PCB automatic function test method and system based on parallel test technology

Abstract

The application discloses a PCB automatic function test method and system based on parallel test technology, and relates to the technical field of electrical performance test, which comprises the following steps: according to a multi-dimensional twin model, the static pre-estimation time length of each test item in a to-be-tested PCB is matched, the static pre-estimation time length is adaptively compensated and calibrated to obtain the pre-estimation test time length of a to-be-tested task; based on the pre-estimation test time length, the to-be-allocated queue of the to-be-tested task is obtained, the to-be-allocated queue is subjected to station allocation to obtain the station task queue of a test station; a dynamic backfill mechanism is introduced, short tasks in the to-be-allocated queue are tested in parallel according to the real-time resource list of the test station during long task testing; and according to the pre-estimation deviation rate of the pre-estimation test time length of unexecuted tasks before and after updating, the station task queue of the unexecuted tasks in the test station is updated. The application is used to solve the problems in the prior art that the function prediction time length cannot be accurately estimated, the station resource utilization rate is low, and the scheduling scheme of test tasks lacks dynamic correction capability.

Description

Technical Field

[0001] This invention relates to the field of electrical performance testing technology, and more specifically, to a method and system for automated functional testing of PCBs based on parallel testing technology. Background Technology

[0002] As the electronics manufacturing industry rapidly develops towards high density, multiple varieties, and large batches, functional testing of PCB circuit boards with completed component assembly has become a core performance verification step before products leave the factory. Multi-station parallel testing technology improves throughput per unit time by simultaneously testing multiple PCBs and has become the mainstream technology route in the industry. However, with the increase in the number of workstations and the acceleration of product iteration, problems such as resource waste, cycle time loss, and prediction distortion under the traditional static scheduling architecture are becoming increasingly prominent.

[0003] For example, Chinese invention patent application document CN119716477A discloses a multi-task parallel testing method for a multi-station FCT testing system. It achieves the function of simultaneous testing at multiple stations by loading multiple PCBs (PCBAs) with completed component assembly to each station of the bed of needles, scanning to obtain production work orders, burning test programs, executing tests and acquiring signals, performing qualification certification, acquiring load working signals and judging whether they are qualified, uploading to MES and indicating qualified stations.

[0004] However, the above method has the following shortcomings: (1) The functional test duration was not pre-assessed, and the impact of batch differences, environmental temperature and humidity fluctuations, and instrument performance aging on the test duration was not considered, which resulted in the inability to provide accurate duration basis for the scheduling decision of test tasks; (2) The test tasks were statically and independently bound to the workstation, resulting in a large number of instrument resources in a single workstation being idle in non-corresponding test phases, low workstation resource utilization and parallel test efficiency, and the test tasks in the test workstation were fixed after allocation, and could not be adaptively adjusted according to the test duration deviation in the actual test process, which resulted in the scheduling effect of test tasks gradually deteriorating with changes in operating conditions.

[0005] To address the aforementioned shortcomings, there is an urgent need for a functional testing method and system that can accurately predict test duration, improve the utilization rate of workstation instruments, and have the ability to dynamically correct test tasks, so as to perform parallel functional testing on PCB circuit boards that have completed component assembly and need to undergo functional verification. Summary of the Invention

[0006] To overcome the aforementioned deficiencies of the prior art, embodiments of the present invention provide a PCB automated functional testing method and system based on parallel testing technology, in order to solve the problems in the prior art, such as the inability to accurately estimate the functional prediction time, low utilization of workstation resources, and the lack of dynamic correction capability in the scheduling scheme of test tasks, which lead to the inability of PCB functional testing to achieve continuous optimal parallel testing.

[0007] To achieve the above objectives, this application provides the following technical solution: In a first aspect, the present invention provides an automated functional testing method for PCBs based on parallel testing technology, comprising: constructing a multidimensional twin model based on multidimensional PCB acquisition data; matching the static estimated duration of each test item in the PCB under test according to the multidimensional twin model; and adaptively compensating and calibrating the static estimated duration of each test item to obtain the estimated test duration of the corresponding test task for the PCB under test; dividing the test task into long tasks and short tasks based on the estimated test duration; sorting the divided test tasks using a hybrid strategy to obtain a queue to be assigned; and introducing a short task priority window and workstation attenuation degree for the test task. The tasks to be tested in the allocation queue are assigned to workstations to obtain the workstation task queues for the test workstations; the tasks of the test workstations are executed based on the workstation task queues, and a resource-aware dynamic backfilling mechanism is introduced to obtain the real-time resource list of the test workstations during long task testing, and the short tasks in the allocation queues are tested in parallel based on the real-time resource list; the multi-dimensional twin model and the estimated test time of the unexecuted tasks are updated based on the actual test time of the completed test tasks, and the workstation task queue of the unexecuted tasks in the test workstations is updated based on the prediction deviation rate of the corresponding estimated test time of the unexecuted tasks before and after the update.

[0008] In a preferred embodiment, constructing a multidimensional twin model based on PCB multidimensional acquisition data includes: the PCB multidimensional acquisition data including test item duration data, test instrument data, and environmental sensor data; constructing a PCB twin layer with a three-layer index structure based on the test item duration data, wherein the three-layer index structure includes the model and batch of the tested PCB, as well as the test item duration and average test duration of each test item; constructing a workstation twin layer with test station ID and workstation instrument information based on the test instrument data; constructing an environmental twin layer with a duration compensation function based on the environmental sensor data, wherein the duration compensation function is generated based on environmental baseline values; and constructing a multidimensional twin model based on the PCB twin layer, workstation twin layer, and environmental twin layer.

[0009] In a preferred embodiment, matching the static estimated duration of each test item in the PCB under test based on the multidimensional twin model includes: obtaining the test parameters of the PCB under test, the test parameters including model, batch and test items, the PCB under test containing at least one test item; based on the PCB twin layer in the multidimensional twin model, performing a matching query on the model and batch of the PCB under test to obtain the average test duration corresponding to each test item in the PCB under test, and using the average test duration as the static estimated duration of each test item.

[0010] In a preferred embodiment, adaptive compensation calibration is performed on the static estimated duration of each test item to obtain the estimated test duration of the test task corresponding to the PCB under test. This includes: calculating the batch compensation amount for each batch based on the test item duration and static estimated duration of each batch of test items in the PCB twin layer, and obtaining the batch compensation amount for each test item in the PCB under test according to the batch of the PCB under test; obtaining the test environment parameters of the PCB under test, and calculating the environmental compensation amount corresponding to the test environment parameters based on the duration compensation function in the environmental twin layer; obtaining the associated instrument list for each test item in the PCB under test, obtaining the annual attenuation rate and service life of each instrument in the associated instrument list based on the workstation twin layer, and calculating the aging compensation amount for each instrument, taking the largest aging compensation amount of each instrument in the associated instrument list as the instrument compensation amount for the test item; adaptively compensating the static estimated duration of each test item based on the batch compensation amount, environmental compensation amount, and instrument compensation amount, and summing the compensated static estimated durations of each test item to obtain the estimated test duration of the test task corresponding to the PCB under test.

[0011] In a preferred embodiment, the tasks to be tested are divided into long tasks and short tasks based on the estimated test duration. A mixed-strategy sorting process is then applied to the divided tasks to obtain a queue for allocation. This includes: generating a queue of tasks to be tested; using the median of the estimated test duration for each task in the queue as a classification benchmark; and calculating a dynamic classification duration based on the benchmark and a preset classification threshold. If the estimated test duration of a task is less than the dynamic classification duration, the task is classified as a short task. If the estimated test duration of a task is greater than or equal to the dynamic classification duration, the task is classified as a long task. Short tasks are arranged in ascending order of estimated test duration to obtain a short task sequence. Long tasks are arranged in descending order of estimated test duration to obtain a long task sequence. The short task sequence is then placed before the long task sequence to generate the queue for allocation.

[0012] In a preferred embodiment, a short task priority window and a workstation attenuation degree are introduced, including: setting a workstation status variable for the test workstation, the workstation status variable including the expected completion time and the list of assigned tasks; generating a short task priority window based on the number of test workstations and a preset window definition constant, and defining the priority allocation range of the tasks to be tested according to the short task priority window; obtaining the annual attenuation rate and a preset instrument weight value of each test instrument in the test workstation, and calculating the workstation attenuation degree according to the annual attenuation rate and the instrument weight value.

[0013] In a preferred embodiment, the process of assigning test tasks in the allocation queue to obtain a test task queue includes: obtaining the queue position number of the test task in the allocation queue; obtaining test tasks of short task type and whose queue position number is within the short task priority window, and assigning the test task to the test station with the shortest expected completion time; obtaining test tasks of long task type and whose queue position number is outside the short task priority window, traversing and calculating the simulated test time of the test task at each test station, and assigning the test task to the test station with the minimum simulated test time and the minimum station decay rate; generating a test task queue based on the test tasks assigned in the test station, and updating the station status variable of the test station based on the test task queue.

[0014] In a preferred embodiment, short tasks in the allocation queue are tested in parallel based on a real-time resource list, including: the real-time resource list of the test station contains a list of idle resources and a list of occupied resources for the test station; when the test station starts executing a long task, a dynamic backfilling mechanism is triggered, and the list of idle resources of the test station is obtained, and dynamic backfilling conditions for the test station are generated based on the list of idle resources; the task parameters of each short task in the allocation queue are obtained, and short tasks whose task parameters meet the dynamic backfilling conditions are backfilled to the test station, and the short tasks are executed in parallel while the long task is being executed at the test station.

[0015] In a preferred embodiment, the method based on the estimated deviation rate of the estimated test duration corresponding to the unexecuted tasks before and after the update includes: obtaining the actual test duration of each test item in the completed test tasks, and updating the average test duration of the corresponding test item in the PCB twin layer based on the actual test duration; obtaining all unexecuted tasks in the workstation task queue, obtaining the estimated test duration of the unexecuted tasks after the update based on the updated PCB twin layer, and obtaining the estimated deviation rate of the unexecuted tasks based on the estimated test duration of the unexecuted tasks before and after the update; comparing the estimated deviation rate of all unexecuted tasks with a preset deviation threshold, and when the estimated deviation rate of an unexecuted task is greater than the preset deviation threshold, it is counted as an update trigger event, and the trigger timestamp of the update trigger event is recorded; when there are two consecutive update trigger events, and the interval between the trigger times of the update trigger events is greater than the minimum test duration of the test item, the workstation task queue of the unexecuted task in the test workstation is updated, wherein the minimum test duration of the test item is obtained based on the updated PCB twin layer.

[0016] A second aspect of the present invention provides an automated functional testing system for PCBs based on parallel testing technology, comprising: a duration estimation module, used to construct a multidimensional twin model based on multidimensional PCB acquisition data, match the static estimated duration of each test item in the PCB under test according to the multidimensional twin model, and perform adaptive compensation calibration on the static estimated duration of each test item to obtain the estimated test duration of the corresponding test task for the PCB under test; and a workstation allocation module, used to divide the test task into long tasks and short tasks based on the estimated test duration, sort the divided test tasks using a hybrid strategy to obtain a queue to be allocated, and introduce a short task priority window and workstation attenuation degree to allocate the tasks. The task queue is assigned to test stations based on the task queues. The parallel testing module executes task tests at the test stations based on the task queues and introduces a resource-aware dynamic backfilling mechanism to obtain a real-time resource list for long task testing at the test station. Short tasks in the assigned queue are tested in parallel based on the real-time resource list. The task queue update module updates the multidimensional twin model and the estimated test time of unexecuted tasks based on the actual test time of completed test tasks. Based on the estimated deviation rate of the estimated test time of unexecuted tasks before and after the update, the task queue of unexecuted tasks at the test station is updated.

[0017] The beneficial effects of this invention are:

[0018] 1. This invention constructs a multidimensional twin model based on multidimensional PCB acquisition data, matches the static estimated duration of each test item in the PCB under test according to the multidimensional twin model, and performs adaptive compensation calibration on the static estimated duration of each test item to obtain the estimated test duration of the corresponding test task for the PCB under test. Through the multidimensional twin model, the functional test duration of the PCB under test is pre-evaluated in multiple dimensions, providing an accurate duration basis for subsequent scheduling decisions of the test task.

[0019] 2. This invention introduces a short task priority window and a workstation decay rate to allocate test tasks in the task allocation queue to obtain the workstation task queue for the test workstation. It also introduces a resource-aware dynamic backfilling mechanism to perform parallel testing of short tasks in the task allocation queue based on the real-time resource list of the test workstation during long task testing. This improves the utilization rate of instrument resources at the test workstation and effectively solves the problems of low workstation resource utilization and low parallel testing efficiency in existing technologies. Furthermore, this invention updates the workstation task queue for unexecuted tasks at the test workstation based on the estimated deviation rate of the estimated test duration corresponding to the unexecuted tasks before and after the update. This enables adaptive adjustment of the workstation task queue for each test workstation based on the actual test duration deviation during the testing process, ensuring the testing efficiency of each test workstation. Attached Figure Description

[0020] Figure 1 This is a flowchart illustrating an automated functional testing method for PCBs based on parallel testing technology. Figure 2 This is a schematic diagram of the method for constructing a multidimensional twin model based on multidimensional PCB acquisition data provided in an embodiment of the present invention; Figure 3 This is a flowchart illustrating a method for adaptively compensating and calibrating the static estimated duration of each test item to obtain the estimated test duration of the corresponding test task for the PCB under test, as provided in an embodiment of the present invention. Figure 4 This is a schematic flowchart of a method for introducing a short task priority window and workstation attenuation degree according to an embodiment of the present invention; Figure 5 This is a flowchart illustrating a method for allocating test tasks in a queue to obtain a test task queue, according to an embodiment of the present invention. Figure 6 This is a flowchart illustrating a method for estimating the deviation rate of the estimated test duration based on the corresponding estimated test duration before and after the update, as provided in an embodiment of the present invention. Figure 7 This is a schematic diagram of a PCB automated functional testing system based on parallel testing technology. Detailed Implementation

[0021] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention. Example 1:

[0022] Please see Figure 1 , Figure 1 This invention presents an automated functional testing method for PCBs based on parallel testing technology, comprising the following steps: constructing a multidimensional twin model based on multidimensional PCB acquisition data; matching the static estimated duration of each test item in the PCB under test according to the multidimensional twin model; adaptively compensating and calibrating the static estimated duration of each test item to obtain the estimated test duration of the corresponding test task for the PCB under test; dividing the test task into long tasks and short tasks based on the estimated test duration; sorting the divided test tasks using a hybrid strategy to obtain a queue to be assigned; and introducing a short task priority window and workstation attenuation degree to the queue to be assigned. The tasks to be tested in the queue are assigned to workstations to obtain the workstation task queues for the test workstations; the tasks of the test workstations are tested based on the workstation task queues, and a resource-aware dynamic backfilling mechanism is introduced to obtain the real-time resource list of the test workstations during long task testing. Short tasks in the queue to be assigned are tested in parallel based on the real-time resource list; the multidimensional twin model and the estimated test time of unexecuted tasks are updated based on the actual test time of completed test tasks, and the workstation task queue of unexecuted tasks in the test workstations is updated based on the prediction deviation rate of the corresponding estimated test time of unexecuted tasks before and after the update.

[0023] This invention constructs a multidimensional twin model based on multidimensional PCB acquisition data. It matches the static estimated duration of each test item in the PCB under test (TBT) according to the multidimensional twin model and performs adaptive compensation calibration on the static estimated duration of each test item to obtain the estimated test duration of the corresponding TBT task for the PCB under test. The multidimensional twin model enables multidimensional pre-evaluation of the functional test duration of the PCB under test, providing accurate duration basis for subsequent scheduling decisions of the TBT task. Furthermore, based on the estimated test duration, the TBT task is divided into long tasks and short tasks. A hybrid strategy is used to sort the divided TBT tasks to obtain a queue to be allocated. A short task priority window and workstation attenuation are introduced. Workstation allocation is performed on the TBT tasks in the allocation queue to obtain the workstation task queue for the test workstations, thus achieving… The invention accurately allocates tasks to be tested and introduces a resource-aware dynamic backfilling mechanism. Based on the real-time resource list of the test station during long task testing, it tests short tasks in the assigned queue in parallel, improving the utilization rate of instrument resources at the test station. This effectively solves the problem in the prior art where test tasks are statically and independently bound to the test station, resulting in a large number of instrument resources in a single station being idle during non-corresponding test phases, leading to low station resource utilization and parallel testing efficiency. Furthermore, the invention updates the station task queue of unexecuted tasks at the test station based on the estimated deviation rate of the estimated test duration corresponding to the unexecuted tasks before and after the update. This enables adaptive adjustment of the station task queue of each test station according to the test duration deviation during the actual testing process, ensuring the testing efficiency of each test station.

[0024] S1. Construct a multidimensional twin model based on the multidimensional acquisition data of the PCB, match the static estimated duration of each test item in the PCB under test according to the multidimensional twin model, and perform adaptive compensation calibration on the static estimated duration of each test item to obtain the estimated test duration of the corresponding test task of the PCB under test.

[0025] Please see Figure 2 S11. Construct a multidimensional twin model based on PCB multidimensional acquisition data, including:

[0026] S111, PCB multidimensional acquisition data includes test item duration data, test instrument data, and environmental sensor data;

[0027] The test duration data is obtained by embedding timing probes within the test program of each test station. Specifically, timestamps are recorded at the beginning and end of each test, with time accuracy controlled within 10 milliseconds, forming structured data records. The record fields must include at least: station ID, product model, batch number, test item name, start timestamp, end timestamp, and test duration.

[0028] The test instrument data is sent to the test instruments configured at each test station at preset intervals via a general interface bus or local area network interface to read and record the key health index parameters of the test instruments. These key health index parameters include, but are not limited to, the power-on slope value and output ripple value of the programmable power supply, the sampling rate deviation value and noise floor level value of the oscilloscope / digitizer, and the frequency stability value and output amplitude deviation value of the signal source. Each record includes: station ID, instrument ID, parameter name, current measured value, factory nominal value, and deviation rate.

[0029] Environmental sensing data is collected by deploying temperature and humidity sensors above or to the side of each test station. The temperature and humidity values ​​of the test station are continuously collected at a sampling frequency of not less than 1 Hz, forming an environmental data sequence with timestamps.

[0030] It should be noted that in this application, PCB refers to a printed circuit board assembly with components already assembled.

[0031] S112. Construct a PCB twin layer with a three-layer index structure based on the test item duration data. The three-layer index structure includes the model and batch of the tested PCB, as well as the test item duration and average test duration of each test item. The PCB twin layer uses a three-layer index structure of "product model - production batch - test item". The product model, production batch and test item correspond to each other. That is, when matching and querying the PCB under test based on the PCB twin layer, the product model, production batch and each test item of the PCB under test are matched and queried in sequence. In addition, a set of duration statistics is maintained independently for each test item in the PCB twin layer. Based on the duration statistics, the test item duration and the average test duration can be obtained.

[0032] S113. Construct a workstation twin layer containing test station ID and workstation instrument information based on test instrument data; The workstation ID serves as the primary index for the workstation twin layer, recording all instrument information configured at that test workstation. The information for each instrument includes at least the instrument ID, model and specifications, the date of the most recent calibration, a set of health indicator data, the cumulative years of use, and the annual degradation rate obtained by linear fitting based on historical health indicator data. The workstation twin layer can record and continuously track the instrument information and performance degradation trend of each instrument in each test workstation.

[0033] S114. Construct an environmental twin layer containing a duration compensation function based on environmental sensing data. The duration compensation function is generated based on the environmental baseline value. The environment twin layer is used to store the duration compensation function for each test item under different environmental conditions. The calculation formula for the duration compensation function is as follows: ; In the formula, This is the temperature difference between the real-time ambient temperature and the temperature reference value in the environmental reference value, which can be set to 25℃. This is the humidity difference between the real-time ambient relative humidity and the relative humidity reference value in the environmental reference value, which can be set to 50%RH. and To compensate for the parameters, pre-calibration experiments were conducted, which involved repeatedly performing the same test item under different temperature and humidity combinations, recording the duration changes, and then fitting the results using linear regression. Different compensation parameters were configured for different types of test items. For example, since digital signal test items are not sensitive to temperature and humidity, their compensation parameters are... and It can be set to zero.

[0034] S115. Construct a multi-dimensional twin model based on the PCB twin layer, workstation twin layer, and environment twin layer.

[0035] Based on the multi-dimensional twin model constructed from the PCB twin layer, the workstation twin layer, and the environment twin layer, it can perform static prediction and adaptive dynamic compensation calibration of the test duration of each test item in each PCB under test from three test factors: the PCB under test, the test workstation, and the test environment. This provides a real-time corrected and accurate predicted test duration for the allocation and scheduling of the corresponding test tasks for the PCB under test, thereby improving the accuracy of the test task allocation results.

[0036] S12. Based on the multi-dimensional twin model, match the static estimated duration of each test item in the PCB under test, including: S121. Obtain the test parameters of the PCB under test. The test parameters include the model, batch and test items. The PCB under test contains at least one test item. The model and batch number specifically refer to the product model and batch number of the PCB under test. After obtaining the product model and batch number of the PCB under test, test items of the same model and batch as the PCB under test can be matched through the PCB twin layer. Based on the average test duration of the test item, the static estimated duration of each test item in the PCB under test can be accurately estimated.

[0037] It should be noted that each PCB under test contains at least one test item. A test item refers to an independent and indivisible single test step or test category when performing functional testing on a PCB. Each test item completes the detection of a specific functional module or a specific electrical indicator on the PCB, such as power-on detection to verify whether the power supply circuit is working properly, and power consumption detection to verify the current consumption under working conditions.

[0038] S122. Based on the PCB twin layer in the multidimensional twin model, a matching query is performed on the model and batch of the PCB under test to obtain the average test duration corresponding to each test item in the PCB under test, and the average test duration is used as the static estimated duration of each test item, specifically: (1) If the model and corresponding batch of the PCB to be tested are already recorded in the PCB twin layer, the average test duration of each test item under that batch is extracted, and the test duration of each test item is used as the static estimated duration of the corresponding test item in the PCB to be tested.

[0039] (2) If the model of the PCB under test exists but the corresponding batch does not exist, the average test duration of each test item in the most recent batch under the model is extracted, and the average test duration of each test item in the batch is used as the static estimated duration of the corresponding test item in the PCB under test. At the same time, the batch of the PCB under test is recorded and the batch is marked as "to be updated" in the PCB twin layer. After each test item of the PCB under test is completed, the test item data record of the corresponding batch of the PCB under test is automatically established in the PCB twin layer.

[0040] (3) If the model of the PCB under test does not exist in the PCB twin layer, the static estimated duration of the PCB under test is generated by the test classification matching strategy. Specifically, all test items are divided into five categories according to physical attributes: power supply, digital IO, analog acquisition, communication interface and radio frequency. Each category has a corresponding baseline duration database. The list of all test items of the PCB under test is obtained. The corresponding category of each test item in the list is matched item by item according to the five categories. The average baseline duration of the test items of the same category in the baseline duration database is obtained. The average baseline duration of each test item is used as the static estimated duration of each test item. After obtaining the static estimated duration, the model of the PCB under test is marked as "to be created" in the PCB twin layer. After the test items of the PCB under test are completed, a complete three-level index structure containing "product model - production batch - test item" is created in the PCB twin layer for the model corresponding to the PCB under test.

[0041] It should be noted that in the process of classifying all test items into five categories based on physical attributes—power supply, digital I / O, analog acquisition, communication interface, and radio frequency—physical attributes refer to the attributes of the core physical quantity being measured. Specifically: test items whose core physical quantity is power level voltage / current and which focus on power transmission and load characteristics are classified as power supply; test items whose core physical quantity is discrete binary level with only high and low logic states are classified as digital I / O; test items whose core physical quantity is continuously changing analog voltage / current and which focus on amplitude, linearity, and accuracy are classified as analog acquisition; test items whose core physical quantity is serial digital data stream with protocols and which focus on connectivity, bit error rate, and interaction flow are classified as communication interface; and test items whose core physical quantity is high-frequency electromagnetic wave signal and which focus on frequency, power, spectrum, and impedance matching characteristics are classified as radio frequency.

[0042] Please see Figure 3 S13. Adaptive compensation calibration is performed on the static estimated duration of each test item to obtain the estimated test duration of the corresponding test task for the PCB under test, including: S131. Based on the test duration and static estimated duration of each batch of test items in the PCB twin layer, calculate the batch compensation amount corresponding to each batch, and obtain the batch compensation amount of the PCB under test according to the batch of the PCB under test. Based on the test duration and static estimated duration of each batch in the PCB twin layer, the batch compensation amount corresponding to each batch is calculated. Specifically, after the first PCB of a batch completes all test items, the ratio of the actual total test duration of the PCB to the static estimated total duration is used as the batch compensation amount of the PCB in that batch. The batch in the PCB twin layer that matches the batch of the PCB under test is obtained, and the batch compensation amount of the batch is used as the batch compensation amount of each test item in the PCB under test.

[0043] S132. Obtain the test environment parameters of the PCB under test, and calculate the environmental compensation amount corresponding to the test environment parameters based on the duration compensation function in the environment twin layer. For each test item of the PCB under test, the environmental compensation amount for the test item is obtained by combining the real-time ambient temperature and relative humidity collected at the test station at the current moment with the duration compensation function in the environmental twin layer and the compensation constant corresponding to the test item. For example, if the digital test item is not sensitive to changes in temperature and humidity, its corresponding compensation constant is 0. Then, the environmental compensation amount for the digital test item is 0 after calculation by the duration compensation function.

[0044] S133. Obtain the list of associated instruments for the PCB under test. Based on the workstation twin layer, obtain the annual attenuation rate and service life of each instrument in the list of associated instruments, and calculate the aging compensation amount of each instrument. Take the maximum aging compensation amount of each instrument in the list of associated instruments as the instrument compensation amount corresponding to the PCB under test. Based on the instruments required for each test item in the PCB under test, a list of associated instruments for the PCB under test is generated. According to this list, the annual attenuation rate and cumulative service life of each instrument are obtained from the workstation twin layer. The aging compensation amount corresponding to each instrument in the list is calculated one by one. If a test item involves multiple instruments, the maximum value of the aging compensation amount of each instrument is taken as the instrument compensation amount for that test item. The calculation formula for the aging compensation amount is as follows: ; In the formula, L is the aging compensation amount of the instrument; The annual decay rate of the instrument refers to the percentage of time that the instrument's performance indicators deteriorate each year. It is obtained by linear fitting of historical health indicator data and stored in the workstation twin layer. The cumulative service life of the instrument refers to the number of years from when the instrument was put into use to the present time, and is stored in the workstation twin layer.

[0045] S134. Based on batch compensation, environmental compensation and instrument compensation, adaptive compensation is performed on the static estimated time of each test item, and the static estimated time of each test item after compensation is summed to obtain the estimated test time of the test task corresponding to the PCB under test. Based on the batch compensation, environmental compensation, and instrument compensation of each test item in the PCB under test, adaptive compensation is performed on the static estimated duration of each test item. This yields the post-compensation estimated duration of each test item in the PCB under test. The post-compensation estimated durations of each test item in the PCB under test are then summed to obtain the estimated test duration for the corresponding test task. The calculation formula for the post-compensation estimated duration of each test item in the PCB under test after adaptive compensation is as follows: ; In the formula, The estimated duration after compensation for the test item; The static estimated duration for the test item; These are the batch compensation amount, environmental compensation amount, and instrument compensation amount, respectively. The estimated test time of each test item in the PCB under test after adaptive compensation is summed to obtain the estimated test time of the corresponding test task for the PCB under test.

[0046] It should be noted that the test task of the PCB under test is generated by combining all the test items of the PCB under test. That is, the test task includes all the test items of the PCB under test. In addition, when combining all the test items of the PCB under test, the weight value of each test item can be preset based on the degree of influence of each test item on the overall performance of the PCB under test. Based on the weight value, all the test items of the PCB under test are sorted and combined in descending order of weight value to obtain the test task of the PCB under test.

[0047] S2. Based on the estimated test duration, the tasks to be tested are divided into long tasks and short tasks. The divided tasks are sorted using a mixed strategy to obtain the queue to be assigned. Short task priority window and workstation decay are introduced. The tasks to be tested in the queue to be assigned are assigned to workstations to obtain the workstation task queue of the test workstation. S21. Based on the estimated test duration, divide the tasks to be tested into long tasks and short tasks. Then, sort the divided tasks using a mixed strategy to obtain a queue for assignment, including: S211. Generate a queue of tasks to be tested based on the tasks to be tested, take the median of the estimated test duration of each task to be tested in the queue as the classification benchmark value, and calculate the dynamic classification duration based on the classification benchmark value and the preset classification threshold. When performing functional testing on PCBs, multiple PCBs are often tested simultaneously. Each PCB under test corresponds to a test task. Therefore, a test task queue is generated based on the test tasks corresponding to the multiple PCBs under test. The test task queue contains the test tasks of all PCBs under test, and the test tasks in the test task queue are sorted in ascending order according to the estimated test duration.

[0048] The formula for calculating the dynamic classification time based on the classification baseline and the preset classification threshold is as follows: ; In the formula, For dynamic classification duration; As the classification baseline value; The preset classification threshold is where The weight range is In this embodiment, the value can be 1.2.

[0049] It should be noted that the classification baseline value changes dynamically with the composition of the test tasks in the test task queue, which can adapt to the duration distribution characteristics of different test task queues. In addition, using the median instead of the mean as the classification baseline value can effectively resist the interference of extreme estimated test durations in the test task queue, making the classification results more consistent with the estimated test duration distribution of most test tasks.

[0050] S212. When the estimated test duration of the task to be tested is less than the dynamic classification duration, the task to be tested is judged as a short task; when the estimated test duration of the task to be tested is greater than or equal to the dynamic classification duration, the task to be tested is judged as a long task.

[0051] S213. Arrange the short tasks in ascending order of estimated test duration from smallest to largest to obtain the short task sequence; arrange the long tasks in descending order of estimated test duration from largest to smallest to obtain the long task sequence; place the short task sequence before the long task sequence to generate the queue to be assigned. All tasks identified as short tasks are sorted in ascending order of estimated test duration from shortest to longest, and all tasks identified as long tasks are sorted in descending order of estimated test duration from longest to shortest. Finally, the short task sequence is placed at the front of the queue and the long task sequence is placed at the back of the queue, and then they are merged to form an ordered queue to be assigned.

[0052] It should be noted that short tasks are prioritized and arranged in ascending order, which can quickly process lightweight tasks and reduce the average waiting time of the overall test; long tasks are arranged in descending order, which can prioritize the most time-consuming tasks in subsequent allocations and prevent them from being delayed to the end and prolonging the overall completion time. During the test, after the estimated test time of a new PCB under test is completed, it is dynamically inserted into the corresponding queue according to its task classification results.

[0053] Please see Figure 4 S22. Introduce short task priority window and workstation attenuation, including: S221. Set the station status variable of the test station. The station status variable includes the expected completion time and the list of assigned tasks. Each test station has a status variable created and maintained, and each test station is a station that can be tested in parallel; the expected completion time refers to the expected time for the test station to complete each test task within the station (initial value is 0); the assigned task list refers to the list of tasks to be tested assigned to the test station (initially empty).

[0054] S222. Based on the number of test stations and the preset window definition constant, generate a short task priority window, and define the priority allocation range of the test tasks according to the short task priority window. The size of the short task priority window is directly proportional to the number of test stations. This window defines the scope of the first few tasks in the station's task queue. Within this scope, a short task priority allocation strategy can quickly fill the initial test slots for each test station, reducing the waiting time for each station to perform its first round of testing. The quantitative formula for the size of the short task priority window is as follows: ; In the formula, Set the window size for the shortest task priority window; The number of test stations; Define a constant for the preset window, with a value range of 100%. In this embodiment 1.75 is a good value.

[0055] S223. Obtain the annual attenuation rate and preset instrument weight value of each test instrument in the test station, and calculate the station attenuation degree based on the annual attenuation rate and instrument weight value. The annual attenuation rate of each testing instrument in the testing station can be obtained based on the station twin layer; The instrument types of each test instrument in the test station are divided into high-precision test instruments (such as source meters, oscilloscopes, spectrum analyzers, etc.) and general auxiliary instruments (such as I / O cards, relay modules, etc.). The instrument weight value of each test instrument is set according to the instrument type. The instrument weight value of high-precision test instruments can be preset to 0.7, and the instrument weight value of high-precision test instruments can be preset to 0.3. The attenuation degree of the test station is calculated based on the annual attenuation rate and the instrument weight value. Specifically, the attenuation degree of a single test instrument is calculated based on its annual attenuation rate and cumulative service life. The attenuation degree of the entire test station is then calculated by comprehensively considering the attenuation degree of each instrument in the test station and its instrument weight value. The formula for calculating the attenuation degree of a single test instrument is as follows: ; In the formula, For the first The attenuation of the test instrument; For the first Annual attenuation rate of the test instrument; For the first The cumulative service life of the testing instruments can be obtained through the workstation twin layer.

[0056] The formula for calculating the station attenuation of the test station is as follows: ; In the formula, For the first The station attenuation degree of each test station is used to comprehensively characterize the aging degree of the instrument and the overall health status of the instrument at that test station. This refers to the attenuation of each testing instrument in this testing station; These are the instrument weight values ​​corresponding to each testing instrument.

[0057] Please see Figure 5 S23. Assign workstations to the tasks to be tested in the task assignment queue to obtain the workstation task queue for the test workstations, including: S231. Obtain the queue position number of the task to be tested in the queue to be assigned; The queue position number of each test task in the queue to be assigned is defined based on the previous and next positions of each test task in the queue to be assigned. Each test task corresponds to a queue position number. Based on the queue position number of the test task in the queue to be assigned and the length of the task, the hierarchical allocation decision of S232 and S233 is executed.

[0058] S232. If the task type is short task and the queue position number is within the short task priority window, then the task to be tested will be assigned to the test station with the shortest expected completion time. The task to be tested whose queue position number is within the shortest task priority window refers to the task to be tested whose queue position number is less than or equal to the window size value of the shortest task priority window. If the current task under test is a short task, and its queue position number is less than or equal to the window size of the short task priority window, then the expected completion time of all test workstations is scanned, and the current task under test is assigned to the test workstation with the shortest expected completion time. This strategy allows short tasks to be quickly filled into the currently idle test workstation, effectively reducing the idle time of the test workstations.

[0059] S233. Obtain the tasks to be tested that are long tasks and whose queue position number is outside the short task priority window. Iterate through and calculate the simulated test time of the tasks to be tested at each test station, and assign the tasks to be tested to the test station with the smallest simulated test time and station decay. A task to be tested whose queue position number is outside the shortest task priority window refers to a task to be tested whose queue position number is greater than the window size of the shortest task priority window. If the current test task is a long task, or the queue position number is greater than the window size of the short task priority window, then all test stations are traversed, and the virtual trial allocation calculation of the long task is performed for each test station. The test task is simulated and allocated to each test station, and the simulated test time of each test station after the simulated allocation is obtained. The test station with the smallest simulated test time and station decay rate is selected, and the current test task is allocated to that test station.

[0060] In another embodiment of this application, since the difference between the simulated test time after simulation allocation and the expected completion time before simulation allocation is within 2%, the test station can be determined to be equivalent to the optimal station in terms of production capacity, without sacrificing the overall maximum test time. Therefore, test stations with a difference of within 2% between the simulated test time after simulation allocation and the expected completion time before simulation allocation can be obtained. Based on such test stations, an equivalent optimal station set is constructed, and the current task to be tested is assigned to the test station with the smallest station decay rate in the equivalent optimal station set, thus completing the allocation of the current task to be tested. This embodiment takes minimizing the expected completion time of all stations as the first objective and the station decay rate of the test stations as the second objective. Through dual-objective hierarchical allocation, reasonable and accurate allocation of tasks to be tested that are long tasks or whose queue position number is greater than the window size value of the short task priority window can be further achieved.

[0061] S234. Generate a station task queue based on the tasks to be tested assigned in the test station, and update the station status variables of the test station based on the station task queue. Based on the allocation order of the assigned tasks in the test station, a list of assigned tasks for that test station is generated. Based on this list of assigned tasks, a task queue for the test station is obtained. Then, based on the estimated test duration of each task in the task queue, the expected completion time of the test station is updated.

[0062] S3. Execute task tests of test workstations based on workstation task queues, and introduce a resource-aware dynamic backfilling mechanism to obtain the real-time resource list of test workstations during long task testing, and test short tasks in the queue to be allocated in parallel based on the real-time resource list. S31. The real-time resource list of the test station includes a list of idle resources and a list of occupied resources for the test station. The list of idle resources refers to the list of instrument resources at the test station that are not currently occupied by any task; the list of occupied resources refers to the list of instrument resources occupied by the task currently being executed at the test station.

[0063] It should be noted that when a test task starts executing at the test station, the instrument resources in the resource requirement set are deducted from the idle resource list of the test station based on the instrument resources required by the test task, and an instrument resource record is added to the occupied resource list. When the test task is completed, the instrument resources in its resource requirement set are released from the occupied resource list to the idle resource list, thus completing the real-time update of the test station resource list.

[0064] S32. When a test station starts executing a long task, a dynamic backfilling mechanism is triggered, and the list of idle resources of the test station is obtained. Based on the list of idle resources, dynamic backfilling conditions for the test station are generated. The dynamic backfill mechanism is not triggered when the test station starts executing a short task; it is only triggered after the test station starts executing a long task. When a test station starts executing a long task, the idle resource list of that test station is periodically retrieved at preset fixed time intervals. Based on this idle resource list, a backfill condition check is performed on the test station to obtain the dynamic backfill condition corresponding to that test station. In real time, it is detected whether there are any short task-type test tasks in the waiting queue that meet the dynamic backfill condition of the test station. If there are short task-type test tasks that meet the dynamic backfill condition of the test station, the dynamic backfill of the test station is triggered, and the short task-type test task is run in parallel with the long task being executed in the test station. Both long and short tasks are executed simultaneously in the test station.

[0065] It should be noted that the dynamic backfilling conditions include, in order, duration adaptation conditions, resource compatibility conditions, and task status conditions, specifically: The duration adaptation condition refers to the existence of at least one short task in the current test queue, and the estimated test duration of the short task is less than or equal to the remaining execution time of the long task being executed at the test station.

[0066] Resource compatibility means that the resource requirements of the short task are completely contained in the current list of idle resources of the test station, that is, all the instrument resources required by the short task are not currently occupied by the long task.

[0067] The task status condition refers to the fact that the short task has not yet been assigned to any other test station.

[0068] S33. Obtain the task parameters of each short task in the queue to be assigned, backfill the short tasks whose task parameters meet the dynamic backfilling conditions to the test station, and execute the short tasks in parallel while the long task is being executed at the test station. The task parameters for a short task include the estimated test duration, resource requirements, and allocation status. If the short task is not assigned to a test station, its allocation status is "unassigned." If the short task has been assigned to any test station, its allocation status is "assigned."

[0069] If the estimated test duration, resource requirements, and allocation status of a short task meet all the dynamic backfill conditions of the test station, the short task will be backfilled to that test station and run in parallel with the long task currently being executed in the test station. That is, both the long task and the short task will be executed simultaneously in that test station.

[0070] It should be noted that when multiple short tasks simultaneously meet the dynamic backfilling conditions of the test station, the short task with the highest resource matching degree in the resource requirement set and the smallest difference between the estimated test duration and the remaining test duration of the long task should be selected first to maximize the utilization of idle time.

[0071] S4. Update the multidimensional twin model and the estimated test duration of unexecuted tasks based on the actual test duration of completed test tasks, and update the task queue of unexecuted tasks at the test station according to the estimated deviation rate of the estimated test duration of unexecuted tasks before and after the update. Please see Figure 6 S41. Based on the estimated deviation rate of the estimated test duration corresponding to the unexecuted tasks before and after the update, including: S411. Obtain the actual test duration of each test item in the completed test task, and update the average test duration of the corresponding test item in the PCB twin layer based on the actual test duration. After any test task is completed at the test station, the actual test duration of each test item in the completed test task is obtained in real time, and the average test duration of the corresponding test item in the PCB twin layer is updated based on the actual test duration.

[0072] S412. Obtain all unexecuted tasks in the workstation task queue, and obtain the estimated test duration of the unexecuted tasks after the update based on the updated PCB twin layer; Obtain the unexecuted tasks in the task queue of all test stations, and obtain the static estimated duration of each test item in the unexecuted tasks after the update based on the updated PCB twin layer. Perform adaptive compensation on the static estimated duration of each test item after the update to obtain the estimated test duration of all unexecuted tasks after the update.

[0073] S413. Based on the estimated test duration of unexecuted tasks before and after the update, calculate the estimated deviation rate of unexecuted tasks. The formula for calculating the estimated deviation rate of unexecuted tasks is as follows: ; In the formula, The estimated deviation rate for tasks not performed; This is the estimated test duration for tasks not executed after the update. This is the estimated test duration for tasks not yet executed before the update.

[0074] S42. Update the task queue for unexecuted tasks at the test workstation; All unexecuted tasks in the task queues corresponding to all test workstations are updated. Specifically, after obtaining the estimated test duration after updating the unexecuted tasks, the task queue to be assigned is updated, and all unexecuted tasks are reassigned to test workstations based on the updated task queue. This completes the update of the task queues for each test workstation. S421. Compare the estimated deviation rate of unexecuted tasks with the preset deviation threshold. If the estimated deviation rate of an unexecuted task is greater than the preset deviation threshold, it is counted as an update trigger event, and the trigger timestamp of the update trigger event is recorded. The preset deviation threshold can be set to a typical value of 15%. When the estimated deviation rate of an unexecuted task is greater than 15%, it is counted as an update trigger event, and the time of the comparison result is recorded as the trigger timestamp of the update trigger event.

[0075] S422. When there are two consecutive update trigger events, and the interval between the trigger events is greater than the minimum test duration of the test item, update the workstation task queue of the unexecuted task at the test station, wherein the minimum test duration of the test item is obtained based on the updated PCB twin layer.

[0076] The minimum test duration of a test item refers to the minimum test duration among all test items in the updated PCB twin layer. Since if the estimated deviation rate of two consecutive unexecuted tasks is greater than 15%, and the minimum test duration representing a complete test item has elapsed between the two unexecuted tasks, it indicates that there is a continuous systematic shift in temperature, humidity, or instrument aging, which is a condition that requires correction of the workstation task queue. Therefore, the minimum test duration of the test item is introduced as the trigger judgment criterion for updating the workstation task queue.

[0077] In one specific embodiment, when the estimated deviation rate of an unexecuted task is detected to be greater than 15% for the first time, an update trigger event A is recorded, along with its trigger timestamp tA. When the estimated deviation rate of an unexecuted task is detected to be greater than 15% again, an update trigger event B is recorded, along with its trigger timestamp tB. If the difference between trigger timestamp tA and trigger timestamp tB is less than or equal to the minimum test duration of the test item, the update of the task queue for each test station is not triggered. If the difference between trigger timestamp tA and trigger timestamp tB is greater than the minimum test duration of the test item, the update of the task queue for each test station is triggered. A new queue to be assigned is generated based on the estimated test duration of all unexecuted tasks after the update. The unexecuted tasks are then reassigned to test stations based on the queue to be assigned, resulting in the updated task queue for each test station.

[0078] Example 2

[0079] Please see Figure 7 The present invention provides an automated functional testing system for PCBs based on parallel testing technology, comprising: The duration estimation module constructs a multidimensional twin model based on multidimensional PCB acquisition data. It then matches the static estimated duration of each test item on the PCB under test to the multidimensional twin model and performs adaptive compensation calibration on the static estimated duration of each test item to obtain the estimated test duration for the corresponding task on the PCB under test. The workstation allocation module divides the tasks under test into long and short tasks based on the estimated test duration. It then sorts the divided tasks using a hybrid strategy to obtain a queue for allocation, and introduces a short task priority window and workstation attenuation to allocate the tasks in the queue to the required workstations. The test station has a task queue; a parallel testing module is used to execute task tests at the test station based on the task queue, and introduces a resource-aware dynamic backfilling mechanism to obtain the real-time resource list of the test station during long task testing, and to test short tasks in the queue to be allocated in parallel according to the real-time resource list; a task queue update module is used to update the multi-dimensional twin model and the estimated test time of unexecuted tasks based on the actual test time of completed test tasks, and update the task queue of unexecuted tasks at the test station according to the prediction deviation rate of the corresponding estimated test time of unexecuted tasks before and after the update.

[0080] The above formulas are all dimensionless calculations. The formulas are derived from software simulations based on a large amount of collected data to obtain the most recent real-world results. The preset parameters in the formulas are set by those skilled in the art according to the actual situation.

[0081] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented, in whole or in part, in the form of a computer program product.

[0082] Those skilled in the art will recognize that the modules and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0083] In addition, the functional modules in the various embodiments of this application can be integrated into one processing module, or each module can exist physically separately, or two or more modules can be integrated into one module.

[0084] The above are merely specific embodiments of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

[0085] In conclusion, the above are merely preferred embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A PCB automated functional testing method based on parallel testing technology, characterized in that, Includes the following steps: A multidimensional twin model is constructed based on multidimensional acquisition data of PCB. The static estimated duration of each test item in the PCB under test is matched according to the multidimensional twin model. Adaptive compensation and calibration are performed on the static estimated duration of each test item to obtain the estimated test duration of the corresponding test task of the PCB under test. Based on the estimated test duration, the tasks to be tested are divided into long tasks and short tasks. The divided tasks are sorted using a mixed strategy to obtain the queue to be assigned. A short task priority window and workstation decay are introduced. The tasks to be tested in the queue to be assigned are then assigned to workstations to obtain the workstation task queue for the test workstations. The test station's task test is executed based on the workstation task queue, and a resource-aware dynamic backfilling mechanism is introduced to obtain the real-time resource list of the test station during long task testing, and short tasks in the assigned queue are tested in parallel based on the real-time resource list. The multidimensional twin model and the estimated test duration of unexecuted tasks are updated based on the actual test duration of completed test tasks. The task queue of unexecuted tasks at the test station is also updated based on the prediction deviation rate of the estimated test duration of unexecuted tasks before and after the update.

2. The PCB automated functional testing method based on parallel testing technology according to claim 1, characterized in that, A multidimensional twin model is constructed based on multidimensional acquisition data from PCBs, including: The PCB multidimensional acquisition data includes test item duration data, test instrument data, and environmental sensor data. A PCB twin layer with a three-layer index structure is constructed based on the test item duration data. The three-layer index structure includes the model and batch of the tested PCB, as well as the test item duration and average test duration of each test item. Construct a workstation twin layer based on test instrument data, which includes test station ID and workstation instrument information; An environmental twin layer containing a duration compensation function is constructed based on environmental sensor data, wherein the duration compensation function is generated based on environmental reference values; A multidimensional twin model is constructed based on the PCB twin layer, the workstation twin layer, and the environment twin layer.

3. The PCB automated functional testing method based on parallel testing technology according to claim 1, characterized in that, Based on the multidimensional twin model, the static estimated duration of each test item in the PCB under test is matched, including: Obtain the test parameters of the PCB under test, which include model, batch and test items, and the PCB under test contains at least one test item; Based on the PCB twin layer in the multidimensional twin model, the model and batch of the PCB under test are matched and queried to obtain the average test duration corresponding to each test item in the PCB under test, and the average test duration is used as the static estimated duration of each test item.

4. The PCB automated functional testing method based on parallel testing technology according to claim 2, characterized in that, Adaptive compensation calibration is performed on the static estimated duration of each test item to obtain the estimated test duration for the corresponding test task of the PCB under test, including: Based on the test duration and static estimated duration of each batch of test items in the PCB twin layer, calculate the batch compensation amount for each batch, and obtain the batch compensation amount for each test item in the PCB under test according to the batch of the PCB under test. Obtain the environmental parameters of the PCB under test, and calculate the environmental compensation amount corresponding to the environmental parameters based on the duration compensation function in the environmental twin layer. Obtain the list of associated instruments for each test item in the PCB under test. Based on the workstation twin layer, obtain the annual decay rate and service life of each instrument in the associated instrument list, and calculate the aging compensation amount of each instrument. Take the maximum aging compensation amount of each instrument in the associated instrument list as the instrument compensation amount of the test item. Based on the batch compensation amount, environmental compensation amount, and instrument compensation amount, the static estimated duration of each test item is adaptively compensated, and the static estimated duration of each test item after compensation is summed to obtain the estimated test duration of the test task corresponding to the PCB under test.

5. The PCB automated functional testing method based on parallel testing technology according to claim 1, characterized in that, Based on the estimated test duration, the tasks to be tested are divided into long tasks and short tasks. A mixed-strategy sorting process is then applied to these tasks to obtain a queue for assignment, including: A queue of tasks to be tested is generated based on the tasks to be tested. The median of the estimated test duration for each task in the queue is used as the classification benchmark value. The dynamic classification duration is calculated based on the classification benchmark value and the preset classification threshold. If the estimated test duration of the task to be tested is less than the dynamic classification duration, the task to be tested is determined to be a short task; if the estimated test duration of the task to be tested is greater than or equal to the dynamic classification duration, the task to be tested is determined to be a long task. Short tasks are sorted in ascending order of estimated test duration from shortest to longest to obtain a short task sequence; long tasks are sorted in descending order of estimated test duration from longest to shortest to obtain a long task sequence; and the short task sequence is placed before the long task sequence to generate a queue to be assigned.

6. The PCB automated functional testing method based on parallel testing technology according to claim 1, characterized in that, Introducing short task priority window and workstation attenuation, including: Set the station status variable for the test station, which includes the expected completion time and the list of assigned tasks; Based on the number of test stations and a preset window definition constant, a short task priority window is generated, and the priority allocation range of the test tasks is defined according to the short task priority window. Obtain the annual attenuation rate and preset instrument weight value of each test instrument in the test station, and calculate the station attenuation degree based on the annual attenuation rate and instrument weight value.

7. The PCB automated functional testing method based on parallel testing technology according to claim 1, characterized in that, The task queue for test stations is obtained by assigning workstations to the tasks to be tested in the task allocation queue, including: Get the queue position number of the task to be tested in the queue to be assigned; Get the task to be tested that is a short task and whose queue position number is in the short task priority window, and assign the task to be tested to the test station with the shortest expected completion time. Get the tasks to be tested that are long tasks and whose queue position number is outside the short task priority window. Iterate through and calculate the simulated test time of the tasks to be tested at each test station, and assign the tasks to be tested to the test station with the smallest simulated test time and station attenuation. A task queue is generated based on the tasks to be tested assigned in the test station, and the station status variables of the test station are updated based on the task queue.

8. The PCB automated functional testing method based on parallel testing technology according to claim 1, characterized in that, Parallel testing of short tasks in the allocation queue based on the real-time resource inventory includes: The real-time resource list of the test station includes a list of idle resources and a list of occupied resources for the test station. When a test station starts executing a long task, a dynamic backfilling mechanism is triggered, and the list of idle resources of the test station is obtained. Based on the list of idle resources, dynamic backfilling conditions for the test station are generated. Obtain the task parameters of each short task in the queue to be assigned, backfill the short tasks whose task parameters meet the dynamic backfilling conditions to the test station, and execute the short tasks in parallel while the long task is being executed at the test station.

9. The PCB automated functional testing method based on parallel testing technology according to claim 1, characterized in that, Based on the estimated deviation rate of the estimated test duration for unexecuted tasks before and after the update, update the task queue for unexecuted tasks at the test workstation, including: Obtain the actual test duration of each test item in the completed test task, and update the average test duration of the corresponding test item in the PCB twin layer based on the actual test duration; Get all unexecuted tasks in the workstation task queue, obtain the estimated test duration of the unexecuted tasks after the update based on the updated PCB twin layer, and obtain the estimated deviation rate of the unexecuted tasks based on the estimated test duration of the unexecuted tasks before and after the update. The estimated deviation rate of all unexecuted tasks is compared with a preset deviation threshold. If the estimated deviation rate of an unexecuted task is greater than the preset deviation threshold, it is counted as an update trigger event, and the trigger timestamp of the update trigger event is recorded. When there are two consecutive update trigger events, and the interval between the trigger events is greater than the minimum test duration of the test item, the workstation task queue of the unexecuted task in the test station is updated, wherein the minimum test duration of the test item is obtained based on the updated PCB twin layer.

10. A system for automated functional testing of PCBs using the parallel testing technology as described in any one of claims 1-9, comprising: The duration estimation module is used to construct a multidimensional twin model based on the multidimensional data collected from the PCB. Based on the multidimensional twin model, the static estimated duration of each test item in the PCB under test is matched, and the static estimated duration of each test item is adaptively compensated and calibrated to obtain the estimated test duration of the corresponding test task of the PCB under test. The workstation allocation module is used to divide the tasks to be tested into long tasks and short tasks based on the estimated test duration. The divided tasks to be tested are sorted using a mixed strategy to obtain the queue to be allocated. A short task priority window and workstation decay are introduced to allocate the tasks to be tested in the queue to obtain the workstation task queue of the test workstation. The parallel testing module is used to execute task tests of test workstations based on the workstation task queue, and introduces a resource-aware dynamic backfilling mechanism to obtain the real-time resource list of the test workstation during long task testing, and to test the short tasks in the queue to be allocated in parallel based on the real-time resource list. The task queue update module is used to update the multidimensional twin model and the estimated test duration of unexecuted tasks based on the actual test duration of completed test tasks, and update the workstation task queue of unexecuted tasks at the test station according to the estimated deviation rate of the estimated test duration of unexecuted tasks before and after the update.

Images