjust the placement angle of the device to ensure precise fit to the placement requirements.

[0051] The X-axis, Y-axis, Z-axis, and rotary actuators enable precise positioning and angle adjustment of the load in space, meeting the requirements of high-speed and micron-level precision pick-up and placement operations for semiconductor devices. This modular and functionally subdivided actuator configuration facilitates independent acquisition and deviation analysis of the action time of each actuator, providing a foundation for subsequent identification of performance degradation of individual actuators and location of fault roots. This avoids difficulties in troubleshooting caused by actuator functional coupling, thereby improving the targeting and efficiency of sorting machine performance monitoring and optimization.

[0052] S120. The action duration of each actuator is compared with the corresponding preset baseline duration average to determine the action duration deviation sequence of each actuator.

[0053] In this embodiment of the invention, the preset baseline duration average can be specifically understood as: the average value of the action duration of each actuator recorded when the sorting machine is debugged to the optimal state during the New Product Introduction (NPI) stage, which is a benchmark reference value for measuring whether the actuator's operating state is normal. The action duration deviation sequence can be specifically understood as: the dataset formed by arranging the differences between the real-time action duration of each actuator and the corresponding preset baseline duration average in the order of the pick-and-place action cycle, reflecting the temporal trend of the actuator's performance deviating from the baseline state.

[0054] Specifically, the system collects the action duration of each of the sorting machine actuators (such as X-axis, Y-axis, Z-axis and rotary actuators) and compares them one-to-one with the pre-set average reference duration of each actuator. This yields the deviation data of the action duration of each actuator in each pick-and-place action cycle from the reference value. Then, these deviation data are sorted and arranged according to the order of the action cycles to obtain the action duration deviation sequence of each actuator.

[0055] Understandably, the system can first select a matching dynamic benchmark performance curve from the benchmark database based on the current sorting machine identifier, product identifier, and firmware version information, and use the statistical mean of the curve as the preset benchmark duration mean for each actuator. The real-time action duration of the sorting machine's X-axis, Y-axis, Z-axis, and rotary actuator is compared and calculated with the corresponding preset benchmark duration mean to obtain the duration deviation value of a single actuator in a single pick-and-place action cycle. According to the time sequence of the pick-and-place action cycle, multiple deviation values ​​of the same actuator are arranged sequentially to form an action duration deviation sequence that can reflect the trend of actuator performance changes.

[0056] The dynamic benchmark performance curve is determined by filtering from the benchmark database based on the sorting machine identifier, product identifier, and firmware version information, rather than using uniform static benchmark data. This ensures that the selected benchmark is compatible with the current sorting machine's equipment model, the type of product to be sorted, and the firmware version, avoiding distortion in deviation analysis caused by mismatch between the benchmark and actual operating conditions. At the same time, this filtering logic can accurately match the benchmark data used for debugging specific equipment and products during the new product introduction phase, ensuring the authority and relevance of the benchmark data. This provides a reliable reference for subsequent actuator action duration deviation analysis, thereby improving the accuracy of performance prediction and optimization schemes, and effectively identifying early subtle performance drift trends.

[0057] Optionally, based on the above embodiments, the duration of each actuator's operation may include the duration from pickup triggering to placement completion.

[0058] Specifically, the action duration of each actuator in the sorting machine can be defined as a unified time range, that is, the duration between the moment the trigger signal is generated and the actuator starts to act, and the moment the device is placed and the actuator stops acting. This forms a standardized statistical caliber for action duration, ensuring that the duration data of each actuator are comparable.

[0059] By defining the motion duration of each actuator as the continuous duration from pickup trigger to placement completion, a unified and standardized duration statistical standard can be provided for all actuators. This ensures the comparability of motion duration data for X-axis, Y-axis, Z-axis, and rotary actuators, avoiding distortion in deviation analysis caused by differences in statistical ranges. At the same time, this duration range fully covers the motion phase of the actuator participating in a single pickup and placement operation, which can truly reflect the actual working efficiency and performance status of the actuator. This provides accurate and reliable basic data for subsequent comparison with preset benchmark durations and generation of motion duration deviation sequences, thereby improving the accuracy of performance degradation trend identification.

[0060] Optionally, based on the above embodiments, determining the action duration sequence of each of the multiple actuators in the sorting machine used to perform semiconductor device pick-and-place actions according to the real-time data stream may include:

[0061] The action events with timestamps are obtained from the real-time data streams of each of the actuators; wherein the action events include pick-up trigger events and placement completion events; and the action duration sequence of each of the multiple actuators in the sorting machine used to perform the pick-up and placement of semiconductor devices is determined based on the action events and timestamps.

[0062] In this embodiment of the invention, the action event can be specifically understood as: key state node events directly related to the actuator's pick-up and place-down operation, which may include a pick-up trigger event (the starting node of the actuator starting the pick-up action) and a place-down completion event (the ending node of the actuator completing the place-down action).

[0063] Specifically, the timestamped pick-up trigger event and placement completion event are extracted from the real-time data stream of each actuator. For each pick-up and placement action cycle, the timestamp difference between the pick-up trigger event and the placement completion event of the same actuator is calculated to obtain the action duration of the actuator in that cycle. According to the order of the pick-up and placement action cycles, the multiple action durations of the same actuator are arranged in order to determine the action duration sequence of each actuator in the sorting machine.

[0064] Understandably, motion events can also include dwell events. Dwell events can be understood as a supplementary type of motion event to the sorting machine actuator, referring to event nodes where the actuator remains stationary after completing actions such as picking or translation. The dwell time corresponding to a dwell event is a preset delay in the sorting machine's motion (used to suppress vibration, etc.). The system monitors its effectiveness; if wear causes an increase in the actual required settling time, it will suggest adjusting the motion parameters.

[0065] By collecting timestamps of dwell events, the static duration within a single pick-up / placement cycle of the actuator can be calculated, and the gaps in the action connections of each actuator can be analyzed. By optimizing timing coordination, the total cycle time can be shortened to improve equipment throughput. It can also help identify abnormal operating conditions of the equipment. If the dwell time exceeds the preset benchmark range, it may correspond to problems such as vacuum adsorption detection timeout or positioning correction failure. Combining pick-up trigger and placement completion events can more accurately locate hidden faults. It can improve the full-cycle analysis of action duration sequences, construct full-cycle time-series data of motion duration and dwell time, make up for the limitations of only analyzing the duration of the motion phase, and improve the comprehensiveness of performance evaluation. It can also guide equipment maintenance and parameter calibration. By monitoring the trend of dwell time changes, the suitability of control parameters can be judged. For example, a gradual increase in dwell time after placement may be due to wear of mechanical parts leading to a decrease in reset accuracy. Based on this, maintenance plans can be formulated in advance to avoid the expansion of faults.

[0066] By acquiring timestamped pick-up trigger events and placement completion events from the real-time data streams of each actuator, and determining the action duration sequence of each actuator based on these events and their corresponding timestamps, the statistical interval of actuator action duration can be defined with standardized time nodes. This ensures the accuracy and consistency of action duration calculation and avoids errors caused by manually defining time ranges. At the same time, relying on the accuracy of timestamps, the duration of a single actuator action can be directly obtained through the event time difference. Arranged in the order of the work cycle to form a sequence, this provides a highly reliable time-series data foundation for subsequent comparison with dynamic benchmark durations and generation of action duration deviation sequences. This improves the accuracy and efficiency of actuator performance degradation trend identification and fault location.

[0067] Furthermore, based on the above embodiments, after determining the action duration sequence of each of the multiple actuators in the sorting machine used to perform the semiconductor device pick-and-place action according to the real-time data stream, it may further include:

[0068] Acquire the sensing data of the optical encoder of each actuator, and determine the position and corresponding time of the actuator based on the sensing data; calculate the action duration of each actuator based on the position and corresponding time of each actuator; correct and supplement the action duration sequence based on the calculated action duration.

[0069] In this embodiment of the invention, the optical encoder sensing data can be specifically understood as: pulse signals or digital data collected by the optical encoder in the sorting machine sensing module to reflect the real-time position of the actuator, used for actuator motion trajectory and position calibration.

[0070] Specifically, after determining the action duration sequence of each actuator in the sorting machine based on the real-time data stream, the system additionally executes an optimization process for the action duration sequence.

[0071] Acquire the optical encoder sensing data corresponding to each actuator, and determine the position of each actuator at different time points during the motion by analyzing the sensing data. Based on the position and corresponding time data of the actuator from the starting position corresponding to the pickup trigger to the ending position corresponding to the placement completion, recalculate the motion duration that is more consistent with the actual motion state.

[0072] The calculation results are used to correct the original motion duration sequence generated solely based on motion event timestamps, correcting duration deviations caused by factors such as signal delays. At the same time, the missing actuator motion sub-stage duration data in the original sequence can be supplemented to form a more complete and accurate motion duration sequence, providing high-quality data support for subsequent performance deviation analysis.

[0073] By acquiring optical encoder sensing data from each actuator and determining its position and corresponding time, the motion duration is calculated to correct and supplement the motion duration sequence. The optical encoder can provide position and time correlation data with micron-level accuracy, effectively correcting errors caused by signal transmission delays and actuator start-stop response lags when calculating duration solely based on the timestamps of the trigger and placement events. Based on position and time data, the duration information of motion sub-stages such as actuator acceleration, constant speed, or deceleration can also be supplemented, ensuring that the motion duration sequence not only includes the overall motion duration but also covers the time sequence data of subdivided motion stages, improving the accuracy and completeness of the motion duration sequence and providing reliable data support for subsequent actuator performance degradation trend analysis and fault root cause location.

[0074] S130. Based on the action duration deviation sequence of each actuator, the system is processed according to the trained performance prediction model to determine the predicted change in the indexing time of the sorting machine.

[0075] In this embodiment of the invention, the performance prediction model can be specifically understood as: an algorithm model (such as a machine learning model) trained based on the historical operating data of the sorting machine and the performance degradation law of the actuator, which has the ability to output performance change prediction results based on input deviation data. The indexing time can be specifically understood as: the complete cycle time for the sorting machine to complete a single semiconductor device action (such as pick-and-place), used to measure the throughput and operating efficiency of the equipment.

[0076] Specifically, after obtaining the action duration deviation sequence of each actuator, the system inputs the sequence into the pre-trained performance prediction model for data processing. The model combines the correlation between actuator performance and indexing time to calculate and output the predicted changes in the sorting machine's indexing time, thereby judging the future operating efficiency trend of the equipment.

[0077] Meanwhile, the system can visualize the predicted changes and corresponding guiding optimization instructions through a graphical user interface (GUI). While presenting the optimization instructions, the interface can also simultaneously display key information such as the predicted future degradation trend, confidence score, root cause analysis hypothesis, and predicted throughput improvement, making it easier for operators to intuitively grasp the equipment performance status, evaluate the effectiveness of optimization solutions, and thus make accurate operation and maintenance decisions.

[0078] Specifically, the graphical user interface can be understood as a visual interactive interface for operators, used to display prediction results and guiding optimization instructions, and to support viewing equipment performance-related data.

[0079] The confidence score can be understood as a quantitative assessment of the reliability of the performance prediction model's output results, reflecting the credibility of the prediction conclusions. The root cause analysis hypothesis can be understood as a hypothesis derived from the deviation sequence and prediction results, suggesting possible causes of actuator performance degradation (such as wear of mechanical components or drift in control parameters). The predicted throughput improvement can be understood as an estimate of the increased processing capacity per unit time that the optimized equipment can achieve, based on the predicted changes in indexing time.

[0080] Optionally, based on the above embodiments, processing is performed according to the action duration deviation sequence of each actuator, based on a trained performance prediction model, to determine the predicted change in the sorting time of the sorting machine. This may include:

[0081] Based on the action duration deviation sequence of each actuator and the sensor detection sequence, the system is processed using a trained performance prediction model to determine the predicted change in the indexing time of the sorting machine; wherein, the sensor detection sequence includes the nozzle action duration sequence and the drive current characteristic value of the actuator.

[0082] In this embodiment of the invention, the sensor detection sequence can be specifically understood as: a time-series dataset of key operating parameters, excluding actuator action duration, collected by the sorting machine's sensing module. This dataset may include a nozzle action duration sequence and actuator drive current characteristic values, serving as important supplementary data for evaluating the equipment's operating status. The nozzle action duration sequence can be specifically understood as: a sequence formed by arranging the duration of the nozzle's actions to pick up and release semiconductor devices according to the work cycle, reflecting the efficiency and stability of processes such as vacuum adsorption and release. The actuator drive current characteristic values ​​can be specifically understood as: indicators obtained through statistical analysis of the actuator drive motor current data, such as peak current or average current, which can indirectly reflect changes in the actuator's mechanical load and the degree of component wear.

[0083] Specifically, the process of determining the predicted changes in the sorting time of the sorting machine based on the action duration deviation sequence of each actuator can be further supplemented by the sensor detection sequence as a supplementary input. Specifically, the action duration deviation sequence of each actuator is combined with the sensor detection sequence containing the nozzle action duration sequence and the actuator drive current characteristic value, and then input together into the trained performance prediction model for analysis and processing. By exploring the intrinsic relationship between action duration deviation, nozzle action efficiency, actuator drive current characteristics and sorting time, the model can more accurately output the predicted changes in the sorting time of the sorting machine.

[0084] By combining the action duration deviation sequences of each actuator with sensor detection sequences containing nozzle action duration sequences and actuator drive current characteristic values, and inputting them into a trained performance prediction model, the predicted changes in the sorting machine's indexing time are determined. This overcomes the limitations of relying solely on action duration data and enables joint analysis of multi-dimensional operating parameters. The nozzle action duration sequence reflects the efficiency of the device's pick-up and release process, while the actuator drive current characteristic values ​​indirectly reflect mechanical load and component wear. The complementary verification of these two with the action duration deviation sequences allows the model to more comprehensively capture key factors affecting indexing time, improving the accuracy and reliability of indexing time prediction changes. Simultaneously, the fusion analysis of multi-dimensional data provides richer evidence for root cause analysis, helping operators to quickly locate the causes of indexing time changes and thus formulate more targeted equipment optimization and maintenance strategies.

[0085] The technical solution of this invention, by acquiring the action duration sequence of multiple actuators from the sorting machine in real time, can capture the timing performance of each actuator in actual operation, avoiding the influence of randomness in single-cycle data. Furthermore, by comparing the real-time action duration with the average of a preset benchmark duration, the action duration deviation sequence of each actuator is determined. Based on a trained performance prediction model, the deviation sequence is processed to determine the predicted changes in the grading time. This allows for the prediction of future performance degradation trends, such as abnormal events, degradation magnitude, and occurrence cycles, from early subtle deviations, rather than passively responding only after severe performance degradation. It can identify potential progressive performance degradation risks of the sorting machine in advance, providing a basis for subsequent targeted optimization, thereby reducing unplanned downtime, ensuring equipment throughput and product yield, and reducing reliance on manual real-time monitoring and experience-based judgment, thus improving the stability and efficiency of the semiconductor device testing and sorting process.

[0086] Example 2

[0087] Figure 2 This is a flowchart illustrating another performance prediction and optimization method for a semiconductor device testing system sorter provided in Embodiment 2 of the present invention. This embodiment is a refinement of the performance prediction and optimization method for the semiconductor device testing system sorter in the above embodiments. Figure 2 As shown, the method includes:

[0088] S210. Obtain a real-time data stream from the sorting machine, and determine the action duration sequence of each of the multiple actuators in the sorting machine used to perform semiconductor device pick-and-place actions based on the real-time data stream.

[0089] The action duration sequence includes the action duration of multiple pick-up and put-down action cycles.

[0090] S220. Based on the metadata of the sorting machine, query and obtain the preset benchmark data of the sorting machine that has been stored and corresponds to the metadata from the preset benchmark library.

[0091] The metadata includes the model, identifier, and operating environment parameters of the sorting machine; the preset benchmark data includes the average benchmark duration and statistical tolerance range of each actuator's motion phase, the optimal control parameters of each actuator, and the environmental parameters recorded during benchmark acquisition; the preset benchmark data is the benchmark data determined when the sorting machine is configured for new product introduction.

[0092] In this embodiment of the invention, the sorting machine metadata can be specifically understood as: a set of basic data used to identify the sorting machine's own attributes and operating conditions, which may include the sorting machine model (equipment hardware specifications and configuration information), sorting machine identifier (equipment unique identification code), and operating environment parameters (such as external conditions that affect the equipment's operating status, such as operating temperature, humidity, and air pressure).

[0093] The preset benchmark library can be understood as a database specifically storing benchmark data recorded when the sorting machine is debugged to its optimal state during the NPI (New Product Integration) phase, providing a reference standard for subsequent equipment performance comparison and optimization. Preset benchmark data can be understood as benchmark reference data retrieved from the preset benchmark library that matches the current sorting machine's metadata. This data can include: the average benchmark duration and statistical tolerance range for each actuator's motion phase, the optimal control parameters for each actuator, and the environmental parameters recorded during benchmark acquisition. This data is determined during the new product introduction and configuration phase of the sorting machine and is standardized and specific.

[0094] Specifically, based on the metadata of the current sorting machine, the system performs a matching query in the preset benchmark library to extract the preset benchmark data of the same type or the same identifier of the stored sorting machine corresponding to the metadata, providing a reference for subsequent actuator action time deviation analysis and performance prediction.

[0095] S230. The action duration of each actuator is compared with the corresponding preset baseline duration average to determine the action duration deviation sequence of each actuator.

[0096] S240. Based on the action duration deviation sequence of each actuator, the system is processed according to the trained performance prediction model to determine the predicted change in the indexing time of the sorting machine.

[0097] Optionally, based on the above embodiments, the performance prediction model can be a machine learning model capable of processing time-series data.

[0098] Specifically, the performance prediction model can be a machine learning model capable of processing time-series data, such as Long Short-Term Memory (LSTM) networks and Transformer models. LSTM can effectively capture long-term dependencies in the actuator action duration deviation sequence through a gating structure, accurately identifying the slow trend of performance degradation. The Transformer model relies on a self-attention mechanism to achieve global correlation analysis of multi-dimensional sensor data (such as action duration sequences and current characteristic values). Both can adapt to the analysis needs of sorting machine time-series operation data, providing reliable algorithmic support for the output of the indexing time prediction change.

[0099] The performance prediction model employs a machine learning model capable of processing time-series data. This type of model has the ability to deeply mine multi-dimensional time-series data, such as actuator action duration deviation sequences and sensor detection sequences arranged in chronological order. It can effectively capture the long-term dependencies and performance change patterns hidden in the data, avoiding the problem of missing subtle decay trends that are difficult to identify by traditional statistical models. At the same time, the time-series machine learning model can adapt to the operating conditions and data characteristics of different sorting machines without the need for manual adjustment of a large number of analysis parameters, improving the accuracy and generalization ability of the sorting time prediction, and providing more reliable algorithmic support for equipment performance prediction and preventive maintenance.

[0100] Optionally, based on the above embodiments, the predicted changes in the sorting time of the sorting machine may include: abnormal events, sorting time increments, predicted cycle counts, and confidence levels.

[0101] In this embodiment of the invention, abnormal events can be specifically understood as: future faults or performance degradation events predicted by the model that may cause abnormal fluctuations in the sorting time of the sorting machine, such as actuator mechanical jamming, insufficient vacuum in the suction nozzle, and abnormal increase in drive current. The sorting time increment can be specifically understood as: based on the current actuator performance deviation trend, the increase in the sorting time of a single pick-and-place operation of the sorting machine relative to the baseline time within a future cycle predicted by the model, used to measure the degree of equipment efficiency decline. The predicted cycle number can be specifically understood as: the number of cycles in which abnormal events may occur, or the number of sorting machine operation cycles in which the sorting time increment reaches the warning threshold, as output by the performance prediction model, providing operators with clear time prediction nodes.

[0102] The predicted changes in the sorting time of a sorting machine can include abnormal events, sorting time increments, predicted cycle counts, and confidence levels, rather than just a single time change value. This allows for a comprehensive depiction of the equipment's future operating status from four dimensions: fault type, efficiency degradation degree, time point, and result reliability. It provides operators with multi-dimensional decision-making support. Identifying abnormal events allows for early identification of potential fault causes, facilitating the development of targeted maintenance plans. The sorting time increment quantifies the degree of equipment efficiency degradation, helping to determine whether immediate parameter optimization is necessary. The predicted cycle count accurately pinpoints the operational cycle points where abnormalities occur, enabling precise planning of preventative maintenance. The confidence level helps operators assess the reliability of the prediction results, avoiding unnecessary downtime or wasted maintenance costs due to misjudgments.

[0103] Furthermore, based on the above embodiments, the predicted changes in the sorting time of the sorting machine may also include: abnormal causes, and / or, optimized parameters and the amount of optimization parameter adjustment.

[0104] In this embodiment of the invention, the cause of the anomaly can be specifically understood as: the inducement directly related to the predicted abnormal event, derived from multi-dimensional input data of the performance prediction model, combined with historical equipment failure cases and operating patterns, such as wear of actuator mechanical parts, insufficient vacuum pressure of the suction nozzle, and drift of control parameters. The optimization parameter can be specifically understood as: the control parameter that can improve the equipment's operating state and reduce the indexing time for the predicted cause of the anomaly or indexing time increment, such as the actuator's motion acceleration parameter, the suction nozzle's vacuum adsorption duration parameter, and the timing coordination parameter of each actuator's action. The adjustment amount of the optimization parameter can be specifically understood as: the specific value or proportion of adjusting the optimization parameter to restore the equipment to its baseline operating state or optimal efficiency range, which is a quantitative control basis that can directly guide on-site operation.

[0105] By predicting changes in the sorting time of the sorting machine, this method further incorporates the causes of anomalies, optimized parameters, and adjustments to those parameters, building a closed-loop guidance system for predicting anomalies, locating root causes, and developing optimization solutions. This overcomes the limitations of simply providing trend warnings: clearly identifying the causes of anomalies helps operators quickly pinpoint the triggers for changes in sorting time, avoiding prolonged downtime due to blind troubleshooting; providing optimized parameters and adjustments directly provides actionable operational guidelines for equipment parameter calibration, eliminating the need for additional debugging tests and improving the efficiency and accuracy of operational decisions; and the combined output of anomaly causes and optimization solutions reduces reliance on operator experience, improving equipment optimization and maintenance efficiency.

[0106] Optionally, based on the above embodiments, the abnormal event may include at least one of the following: no event, indexing time increase event, actuator timeout event, placement accuracy failure event, pickup failure event, sequence out-of-step event, and motor overcurrent warning event;

[0107] The causes of the anomaly may include at least one of the following: mechanical wear, belt tension drift, bearing wear, pneumatic system performance degradation, vacuum leakage, filter blockage, pressure regulator drift, servo parameter tuning drift, environmental parameter drift, and timing changes caused by firmware.

[0108] The optimization parameters may include at least one of the following: X-axis acceleration, X-axis deceleration, Y-axis acceleration, Z-axis lifting vacuum setting value, Z-axis lowering vacuum release timing value, X-axis movement delay timing value, Y-axis start-up waiting X-axis completion timing value, and X-axis servo proportional gain.

[0109] In this embodiment of the invention, "no event" can be specifically understood as: the equipment is operating normally without any abnormal working conditions. "Increased indexing time" can be specifically understood as: the complete cycle time of a single pick-and-place operation by the sorting machine exceeds a baseline threshold. "Actuator timeout" can be specifically understood as: the time taken for the actuator to complete actions such as picking and placing exceeds a preset duration. "Placement accuracy failure" can be specifically understood as: the semiconductor device placement position deviation exceeds the allowable range. "Pick-up failure" can be specifically understood as: the nozzle fails to successfully pick up the device or the device falls off after picking it up. "Sequence out-of-synchronization" can be specifically understood as: the timing coordination of multiple actuator actions is disordered, resulting in misaligned action connections. "Motor overcurrent warning" can be specifically understood as: the operating current of the actuator drive motor exceeds the rated range, triggering a protection warning.

[0110] Servo proportional gain can be specifically understood as a parameter in the proportional control loop of the sorting machine actuator control system, used to characterize the servo motor's response sensitivity to position deviation. Vacuum setpoint can be specifically understood as the target vacuum threshold that the vacuum system must achieve when the sorting machine nozzle adsorbs semiconductor devices, pre-configured by the equipment control system. Vacuum release timing can be specifically understood as the delay time for maintaining a vacuum state after the sorting machine nozzle places the device at the target position, or the time parameter for triggering vacuum cutoff. Drift can be specifically understood as an external environmental parameter affecting the normal operation of the equipment deviating from the equipment's calibrated or preset reference range, exhibiting a continuous and slow changing trend, rather than a sudden large fluctuation. Environmental parameter drift can include temperature drift, humidity drift, and air pressure drift.

[0111] By clearly defining the specific scope of abnormal events, abnormal causes, and optimization parameters, abnormal events include all types of equipment abnormalities under various operating conditions, such as no events, indexing time increase events, actuator timeout events, placement accuracy failure events, pickup failure events, sequence synchronization loss events, and motor overcurrent warning events. Abnormal causes include multi-dimensional fault triggers such as mechanical wear, belt tension drift, bearing wear, pneumatic system performance degradation, vacuum leakage, filter blockage, pressure regulator drift, servo parameter setting drift, environmental parameter drift, and timing changes caused by firmware. Optimization parameters include X-axis acceleration, X-axis deceleration, Y-axis acceleration, and Z-axis lifting vacuum setting values. Key control parameters such as the Z-axis descent vacuum release timing value, the X-axis movement delay timing value, the Y-axis start-up waiting X-axis completion timing value, and the X-axis servo proportional gain can establish a standardized technical link for anomaly identification, fault root cause classification and location, and parameter optimization and targeted measures. Comprehensive anomaly classification enables complete monitoring of equipment operating status and avoids overlooking potential fault risks. The correspondence between anomaly causes and optimization parameters can shorten the time for fault diagnosis and parameter adjustment, reduce reliance on operator experience, and ensure that the equipment can quickly return to its optimal operating state, improving the sorting machine's operational stability and efficiency.

[0112] Furthermore, based on the above embodiments, after determining the predicted change in the sorting time of the sorting machine, the method may further include:

[0113] The optimized parameters and their adjustment values ​​are transmitted to the sorting machine to control the machine to adjust the actuator parameters; or the optimized parameters and their adjustment values ​​are provided to the user, and the machine is controlled to adjust the actuator parameters in response to the user's selection confirmation.

[0114] Specifically, after determining the predicted changes in the sorting time of the sorting machine, the system can perform parameter optimization in two ways: directly transmit the optimized parameters and the adjustment amount to the sorting machine to automatically complete the actuator parameter adjustment; or provide the optimized parameters and adjustment amount to the user, and control the sorting machine to adjust the parameters after the user selects and confirms.

[0115] Understandably, after determining the predicted changes in the sorting time of the sorting machine, the system can first determine whether the confidence score of the predicted future attenuation trend exceeds a preset threshold. Only when it exceeds the preset threshold will a guiding optimization instruction (based on the optimization parameters and the adjustment amount of the optimization parameters) be generated. This can effectively filter out low-reliability prediction results, avoid blind adjustment of equipment parameters due to misjudgment instructions, and ensure the accuracy and safety of parameter optimization.

[0116] Before generating instructions, the kinematic effects of preventative adjustments can be simulated and verified. For example, relying on the digital twin model or kinematic simulation algorithm of the sorting machine, the optimization parameters and adjustment amounts to be executed can be input into the simulation system. In the virtual environment, the motion trajectory, action sequence, spatial positional relationship between various components, and the interaction process of picking up and placing semiconductor devices can be simulated. By monitoring the distance between the actuator and the equipment rack and other actuators during the simulation process, as well as verifying the positional accuracy during device picking and placement, it can be predicted whether parameter adjustments will cause mechanical collisions, device offsets, or picking and placing failures. Ultimately, it can be ensured that the output optimization instructions will not have a negative impact on the safety of equipment operation and the accuracy of operation when actually executed.

[0117] Then, the system performs parameter optimization in two ways. After the parameter adjustment is completed, the system can collect the operating data of the adjusted equipment (such as changes in indexing time, whether abnormal events are eliminated, or the accuracy of actuator movements) as feedback data characterizing the implementation effect of the optimization instructions. The system then associates this feedback data with the corresponding multivariate deviation vector (such as actuator movement duration deviation and sensor detection data deviation) as labeled training samples to iteratively train the performance prediction model, continuously improving the model's prediction accuracy and the reliability of the optimization scheme.

[0118] In a specific example, model training can use performance degradation events as labels. These labels can include various types: exceeding the gradation time threshold (recording the delay magnitude and corresponding loop number), actuator timing failures (such as Z-axis lift timeout), placement accuracy failures (such as misalignment due to insufficient Theta-axis positioning), vacuum system failures (such as vacuum establishment timeout causing pickup failure), motor or drive anomalies (such as X-axis overcurrent peaks), actuator synchronization loss (such as reduced overlap between X and Y axis movements), and events increasing settling time (such as longer stabilization time required for the Y-axis). Training data can include multivariate sequence data prior to the event, along with corresponding corrective measures, allowing the model to learn early characteristics of the event and generate optimization schemes.

[0119] Typically, the training label vector format can be [event_code, delta_t, horizon, cause_code, param_code, delta_p, confidence]. `event_code` encodes the type of performance degradation event using an integer, such as 1 for increased indexing time, 4 for pickup failure, 5 for actuator sequence out of step, and 0 for no abnormal event within the prediction cycle. `delta_t` records the actual measured indexing time extension value in milliseconds using a floating-point number; for indirect delay events (such as mounting deviation), it is converted to an equivalent time loss. `horizon` records the number of sorting machine cycles from the end of the deviation sequence to the occurrence of the event, specifying the time node of the event. `cause_code` encodes the root cause of the event using an integer, such as 21 for vacuum leakage or filter blockage, 12 for bearing wear, and 40 for temperature drift. `param_code` (Optimization Parameter Code): Uses integers to encode the key control parameters of the event. For example, 201 represents the Z-axis vacuum setting value, and 301 represents the delay timer after X-axis movement. `delta_p` (Optimization Parameter Adjustment): Uses signed floating-point numbers to record the adjustment range of the parameter. The unit varies depending on the parameter type (e.g., bar for pressure and milliseconds for timers). `+` indicates increasing the parameter, and `-` indicates decreasing the parameter. `confidence` (Confidence Level): Uses floating-point numbers from 0.0 to 1.0 to represent the certainty of the correlation between the cause and the measure. 1.0 means the adjustment measure directly solves the problem, while less than 1.0 means the measure is part of comprehensive maintenance, indicating a weaker causal relationship. Correspondingly, during training, the model input can be a multivariate bias feature sequence containing a preset number (e.g., 500) running cycles, and the output is a corresponding 7-field label vector.

[0120] The performance prediction and optimization method of this invention can also continuously monitor the calibration frequency (e.g., whether the equipment needs to perform calibration more frequently to maintain accuracy) and calibration duration (e.g., whether the completion time of a single calibration process is extended). If an increase in calibration frequency or calibration duration is detected, it indicates that the actuator may have mechanical degradation problems (e.g., bearing wear or encoder drift). Accordingly, to ensure the motion accuracy of the actuators (X, Y, Z, and Theta axes), the semiconductor test sorting machine will periodically perform a reference correction process, i.e., a homing sequence or calibration sequence. For example, the sorting machine will drive the actuators back to the mechanical zero point or calibrate the position reference of the optical encoder every fixed number of production cycles, thereby eliminating positioning deviations caused by long-term operation.

[0121] It is understandable that multiple actuators in a sorting machine (such as X-axis movement and Z-axis lifting) can operate in parallel and collaboratively in time, meaning there is overlap between actuators. This design, by allowing the actions of different actuators to overlap, can significantly reduce the total time consumed in a single pick-up and place-down cycle, avoiding the time wasted due to sequential actuator actions and improving throughput. The technical solution of this invention does not eliminate this action overlap to test the performance of a single actuator, but rather focuses on monitoring the drift of the overlap time. For example, if the Y-axis actuator should start its action 2 milliseconds after the X-axis actuator completes its movement, and if the actual detected start delay becomes 5 milliseconds, it will lead to a shortened effective overlap time, resulting in an increase in indexing time. To address this, the system will adjust the preset delay timer parameters (such as shortening the delay time after the X-axis action is completed) to restore the actuators to the optimal action overlap state, thereby optimizing performance and preventing the continuous decay of indexing time.

[0122] The system offers two execution modes: automatic parameter adjustment and manual adjustment after user confirmation. The automatic adjustment mode directly transmits optimized parameters and adjustment amounts to the sorting machine, quickly completing actuator parameter calibration without manual intervention. This shortens the time from abnormal warnings to performance recovery, improving the continuous operation efficiency of the production line. The user confirmation mode provides the optimized plan back to the operator, allowing the user to determine whether to execute the adjustment based on the actual on-site conditions. This avoids misoperation due to model prediction deviations or special operating conditions, ensuring equipment operation safety. Furthermore, the optional configuration of these two modes adapts to the management standards and maintenance habits of different production lines, enhancing the practicality and compatibility of the technical solution while balancing equipment maintenance efficiency and operational safety.

[0123] Furthermore, based on the above embodiments, after controlling the sorting machine to adjust the actuator parameters, it may further include:

[0124] Monitor the sorting time of the sorting machine; if the monitored sorting time decay exceeds the set threshold or a fault condition is detected, the rollback program is automatically started to restore the parameter configuration to the state before adjustment.

[0125] In this embodiment of the invention, the indexing time decay amplitude can be specifically understood as: the change in the indexing time of a single operation cycle of the sorting machine after parameter adjustment compared to before adjustment, used to measure the effect of parameter optimization. If the monitored indexing time decay amplitude exceeds a set threshold, it means that after parameter adjustment, the indexing time of a single operation cycle of the sorting machine has changed beyond the allowable range compared to before adjustment, indicating that the adjusted parameter configuration does not match the current operating conditions of the equipment.

[0126] Specifically, after the system controls the sorting machine to adjust the actuator parameters, it continuously monitors the real-time indexing time of the sorting machine. At the same time, it can detect whether the equipment has fault states such as actuator timeout, pickup failure, or sequence out-of-synchronization by combining preset abnormal event types. If the monitored indexing time decay exceeds the set threshold, or any fault state is detected, the rollback program is automatically started to restore the actuator parameters to the configuration before adjustment. This avoids the equipment operating efficiency from decreasing or the fault from expanding due to improper parameter optimization, and ensures the stability and safety of the sorting machine operation.

[0127] By continuously monitoring the indexing time and equipment operating status after the actuator parameters are adjusted, and automatically starting the rollback program to restore the parameters before adjustment when the indexing time decays too much or a fault is detected, the system can promptly correct improper parameter optimization operations, avoid decreased equipment operating efficiency and increased fault risk due to optimization errors, and effectively avoid production interruption caused by parameter adjustments, ensuring the continuous and stable operation of the sorting machine and reducing maintenance costs and production losses.

[0128] The technical solution of this invention, by collecting the action duration sequence of multiple actuators from the sorting machine in real time, can capture the timing performance of each actuator in actual operation, avoiding the influence of randomness in single-cycle data. Based on metadata such as the sorting machine's model, identification, and operating environment parameters, corresponding preset benchmark data is retrieved from a preset benchmark library to ensure that the selected benchmark data matches the equipment characteristics and operating scenario of the current sorting machine, avoiding comparison deviations caused by the generalization of benchmark data. At the same time, the statistical tolerance range and environmental parameters contained in the benchmark data can provide a more comprehensive reference dimension for subsequent deviation analysis, helping to accurately distinguish between normal fluctuations and abnormal deviations related to performance degradation. The optimal control parameters can also lay the foundation for the subsequent generation of targeted optimization schemes, thereby improving the accuracy of performance prediction and the effectiveness of optimization measures, and ensuring the scientific nature of sorting machine performance evaluation and intervention. By comparing the real-time action duration with the average of the preset baseline duration, the action duration deviation sequence of each actuator is determined. Based on the trained performance prediction model, the deviation sequence is processed to determine the predicted changes in the indexing time. This allows for the prediction of future performance degradation trends, such as abnormal events, degradation magnitude, and occurrence cycle, from early subtle deviations, rather than passively responding only after performance has severely degraded. It can identify potential progressive performance degradation risks of the sorting machine in advance, providing a basis for subsequent targeted optimization, thereby reducing unplanned downtime, ensuring equipment throughput and product yield, and reducing reliance on manual real-time monitoring and experience judgment, thus improving the stability and efficiency of the semiconductor device testing and sorting process.

[0129] Example 3

[0130] Figure 3This is a flowchart illustrating another performance prediction and optimization method for a semiconductor device testing system sorter provided in Embodiment 3 of the present invention. This embodiment is a refinement of the performance prediction and optimization method for the semiconductor device testing system sorter in the above embodiments. Figure 3 As shown, the method includes:

[0131] S310. Obtain a real-time data stream from the sorting machine, and determine the action duration sequence of each of the multiple actuators in the sorting machine used to perform semiconductor device pick-and-place actions based on the real-time data stream.

[0132] The action duration sequence includes the action duration of multiple pick-up and put-down action cycles.

[0133] S320. The action duration of each actuator is compared with the corresponding preset baseline duration average to determine the action duration deviation sequence of each actuator.

[0134] S330. Based on the action duration deviation sequence of each actuator, the system is processed according to the trained performance prediction model to determine the predicted change in the indexing time of the sorting machine.

[0135] S340. Based on the predicted changes in the sorting time of the sorting machine and the preset domain mapping rule algorithm, determine the analysis result of the sorting machine.

[0136] The analysis results include at least one of the following: description of the cause of the anomaly, throughput impact risk, collision risk, historical performance analysis, processing program status analysis, and estimated throughput change.

[0137] In this embodiment of the invention, the preset domain mapping rule algorithm can be specifically understood as: a dedicated algorithm that pre-sets the correlation and matching of the indexing time prediction data with technical elements such as equipment operating status, abnormal events, abnormal causes, and optimization parameters, based on the operating characteristics of the semiconductor test sorting machine.

[0138] The analysis results can be specifically understood as: comprehensive conclusions reflecting the operating status of the equipment based on the predicted data and the output of the mapping algorithm. These conclusions may include at least one of the following: description of the cause of the anomaly, throughput impact risk, collision risk, historical performance risk, processing program status analysis, and estimated throughput change.

[0139] S350. Display the predicted changes and analysis results on the user's graphical interface.

[0140] Specifically, the predicted changes in the sorting time of the sorting machine are obtained, and the predicted data is input into a preset domain mapping rule algorithm for calculation and analysis. The algorithm associates and matches the predicted data with various status indicators of the equipment operation to determine the analysis results. Finally, the predicted changes in the sorting time and the corresponding analysis results are displayed synchronously on the user's graphical interface.

[0141] Figure 4 This is a schematic diagram of an actuator timing sequence comparison applicable to an embodiment of the present invention, as shown below. Figure 4 As shown, with the test cycle as the horizontal axis and time (milliseconds) as the vertical axis, the three states of actuator timing are displayed: the NPI baseline (optimal) represents the ideal optimal level of actuator timing, the current performance corresponds to the current actual timing state of the actuator, and the predicted drift trajectory shows the decay trend of actuator timing as the test cycle progresses; the figure also marks the "intervention" node (corresponding to about 600 test cycles) and the predicted result of "22 millisecond delay" after intervention, intuitively comparing the change process of actuator timing from the optimal baseline, the current state to the predicted drift, reflecting the trend of timing decay and the delay impact after intervention.

[0142] In a specific example, the graphical user interface may include the following: the top of the interface is marked with a red warning bar indicating "Predicted performance degradation," and displays information such as the device's sorter, product (here referring to the device under test, such as a MEMS accelerometer), and firmware (the device's built-in control program) and current cycle; the main body of the interface can be divided into three modules: "Predicted Impact" with corresponding prompts such as "Prevent an 8% increase in indexing time within the next 48 hours"; "Root Cause" displays the root cause such as "Gradual vacuum degradation due to filter clogging" with a confidence level of 92%; "Suggested Actions" such as "Increase the vacuum setpoint by 0.2 bar," and the bottom may also have "Schedule Maintenance Window" and "Apply Now" operation buttons for operators to select.

[0143] In a specific example, a performance prediction and optimization system for a semiconductor device testing system sorting machine can include a microelectromechanical system (MEMS) testing sorting machine, a log access module, a new product import benchmark database, a machine learning engine, a recommendation engine, and an output interface (graphical user interface and control application programming interface). The sorting machine generates time-series equipment logs and receives parameter adjustment instructions. The log access module collects information such as event timestamps, actuator timing, and sensor data from the sorting machine via industry-standard protocols. The new product import benchmark database stores the benchmark curve of the sorting machine's optimal operating state. The machine learning engine uses a time-series model to calculate multivariate deviation scores and predict performance degradation trends. The recommendation engine transforms the prediction results into priority-ranked preventative parameter adjustment schemes. The output interface is used to display information to users or autonomously issue adjustment instructions. This system ensures the stability of the overall equipment efficiency by proactively preventing performance degradation, issuing early warnings before component failures to reduce downtime, and integrating kinematic simulation functions to protect product yield. It transforms the sorting machine's control paradigm from passive correction to active optimization, possesses cross-platform compatibility to adapt to mainstream sorting machine manufacturers' equipment, improves throughput, and reduces operation and maintenance costs.

[0144] The technical solution of this invention, by acquiring the action duration sequence of multiple actuators from the sorting machine in real time, can capture the timing performance of each actuator in actual operation, avoiding the influence of randomness in single-cycle data. Furthermore, by comparing the real-time action duration with the average of a preset benchmark duration, the action duration deviation sequence of each actuator is determined. Based on a trained performance prediction model, the deviation sequence is processed to determine the predicted changes in the grading time. This allows for the prediction of future performance degradation trends, such as abnormal events, degradation magnitude, and occurrence cycles, from early subtle deviations, rather than passively responding only after severe performance degradation. It can identify potential progressive performance degradation risks of the sorting machine in advance, providing a basis for subsequent targeted optimization, thereby reducing unplanned downtime, ensuring equipment throughput and product yield, and reducing reliance on manual real-time monitoring and experience-based judgment, thus improving the stability and efficiency of the semiconductor device testing and sorting process. After determining the predicted changes in the sorting time of the sorting machine, the algorithm, combined with preset domain mapping rules, further derives analysis results including at least one of the following: description of abnormal causes, throughput impact risk, collision risk, historical performance risk, processing program status risk, and estimated throughput changes. These results are simultaneously displayed on the user's graphical interface, transforming abstract predictive data into intuitive information. This helps users quickly identify key factors, potential risks, and possible production impacts related to performance degradation, eliminating the need for additional time to interpret raw data. Furthermore, risk warnings and impact estimates provide a basis for user decision-making, allowing them to rationally select intervention timing and methods based on actual production needs. This avoids production interruptions caused by blind adjustments and timely mitigation of major risks, improving the convenience and scientific basis of sorting machine performance management and reducing reliance on the experience of professional technicians.

[0145] Example 4

[0146] Figure 5 This is a schematic diagram of the performance prediction and optimization device for a semiconductor device testing system sorting machine provided in Embodiment 4 of the present invention. Figure 5 As shown, the device includes: a sequence determination module 510, a benchmark comparison module 520, and a change prediction module 530, wherein:

[0147] The sequence determination module 510 is used to acquire a real-time data stream from the sorting machine and determine the action duration sequence of each of the multiple actuators in the sorting machine that perform semiconductor device pick-and-place actions based on the real-time data stream; wherein, the action duration sequence includes the action duration of multiple pick-and-place action cycles;

[0148] The benchmark comparison module 520 is used to compare the action duration of each actuator with the corresponding preset benchmark duration average value to determine the action duration deviation sequence of each actuator.

[0149] The prediction change module 530 is used to process the action duration deviation sequence of each actuator based on a trained performance prediction model to determine the predicted change in the sorting time of the sorting machine.

[0150] The technical solution of this invention, by acquiring the action duration sequence of multiple actuators from the sorting machine in real time, can capture the timing performance of each actuator in actual operation, avoiding the influence of randomness in single-cycle data. Furthermore, by comparing the real-time action duration with the average of a preset benchmark duration, the action duration deviation sequence of each actuator is determined. Based on a trained performance prediction model, the deviation sequence is processed to determine the predicted changes in the grading time. This allows for the prediction of future performance degradation trends, such as abnormal events, degradation magnitude, and occurrence cycles, from early subtle deviations, rather than passively responding only after severe performance degradation. It can identify potential progressive performance degradation risks of the sorting machine in advance, providing a basis for subsequent targeted optimization, thereby reducing unplanned downtime, ensuring equipment throughput and product yield, and reducing reliance on manual real-time monitoring and experience-based judgment, thus improving the stability and efficiency of the semiconductor device testing and sorting process.

[0151] Based on the above embodiments, the multiple actuators of the sorting machine may include: an X-axis actuator and a Y-axis actuator for horizontal movement, a Z-axis actuator for vertical movement, and a rotary actuator for rotation about the Z-axis.

[0152] Based on the above embodiments, the duration of each actuator's operation may include the duration from pickup triggering to placement completion.

[0153] Based on the above embodiments, the sequence module 510 is determined to be specifically used for:

[0154] The action events with timestamps are obtained from the real-time data streams of each actuator; wherein the action events include pick-up trigger events and placement completion events; and the action duration sequence of each of the multiple actuators in the sorting machine used to perform semiconductor device pick-up and placement actions is determined based on the action events and timestamps.

[0155] Furthermore, based on the above embodiments, the performance prediction and optimization device for the semiconductor device testing system sorting machine may further include: a sensor data module, an action duration module, and a correction and supplementation module, wherein:

[0156] The sensing data module is used to acquire the sensing data of the optical encoder of each actuator after determining the action duration sequence of each of the multiple actuators in the sorting machine that perform the semiconductor device picking and placing actions according to the real-time data stream, and to determine the position and corresponding time of the actuator according to the sensing data.

[0157] The action duration module is used to calculate the action duration of each actuator based on the position and corresponding time of each actuator.

[0158] The correction and supplementation module is used to correct and supplement the action duration sequence based on the calculated action duration.

[0159] Based on the above embodiments, the change prediction module 530 is specifically used for:

[0160] Based on the action duration deviation sequence of each actuator and the sensor detection sequence, the system is processed using a trained performance prediction model to determine the predicted change in the indexing time of the sorting machine; wherein, the sensor detection sequence includes the nozzle action duration sequence and the drive current characteristic value of the actuator.

[0161] Furthermore, based on the above embodiments, the performance prediction and optimization device for the semiconductor device testing system sorter may further include: a reference module, wherein:

[0162] The benchmark query module is used to retrieve the stored benchmark data of the sorting machine from the benchmark database based on the metadata of the sorting machine before comparing the action duration of each actuator with the corresponding average preset benchmark duration. The metadata includes the model, identifier, and operating environment parameters of the sorting machine. The preset benchmark data includes the average benchmark duration and statistical tolerance range of each actuator's motion phase, the optimal control parameters of each actuator, and the environmental parameters recorded during benchmark acquisition. The preset benchmark data is the benchmark data determined when configuring the sorting machine for new product introduction.

[0163] Based on the above embodiments, the performance prediction model can be a machine learning model capable of processing time-series data.

[0164] Based on the above embodiments, the predicted changes in the sorting time of the sorting machine can include: abnormal events, sorting time increments, predicted cycle counts, and confidence levels.

[0165] Based on the above embodiments, the predicted changes in the sorting time of the sorting machine may also include: abnormal causes, and / or, optimized parameters and the amount of optimization parameter adjustment.

[0166] Based on the above embodiments, abnormal events may include at least one of the following: no event, indexing time increase event, actuator timeout event, placement accuracy failure event, pickup failure event, sequence out-of-step event, and motor overcurrent warning event;

[0167] The causes of the anomaly may include at least one of the following: mechanical wear, belt tension drift, bearing wear, pneumatic system performance degradation, vacuum leakage, filter blockage, pressure regulator drift, servo parameter tuning drift, environmental parameter drift, and timing changes caused by firmware.

[0168] The optimization parameters may include at least one of the following: X-axis acceleration, X-axis deceleration, Y-axis acceleration, Z-axis lifting vacuum setting value, Z-axis lowering vacuum release timing value, X-axis movement delay timing value, Y-axis start-up waiting X-axis completion timing value, and X-axis servo proportional gain.

[0169] Furthermore, based on the above embodiments, the performance prediction and optimization device for the semiconductor device testing system sorting machine may further include: an adjustment module, wherein:

[0170] The adjustment module is used to transmit the optimized parameters and the optimized parameter adjustment amount to the sorting machine after determining the predicted change in the sorting time, so as to control the sorting machine to adjust the actuator parameters; or to provide the optimized parameters and the optimized parameter adjustment amount to the user, and control the sorting machine to adjust the actuator parameters in response to the user's selection confirmation.

[0171] Optionally, based on the above embodiments, the adjustment module may include: a monitoring unit and a rollback unit, wherein:

[0172] A monitoring unit is used to monitor the indexing time of the sorting machine after the parameters of the actuator are adjusted by controlling the sorting machine.

[0173] The rollback unit is used to automatically start the rollback procedure and restore the parameter configuration to the state before adjustment if the detected indexing time decay exceeds the set threshold or a fault condition is detected.

[0174] Optionally, based on the above embodiments, the adjusted module may include: a result determination unit and a display unit, wherein:

[0175] The result determination unit is used to determine the analysis result of the sorting machine based on the predicted change of the sorting time and a preset neighborhood mapping rule algorithm after determining the predicted change of the sorting time of the sorting machine; wherein the analysis result includes at least one of the following: description of the cause of the anomaly, throughput impact risk, collision risk, historical performance analysis, processing program status analysis, and estimated throughput change.

[0176] The display unit is used to display the predicted changes and analysis results on a user graphical interface.

[0177] The performance prediction and optimization device for the semiconductor device testing system sorter provided in this embodiment of the invention can execute the performance prediction and optimization method for the semiconductor device testing system sorter provided in any embodiment of the invention, and has the corresponding functional modules and beneficial effects of executing the method.

[0178] The collection, storage, use, processing, transmission, provision, and disclosure of user personal information involved in the technical solution disclosed herein comply with the provisions of relevant laws and regulations and do not violate public order and good morals.

[0179] Example 5

[0180] Figure 6 A schematic diagram of an electronic device 10, which can be used to implement embodiments of the present invention, is shown. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as personal digital processors, cellular phones, smartphones, wearable devices (e.g., helmets, glasses, watches, etc.), and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely illustrative and are not intended to limit the implementation of the invention described and / or claimed herein.

[0181] like Figure 6 As shown, the electronic device 10 includes at least one processor 11 and a memory, such as a read-only memory (ROM) 12 or a random access memory (RAM) 13, communicatively connected to the at least one processor 11. The memory stores computer programs executable by the at least one processor. The processor 11 can perform various appropriate actions and processes based on the computer program stored in the ROM 12 or loaded into the RAM 13 from storage unit 18. The RAM 13 can also store various programs and data required for the operation of the electronic device 10. The processor 11, ROM 12, and RAM 13 are interconnected via a bus 14. An input / output (I / O) interface 15 is also connected to the bus 14.

[0182] Multiple components in electronic device 10 are connected to I / O interface 15, including: input unit 16, such as keyboard, mouse, etc.; output unit 17, such as various types of displays, speakers, etc.; storage unit 18, such as disk, optical disk, etc.; and communication unit 19, such as network card, modem, wireless transceiver, etc. Communication unit 19 allows electronic device 10 to exchange information / data with other devices through computer networks such as the Internet and / or various telecommunications networks.

[0183] Processor 11 can be a variety of general-purpose and / or special-purpose processing components with processing and computing capabilities. Some examples of processor 11 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various special-purpose artificial intelligence (AI) computing chips, various processors running machine learning model algorithms, a digital signal processor (DSP), and any suitable processor, controller, microcontroller, etc. Processor 11 performs the various methods and processes described above, such as the performance prediction optimization method for a semiconductor device testing system sorting machine, i.e.:

[0184] A real-time data stream is acquired from the sorting machine, and the action duration sequence of each of the multiple actuators in the sorting machine used to perform the picking and placing of semiconductor devices is determined based on the real-time data stream; wherein, the action duration sequence includes the action duration of multiple picking and placing action cycles; the action duration of each actuator is compared with the corresponding preset benchmark duration average to determine the action duration deviation sequence of each actuator; based on the action duration deviation sequence of each actuator, a trained performance prediction model is used to determine the predicted change in the sorting time of the sorting machine.

[0185] In some embodiments, the performance prediction and optimization method for the semiconductor device test system sorter can be implemented as a computer program tangibly contained in a computer-readable storage medium, such as storage unit 18. In some embodiments, part or all of the computer program can be loaded and / or installed on the electronic device 10 via ROM 12 and / or communication unit 19. When the computer program is loaded into RAM 13 and executed by processor 11, one or more steps of the performance prediction and optimization method for the semiconductor device test system sorter described above can be performed. Alternatively, in other embodiments, processor 11 can be configured to execute the performance prediction and optimization method for the semiconductor device test system sorter by any other suitable means (e.g., by means of firmware).

[0186] Various embodiments of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field-programmable gate arrays (FPGAs), application-specific integrated circuits (ASICs), application-specific standard products (ASSPs), systems-on-a-chip (SoCs), payload-programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various embodiments may include implementations in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which may be a dedicated or general-purpose programmable processor, capable of receiving data and instructions from a storage system, at least one input device, and at least one output device, and transmitting data and instructions to the storage system, the at least one input device, and the at least one output device.

[0187] Computer programs used to implement the methods of the present invention may be written in any combination of one or more programming languages. These computer programs may be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing device, such that when executed by the processor, the computer programs cause the functions / operations specified in the flowcharts and / or block diagrams to be performed. The computer programs may be executed entirely on a machine, partially on a machine, or as a standalone software package, partially on a machine and partially on a remote machine, or entirely on a remote machine or server.

[0188] In the context of this invention, a computer-readable storage medium can be a tangible medium that may contain or store a computer program for use by or in conjunction with an instruction execution system, apparatus, or device. A computer-readable storage medium may include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination thereof. Alternatively, a computer-readable storage medium may be a machine-readable signal medium. More specific examples of machine-readable storage media include electrical connections based on one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fibers, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof.

[0189] To provide interaction with a user, the systems and techniques described herein can be implemented on an electronic device having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user; and a keyboard and pointing device (e.g., a mouse or trackball) through which the user provides input to the electronic device. Other types of devices can also be used to provide interaction with the user; for example, feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including sound input, voice input, or tactile input).

[0190] The systems and technologies described herein can be implemented in computing systems that include backend components (e.g., as data servers), or middleware components (e.g., application servers), or frontend components (e.g., user computers with graphical user interfaces or web browsers through which users can interact with implementations of the systems and technologies described herein), or any combination of such backend, middleware, or frontend components. The components of the system can be interconnected via digital data communication of any form or medium (e.g., communication networks). Examples of communication networks include local area networks (LANs), wide area networks (WANs), blockchain networks, and the Internet.

[0191] A computing system can include clients and servers. Clients and servers are generally located far apart and typically interact through communication networks. The client-server relationship is created by computer programs running on the respective computers and having a client-server relationship with each other. The server can be a cloud server, also known as a cloud computing server or cloud host, which is a hosting product within the cloud computing service system to address the shortcomings of traditional physical hosts and VPS services, such as high management difficulty and weak business scalability.

[0192] It should be understood that the various forms of processes shown above can be used, with steps reordered, added, or deleted. For example, the steps described in this invention can be executed in parallel, sequentially, or in different orders, as long as the desired result of the technical solution of this invention can be achieved, and this is not limited herein.

[0193] The specific embodiments described above do not constitute a limitation on the scope of protection of this invention. Those skilled in the art should understand that various modifications, combinations, sub-combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this invention should be included within the scope of protection of this invention.

Claims

1. A method for performance prediction and optimization of a semiconductor device testing system sorter, characterized in that, include: A real-time data stream is acquired from the sorting machine, and the action duration sequence of each of the multiple actuators in the sorting machine used to perform the pick-and-place action of semiconductor devices is determined based on the real-time data stream; wherein, the action duration sequence includes the action duration of multiple pick-and-place action cycles; The action duration of each actuator is compared with the corresponding preset baseline average duration to determine the action duration deviation sequence of each actuator. Based on the action duration deviation sequence of each actuator, the system is processed using a trained performance prediction model to determine the predicted changes in the sorting time of the sorting machine.

2. The method according to claim 1, characterized in that, The sorting machine includes multiple actuators: an X-axis actuator and a Y-axis actuator for horizontal movement, a Z-axis actuator for vertical movement, and a rotary actuator for rotation about the Z-axis.

3. The method according to claim 2, characterized in that, The duration of each actuator's action includes the duration from pickup trigger to placement completion.

4. The method according to claim 3, characterized in that, The action duration sequence of each of the multiple actuators in the sorting machine used to perform semiconductor device pick-and-place actions is determined based on the real-time data stream, including: The timestamped action events are obtained from the real-time data streams of each of the actuators; wherein the action events include pick-up trigger events and placement completion events; Based on the action events and timestamps, the action duration sequence of each of the multiple actuators in the sorting machine used to perform semiconductor device pick-and-place actions is determined.

5. The method according to claim 4, characterized in that, After determining the action duration sequence of each of the multiple actuators in the sorting machine used to perform semiconductor device pick-and-place actions based on the real-time data stream, the method further includes: Acquire the sensing data of the optical encoders of each actuator, and determine the position and corresponding time of the actuator based on the sensing data; Calculate the action duration of each actuator based on its position and corresponding time. The action duration sequence is corrected and supplemented based on the calculated action duration.

6. The method according to claim 4, characterized in that, Based on the action duration deviation sequence of each actuator, a trained performance prediction model is used to determine the predicted change in the indexing time of the sorting machine, including: Based on the action duration deviation sequence of each actuator and the sensor detection sequence, the system is processed using a trained performance prediction model to determine the predicted change in the indexing time of the sorting machine; wherein, the sensor detection sequence includes the nozzle action duration sequence and the drive current characteristic value of the actuator.

7. The method according to claim 1, characterized in that, Before comparing the action duration of each actuator with the corresponding preset baseline average duration, the method further includes: Based on the metadata of the sorting machine, the preset benchmark data of the sorting machine that has been stored is retrieved from the preset benchmark library. The metadata includes the model, identifier, and operating environment parameters of the sorting machine; the preset benchmark data includes the average benchmark duration and statistical tolerance range of each actuator's motion phase, the optimal control parameters of each actuator, and the environmental parameters recorded during benchmark acquisition; the preset benchmark data is the benchmark data determined when the sorting machine is configured for new product introduction.

8. The method according to claim 1, characterized in that, The performance prediction model is a machine learning model capable of processing time-series data.

9. The method according to claim 1, characterized in that, The predicted changes in the sorting time of the sorting machine include: abnormal events, sorting time increments, predicted cycle counts, and confidence levels.

10. The method according to claim 9, characterized in that, The predicted changes in the sorting time of the sorting machine also include: the cause of the abnormality, and / or, the optimization parameters and the adjustment amount of the optimization parameters.

11. The method according to claim 10, characterized in that, The abnormal events include at least one of the following: no event, indexing time increase event, actuator timeout event, placement accuracy failure event, pickup failure event, sequence out-of-step event, and motor overcurrent warning event; The causes of the anomaly include at least one of the following: mechanical wear, belt tension drift, bearing wear, pneumatic system performance degradation, vacuum leakage, filter blockage, pressure regulator drift, servo parameter tuning drift, environmental parameter drift, and timing changes caused by firmware. The optimized parameters include at least one of the following: X-axis acceleration, X-axis deceleration, Y-axis acceleration, Z-axis lifting vacuum setting value, Z-axis lowering vacuum release timing value, X-axis movement delay timing value, Y-axis start-up waiting X-axis completion timing value, and X-axis servo proportional gain.

12. The method according to claim 10, characterized in that, After determining the predicted variation in the indexing time of the sorting machine, the method further includes: The optimized parameters and their adjustment values ​​are transmitted to the sorting machine to control it to adjust the actuator parameters; or The optimization parameters and their adjustment amounts are provided to the user, and in response to the user's selection confirmation, the sorting machine is controlled to adjust the actuator parameters.

13. The method according to claim 12, characterized in that, After controlling the sorting machine to adjust the actuator parameters, the process also includes: Monitor the indexing time of the sorting machine; If the detected indexing time decay exceeds the set threshold or a fault condition is detected, the rollback procedure will be automatically started to restore the parameter configuration to its original state.

14. The method according to claim 1, characterized in that, After determining the predicted variation in the indexing time of the sorting machine, the method further includes: Based on the predicted changes in the sorting time of the sorting machine and the preset domain mapping rule algorithm, the analysis results of the sorting machine are determined; wherein, the analysis results include at least one of the following: description of the cause of the anomaly, throughput impact risk, collision risk, historical performance analysis, processing program status analysis, and estimated throughput change; The predicted changes and analysis results are displayed in the user's graphical interface.

15. A performance prediction and optimization device for a semiconductor device testing system sorting machine, characterized in that, include: A sequence determination module is used to acquire a real-time data stream from the sorting machine and determine the action duration sequence of each of the multiple actuators in the sorting machine that perform semiconductor device pick-and-place actions based on the real-time data stream; wherein, the action duration sequence includes the action duration of multiple pick-and-place action cycles; The benchmark comparison module is used to compare the action duration of each actuator with the corresponding preset benchmark duration average value to determine the action duration deviation sequence of each actuator. The prediction change module is used to process the action duration deviation sequence of each actuator based on a trained performance prediction model to determine the predicted change in the sorting time of the sorting machine.

16. An electronic device, characterized in that, The electronic device includes: At least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores a computer program executable by the at least one processor, the computer program being executed by the at least one processor to enable the at least one processor to perform the performance prediction and optimization method for the sorting machine of the semiconductor device test system according to any one of claims 1-14.

17. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer instructions that are used to cause a processor to execute the performance prediction and optimization method for the sorting machine of the semiconductor device testing system according to any one of claims 1-14.

18. A computer program product, characterized in that, The computer program product includes a computer program that, when executed by a processor, implements the performance prediction and optimization method for the sorting machine of the semiconductor device testing system according to any one of claims 1-14.

Images