Interactive machine parameter optimization system and method for batch and staged inspection

Abstract

An interactive machine (2) parameter optimization system and method for batch and staged inspection. The interactive machine (2) parameter optimization system comprises a database (12) and a processor (11). The database (12) records a plurality of parameter categories to be tested. The processor (11) executes the following procedures: obtaining the number of batch-recommended combinations (S21); generating at least one parameter combination on the basis of a plurality of parameter categories and the number of batch-recommended combinations (S22); performing at least one stage of inspection on a product produced on the basis of each parameter combination (S23), and generating an inspection result corresponding to each stage of inspection (S24), wherein when the current number of stages does not reach a preset number of inspection stages, parameter combinations meeting the criteria of this stage are obtained (S25), and next stage of inspection is performed on products corresponding to these parameter combinations (S26); when the number of inspection stages is reached and a preset stop condition is satisfied (S27), generating an optimized parameter combination (S28); and when the number of inspection stages is reached but the preset stop condition is not satisfied, repeating the procedures above.

Description

Batch and stage detection interactive machine parameter optimization system and method TECHNICAL FIELD

[0001] The present application relates to a machine parameter optimization system and method, and more particularly to a batch and stage detection interactive machine parameter optimization system and method. BACKGROUND

[0002] Currently, the industrial production line mainly adopts a small amount of production mode, and thus needs to be frequently changed. Frequent line change requires frequent adjustment of machine operation parameters. However, the current production line is mainly adjusted by production line personnel based on experience (such as trial and error method). Such adjustment method is inefficient, and is limited by the experience and intuition of production line personnel, resulting in complex and time-consuming parameter adjustment of the machine.

[0003] In addition, during the parameter adjustment process, the products produced by the machine based on various parameters need to be detected. However, with the increase of detection items and the increasing complexity of the process, a longer measurement waiting time is required, resulting in longer quality detection time.

[0004] Currently, the common machine test method is to comprehensively detect the product quality without the concept of stage testing. For example, the pre-planned experimental method (such as Taguchi method and exhaustive method) mainly completes the setting of all experimental parameter combinations at the beginning of the experiment, and all these experimental parameter combinations are tested at one time. In addition, even the interactive experimental method (such as Bayes optimization) adjusts the subsequent experimental plan dynamically according to the experimental process, but only one set of parameters is recommended for experiment in each round. In other words, the current machine test method is extremely unfavorable for industrial production processes that require production efficiency and rapid adjustment to the best parameters, and is not conducive to production lines that need to be frequently changed.

[0005] SUMMARY

[0006] The main purpose of the present application is to provide a batch and stage detection interactive machine parameter optimization system and method, which further saves detection time in the tedious and lengthy detection procedure through stage detection. In addition, the batch recommended parameter mode makes the system reduce the number of rounds required for recommendation, thereby effectively shortening the time to find the best parameters.

[0007] In one embodiment, the interactive machine parameter optimization system of the present application comprises:

[0008] a database recording a plurality of parameter categories; and

[0009] a processor coupled to the database and performing the following procedures:

[0010] (a) obtaining a batch number of recommended groups, a number of detection stages, and a preset stopping condition;

[0011] (b) generating at least one group of parameter combinations based on the plurality of parameter categories and the batch number of recommended groups;

[0012] (c) performing at least one stage of detection on at least one product produced according to each group of parameter combinations to generate a stage detection result corresponding to each stage of detection;

[0013] (d) determining whether a number of stages of the current stage of detection reaches the number of detection stages;

[0014] (e) when the number of stages reaches the number of detection stages, determining whether the preset stopping condition is reached;

[0015] (f) when it is determined that the preset stopping condition is reached, generating an optimized group of parameter combinations based on all the stage detection results obtained in the procedure (c);

[0016] (g) when it is determined that the preset stopping condition is not reached, repeating the procedures (b)-(f);

[0017] wherein the procedure (c) comprises: when the number of stages does not reach the number of detection stages, obtaining the parameter combinations corresponding to the stage detection results meeting a stage standard, and performing a next stage of detection on at least one product produced according to the parameter combinations.

[0018] In an embodiment, the interactive machine parameter optimization method of the present application comprises:

[0019] a) obtaining a plurality of parameter categories in a database;

[0020] b) setting, by a processor, a batch number of recommended groups, a number of detection stages, and a preset stopping condition;

[0021] c) generating, by the processor, at least one group of parameter combinations based on the plurality of parameter categories and the batch number of recommended groups;

[0022] d) performing, by the processor, at least one stage of detection on at least one product produced according to each group of parameter combinations to generate a stage detection result corresponding to each stage of detection;

[0023] e) determining, by the processor, whether a number of stages of the current stage of detection reaches the number of detection stages;

[0024] f) when the number of stages reaches the number of detection stages, determining whether the preset stopping condition is reached;

[0025] g) when judging that the preset stop condition is reached, generating a set of optimized parameter combinations by the processor according to all the stage detection results obtained in the step d); and

[0026] h) when judging that the preset stop condition is not reached, repeating the step c) to the step g);

[0027] wherein the step d) comprises: when the stage number does not reach the detection stage number, obtaining the parameter combinations corresponding to the stage detection results meeting a stage criterion, and performing next stage detection on at least one product produced according to the parameter combinations.

[0028] Compared with the related art, the present application can obtain the best parameters in the shortest time by batch-recommending parameters and combining with stage-by-stage detection, thereby being suitable for production lines which need to frequently change lines and highly emphasize parameter adjustment efficiency. BRIEF DESCRIPTION OF DRAWINGS

[0029] Fig. 1 is an embodiment of a block diagram of an optimization system of the present application;

[0030] Fig. 2 is an embodiment of a flowchart of an optimization method of the present application;

[0031] Fig. 3 is a schematic diagram of an input interface of the present application;

[0032] Fig. 4 is a first embodiment of a schematic diagram of a recommended parameter page of the present application;

[0033] Fig. 5 is a first embodiment of a schematic diagram of a detection result return page of the present application;

[0034] Fig. 6 is a second embodiment of a schematic diagram of a detection result return page of the present application;

[0035] Fig. 7 is a third embodiment of a schematic diagram of a detection result return page of the present application;

[0036] Fig. 8 is a fourth embodiment of a schematic diagram of a detection result return page of the present application;

[0037] Fig. 9 is a second embodiment of a schematic diagram of a recommended parameter page of the present application.

[0038] Reference signs 1: optimization system 11: processor 12: database 121: parameter category 122: adjustment factor 123: detection item 124: quality criterion 13: communication unit 14: human-machine interface 2: machine S20-S28: optimization steps DETAILED DESCRIPTION

[0039] The present application discloses a batch and stage detection interactive machine parameter optimization system and method, which can recommend a plurality of parameter combinations for experiments in batches, thereby shortening the experimental time, and dividing the detection items into a plurality of detection stages to avoid unnecessary detection actions. In addition, the present application discloses an optimization system and method which can perform interactive experiments, and recommend a parameter combination for the next round of experiments (i.e., the nth round) by using all the measurement results (i.e., the 1st round to the (n-1)th round) so far, or using all the historical detection data (detection results) to recommend a parameter combination for the next round of experiments, so as to improve the fitting degree of the recommended parameter combination.

[0040] Specifically, the test method performed by a general machine is a pre-experiment method, that is, before the experiment starts, all experimental parameter combinations are set at one time, and then all the experimental parameter combinations are tested at one time, and experimental results are generated. In contrast, the optimization system and method of the present application involves interactive experiments, that is, the system and the machine have an interactive feeling. In the present application, the system first generates or recommends a batch of parameter combinations, the machine performs the first round of experiments according to the recommended parameter combinations and feeds back the experimental results to the system, and the system recommends a parameter combination for the next round of experiments according to the experimental results of the first round.

[0041] Please refer to FIG. 1, which is an embodiment of a block diagram of the optimization system of the present application. As shown in FIG. 1, the batch and stage detection interactive machine parameter optimization system (hereinafter referred to as optimization system 1) of the present application at least includes a processor 11 and a database 12, wherein the processor 11 is coupled to the database 12.

[0042] In one embodiment, the processor 11 is a Central Processing Unit (CPU), a Micro Control Unit (MCU), a Programmable Logic Controller (PLC), a System on Chip (SoC), or a Field Programmable Gate Array (FPGA), but not limited thereto. The database 12 is a Double Data Rate (DDR) memory, a Flash memory, a Random Access Memory (RAM), a Read Only Memory (ROM), a Hard DisK Drive (HDD), a Solid State Drive (SSD), or a combination thereof, but not limited thereto. The database 12 records computer executable program codes (not shown in the figure) for implementing the interactive machine parameter optimization method (hereinafter referred to as the optimization method) of the present application, which is batch and stage detection, after the optimization system 1 is started and the processor 11 reads and executes the computer executable program codes.

[0043] The optimization system 1 of the present application can also have a communication unit 13 coupled to the processor 11. As shown in FIG. 1, the communication unit 13 of the optimization system 1 is connected to the external machine 2 through wired or wireless means. One of the purposes of the present application is to automatically recommend the applicable parameter combination from the optimization system 1 according to the corresponding product after the machine 2 is changed. Thus, after the machine 2 performs the necessary detection on the parameter combination recommended by the optimization system 1, the best parameter combination for producing the product can be determined.

[0044] As shown in FIG. 1, the database 12 records a plurality of parameter categories 121, one or more adjustment factors 122 of each parameter category 121, one or more detection items 123 included in one or more detection stages, and quality standards 124 of each detection item 123. The data recorded in the database 12 is a problem that needs to be defined in advance by the optimization system 1 of the present application (for example, automatically defined according to the product specification), and recorded in the database 12 after the definition is completed. In another embodiment, the data in the database 12 can also be manually input into the optimization system 1 and stored in the database 12 by the user when the optimization system 1 is to be used.

[0045] In one embodiment, the optimization system 1 records a plurality of parameter categories 121 and corresponding adjustment factors 122 (e.g., minimum adjustment amount, adjustment unit, upper limit value, lower limit value, etc.) for one or more products that can be produced by the machine 2. When the optimization system 1 is triggered to recommend one or more sets of parameter combinations, the processor 11 can generate recommended values for each parameter category 121 based on the adjustment factors 122, so that the machine 2 can use one or more sets of parameter combinations recommended by the optimization system 1 to perform experiments and eventually find the optimal parameter combination.

[0046] The parameter categories 121 and the adjustment factors 122 are shown in the following table, but are not limited thereto.

[0047] In the above table, the parameter categories 121 are exemplified by twelve (i.e., parameter 1 to parameter 12), but are only examples, and different products have different numbers and contents of parameter categories 121, and are not limited to the above table. For example, an injection molding machine, the parameter categories 121 can include, for example, injection speed, injection pressure, holding pressure speed, and holding pressure.

[0048] When the machine 2 produces corresponding products (i.e., products to be tested) according to one or more sets of parameter combinations recommended by the optimization system 1, the machine 2 can test these products based on the detection items 123 and the quality standards 124 of each detection item 123. Finally, the optimization system 1 can recommend one or more sets of parameter combinations for the next round according to the test results of these products. In one embodiment, the quality standards 124 can include units of detection items 123, detection value specifications, optimization targets, specification requirements, and pass standards, but are not limited thereto.

[0049] The detection items 123 and the quality standards 124 are shown in the following table, but are not limited thereto.

[0050] In the above table, the detection items 123 are exemplified by four (including appearance detection 1, appearance detection 2, detail detection 3, and detail detection 4), but are only examples, and different products have different numbers and contents of detection items 123, and are not limited to the above table.

[0051] As shown in FIG. 1, the optimization system 1 of the present application also has a human-machine interface 14 coupled to the processor 11. In one embodiment, the optimization system 1 receives the number of batch recommendation groups, the number of detection stages, and the preset stopping condition from the user through the human-machine interface 14. As mentioned above, the optimization system 1 of the present application shortens the experiment time through batch recommendation and stage-by-stage detection. In one embodiment, the number of batch recommendation groups represents the number of parameter combinations that the optimization system 1 needs to recommend simultaneously in each round (e.g., 3 groups, 5 groups, etc.); the number of detection stages represents the number of stages that the machine 2 needs to distinguish when detecting the product (e.g., in the above table, the multiple detection items 123 can be distinguished into two detection stages (including the appearance detection stage and the detail detection stage)); and the preset stopping condition represents when the optimization system 1 needs to repeat the batch recommendation (i.e., the number of execution rounds).

[0052] In another embodiment, the optimization system 1 can also store the number of batch recommendation groups, the number of detection stages, and the preset stopping condition in the database 12 in advance so that the processor 11 can read them without limitation.

[0053] When the machine 2 needs to recommend parameter combinations due to line change, the user first operates the optimization system 1 through the human-machine interface 14 to select the product to be experimented in the optimization system 1. Then, the optimization system 1 automatically acquires the corresponding template stored in advance and acquires the multiple parameter categories 121, the multiple adjustment factors 122, the multiple detection items 123, and the quality standards 124 of each detection item 123 corresponding to the template from the database 12.

[0054] The optimization system 1 also receives (or reads from the database 12) the number of batch recommendation groups (e.g., 3 groups), the number of detection stages (e.g., 2 stages), and the preset stopping condition (e.g., 6 rounds of execution) required by the user for this detection through the human-machine interface 14. After acquiring the above data, the processor 11 of the optimization system 1 generates one or more parameter combinations based on each parameter category 121 and the number of batch recommendation groups, and the number of the one or more parameter combinations is the same as the number of batch recommendation groups input by the user. For example, if the number of batch recommendation groups is K, the processor 11 generates K groups of parameter combinations at a time, wherein each group of parameter combinations contains recommended values of all parameter categories 121. It is worth mentioning that the processor 11 generates one or more parameter combinations based on one or more adjustment factors 122 of each parameter category 121 and the number of batch recommendation groups, i.e., calculates the recommended values of each parameter category 121, so that the recommended values of each parameter category 121 in each group of parameter combinations are not repeated. The number of batch recommendation groups K is a positive integer, and K≥1.

[0055] After the K sets of parameter combinations are generated, the machine 2 can automatically obtain the K sets of parameter combinations, or the user can input the K sets of parameter combinations into the machine 2. Thus, the machine 2 can produce one or more products according to the K sets of parameter combinations, wherein each product corresponds to one set of parameter combinations. Then, the machine 2 or other detection equipment (not shown in the figure) can detect the one or more products respectively, and generate corresponding detection results.

[0056] It is worth mentioning that when the products are detected, based on the number of detection stages, the machine 2 only needs to perform one or more detection items of the first stage detection on the products, and feed back the corresponding detection results to the optimization system 1. Only when a product passes all detection items of the first stage detection, the machine 2 needs to perform the next stage detection on this product. Thus, the technical effect of improving the detection efficiency by stage detection can be achieved.

[0057] In an embodiment, the optimization system 1 can also obtain the number of repeated detections M required by the user through the human-computer interface 14, wherein the number of repeated detections M represents the number of times each set of parameter combinations needs to be detected as desired by the user. For example, if the number of batch recommended groups K is 3, and the number of repeated detections M is 2, it means that the machine 2 needs to use the optimization system 1 to produce 2 corresponding products for each set of parameter combinations recommended in this round, so the machine 2 needs to produce a total of 6 products in this round, and perform the stage detection procedure on the 6 products respectively.

[0058] When the machine 2 performs all detection items 123 of the i-th stage detection (wherein 1≤i≤L, L is the number of detection stages) on the K products respectively produced according to the K sets of parameter combinations, the optimization system 1 can obtain the detection results of these products in the i-th stage detection. In this application, the optimization system 1 obtains the detection results of each product (i.e., each set of parameter combinations) in the i-th stage detection, and only when a product (i.e., a set of parameter combinations) passes all detection items 123 of the i-th stage detection, the feedback mechanism of the next stage detection (i.e., the (i+1)-th stage detection) of this parameter combination is triggered. And only after the feedback mechanism of the next stage detection of a parameter combination is triggered, the processor 11 of the optimization system 1 can further obtain the detection results of the product corresponding to this parameter combination in the next stage detection through the triggered feedback mechanism (such as the reward field shown in Figure 5). In other words, when a product passes all detection items 123 of the i-th stage detection, the machine 2 has the necessity to perform the (i+1)-th stage detection on this product, and the optimization system 1 also has the necessity to receive the detection results of this product after the (i+1)-th stage detection.

[0059] Therefore, after the machine 2 completes the i-th stage detection on the K products, the processor 11 of the optimization system 1 first determines whether the number of stages of the i-th stage detection currently being performed reaches the number of detection stages (i.e., whether the detection is completed). When the number of stages of the i-th stage detection currently being performed does not reach the number of detection stages, the processor 11 of the optimization system 1 can obtain the parameter combinations corresponding to one or more products that meet the stage criteria of all detection items 123 of the i-th stage detection, respectively, and perform the next stage detection (i.e., the (i+1)-th stage detection) on the products produced according to the parameter combinations, and obtain the detection results of one or more detection items 123 included in the next stage detection performed on the products.

[0060] In an embodiment, the optimization system 1 generates the production instructions corresponding to the K sets of parameter combinations based on the K sets of parameter combinations through the processor 11, and transmits the production instructions to the machine 2 through the communication unit 13. In this way, the machine 2 can automatically produce the corresponding K products. In this embodiment, the optimization system 1 also automatically receives the detection results generated by the machine 2 after performing the detection (e.g., the i-th stage detection, the (i+1)-th stage detection, etc.) on the products through the communication unit 13. In another embodiment, after the processor 11 generates the K sets of parameter combinations, the user manually controls the machine 2 to produce one or more products according to the K sets of parameter combinations, and after the detection is completed, the user also manually inputs the detection results of the products to the optimization system 1.

[0061] In this application, the optimization system 1 and the machine 2 will repeatedly perform the above actions. That is, for one or more products that pass the detection of all detection items 123 of the previous stage detection, the machine 2 performs the detection of one or more detection items 123 of the next stage detection, and the optimization system 1 obtains the detection results of the products in the next stage detection, until the number of stages of the current stage detection reaches the number of detection stages set by the user. It is worth mentioning that if all products do not pass the i-th stage detection, the machine 2 will not perform the (i+1)-th stage detection, but will directly end the detection procedure. Similarly, if all products do not pass the (i+1)-th stage detection, the machine 2 will not perform the (i+2)-th stage detection.

[0062] During the detection and obtaining of the detection results, the processor 11 continuously determines whether the number of stages of the stage detection currently being performed reaches the number of detection stages, i.e., whether all stage detections are completed. If the number of stages of the stage detection currently being performed reaches the number of detection stages, it means that the K sets of parameter combinations generated this time have completed the stage-by-stage detection. At this time, the processor 11 determines whether the preset stop condition input by the user is met, such as whether the number of execution rounds reaches, or whether any product meets the quality criteria 124 of all detection items 123 of all detection stages, without limitation.

[0063] In one embodiment, the preset stopping condition is a number of execution rounds of the action of recommending K sets of parameter combinations. In this embodiment, when the processor 11 judges that the number of stages of the current stage detection has reached the number of detection stages, the processor 11 further judges whether the number of times of recommending K sets of parameter combinations has reached the number of execution rounds. If the number of times of recommending K sets of parameter combinations (e.g. 2 times) has not reached the number of execution rounds (e.g. 3 times), the processor 11 continues the procedure of the next round, and recommends a new K set of parameter combinations (i.e. the third time) in the next round, and performs detection.

[0064] In one embodiment, the preset stopping condition is that the stage detection result of at least one product corresponding to any set of parameter combinations satisfies the quality standard 124 of all detection items of all stage detections. In this embodiment, when the processor 11 judges that the number of stages of the current stage detection has reached the number of detection stages, the processor 11 further judges whether the stage detection result of any product satisfies the quality standard 124 (e.g. the detection value specification, optimization target and specification requirement of each stage detection) of all detection items of all stage detections. If any product satisfies the above condition, the processor 11 can directly take the parameter combination corresponding to this product as the best set of parameter combinations, and stop recommending a new K set of parameter combinations.

[0065] When the processor 11 judges that the number of stages of the current stage detection has reached the number of detection stages but the preset stopping condition has not been satisfied, the processor 11 refers to all detection results obtained so far, and selectively refers to the historical detection data of the machine 2, to automatically generate and recommend a new K set of parameter combinations. At this time, the optimization system 1 and the machine 2 repeat the above actions to perform stage detection on K products corresponding to the new K set of parameter combinations in a stage-by-stage manner, and obtain detection results in a stage-by-stage manner.

[0066] When the processor 11 judges that the number of stages of the current stage detection has reached the number of detection stages and the preset stopping condition has also been satisfied, it means that the detection process of this time can be ended. At this time, the optimization system 1 can automatically generate and recommend a set of optimized parameter combinations based on all detection results obtained in all stage detections, and selectively refer to the historical detection data of the machine 2.

[0067] In one embodiment, the processor 11 generates the parameter combination and the optimized parameter combination based on one or more detection results by using a statistical method, a machine learning method or a Bayesian optimization method. However, the above is only one specific embodiment of the present application, but is not limited thereto.

[0068] Reference is made to FIG. 2 to FIG. 9 concurrently, wherein FIG. 2 is an embodiment of the flow chart of the optimization method of the present application, FIG. 3 is a schematic diagram of the input interface of the present application, FIG. 4 is a first embodiment of the schematic diagram of the recommended parameter page of the present application, FIG. 5 to FIG. 8 are respectively a first embodiment, a second embodiment, a third embodiment and a fourth embodiment of the schematic diagram of the detection result feedback page of the present application, and FIG. 9 is a second embodiment of the schematic diagram of the recommended parameter page of the present application.

[0069] FIG. 2 discloses the optimization method of the present application, and the optimization method is mainly applied to the optimization system 1 shown in FIG. 1, but is not limited thereto.

[0070] As shown in FIG. 2, when the optimization system 1 of the present application is used, the optimization system 1 is first triggered by the user (for example, the parameter recommendation interface of the optimization system 1 is started), and the processor 11 obtains the plurality of parameter categories 121 from the database 12 (step S20). In an embodiment, the processor 11 can also obtain the adjustment factor 122 of each of the plurality of parameter categories 121 and one or more detection items of each circle detection stage from the database 12. In an embodiment, the processor 11 can also obtain the quality standard 124 of each detection item 123 from the database 12.

[0071] In an embodiment, the user can start a new project on the optimization system 1, input the project name, and select the template corresponding to the product to be detected. In the present application, each template corresponds to a different product category, and records the parameter categories 121, the adjustment factors 122 and the detection items 123 corresponding to the product category. Therefore, when the user selects a specific template, the processor 11 automatically obtains the required parameter categories 121, adjustment factors 122 and detection items 123 from the database 12, without the need for the user to input them manually.

[0072] It is worth mentioning that if there is a new experiment or product to be detected, the corresponding template, parameter categories 121, adjustment factors 122 and detection items 123 may not exist in the optimization system 1. In this case, the optimization system 1 can start the human-computer interface 14, and the user can directly input the product to be detected, the parameter categories 121 that the product should have, the adjustment factors 122 of each parameter category 121, and the detection items 123 that the product should undergo on the human-computer interface 14. Thus, when the optimization system 1 automatically generates the recommended parameter combination, it must comply with the above information.

[0073] Next, the processor 11 obtains the batch recommendation group number, the detection stage number and the preset stop condition (step S21). In one embodiment, the processor 11 obtains the batch recommendation group number, the detection stage number and the preset stop condition through the human-machine interface 14 by accepting external setting. In another embodiment, the batch recommendation group number, the detection stage number and the preset stop condition are stored in the database 12 through user input, and in step S21, the processor 11 directly obtains the batch recommendation group number, the detection stage number and the preset stop condition from the database 12.

[0074] In one embodiment, the processor 11 can also obtain the repeated detection number of each group of parameter combinations through the human-machine interface 14, which represents the number of times each group of parameter combinations recommended by the processor 11 needs to be detected. By repeatedly detecting the same group of parameter combinations, the stability of these parameter combinations can be effectively detected.

[0075] As shown in FIG. 3, when setting the experimental target, the optimization system 1 can accept user input through the human-machine interface 14 to recommend the number of rounds R (i.e., the number of execution rounds, which can be one of the preset stop conditions), the number of repetitions M (i.e., the repeated detection number of each group of parameter combinations), and the number of batches recommended per round (i.e., the number of groups of parameter combinations that need to be recommended simultaneously per round, which is the batch recommendation group number K). For example, if the user inputs the number of recommended rounds as 6, the number of repetitions as 2, and the number of batches recommended per round as 3 (i.e., R = 6, M = 2, K = 3), it means that the optimization system 1 needs to perform a recommended program for six rounds, and in each round, three groups of parameter combinations need to be recommended, and each group of parameter combinations needs to be detected twice.

[0076] After step S21, the processor 11 then automatically generates and recommends K groups of parameter combinations based on each parameter category 121 and the batch recommendation group number required by the user (step S22). More specifically, the processor 11 generates K groups of parameter combinations based on one or more adjustment factors 122 of each parameter category 121 and the batch recommendation group number required by the user. In this application, K is a positive integer and K ≥ 1. In other words, according to the batch recommendation group number set by the user, the processor 11 automatically generates a corresponding number of groups of parameter combinations based on the adjustment factors 122, wherein each group of parameter combinations includes recommended values of the multiple parameter categories 121 included in the template (i.e., parameter values X1-Xn shown in FIG. 4). And in one embodiment, the recommended values of each parameter category 121 of each group of parameter combinations recommended by the processor 11 are not repeated.

[0077] As shown in FIG. 4, assuming that the template selected by the user includes n parameter categories 121 and the user requires 3 batch-recommended groups, when the user triggers the start-recommendation button on the human-machine interface 14, the processor 11 will automatically generate three groups of parameter combinations (e.g., the 1st group, the 2nd group, and the 3rd group in FIG. 4) according to the adjustment factors 122 of the n parameter categories 121, and each group of parameter combinations includes n parameter categories 121 (e.g., the 1st parameter to the n parameter in FIG. 4) and the recommended values of the parameter categories 121 (e.g., X1~Xn in FIG. 4).

[0078] The batch-recommended multiple groups of parameter combinations and the single-recommended single group of parameter combinations have little difference in the experimental time for each round for the machine 2. However, the batch-recommended multiple groups of parameter combinations can allow the machine 2 to try more parameter combinations per unit time, thereby shortening the overall experimental time (e.g., the optimal parameter combination can be found with fewer rounds).

[0079] After step S22, the optimization system 1 of the present application combines the machine 2 to detect each group of parameter combinations recommended by the processor 11 (step S23). In an embodiment, the optimization system 1 can automatically transmit the multiple parameter combinations recommended by the processor 11 and the repeated detection number set by the user to the machine 2, so that the machine 2 produces a corresponding number of products. For example, the processor 11 recommends three groups of parameter combinations and the user sets the repeated detection number as two, so that the machine 2 produces a corresponding number of products (e.g., six products, but not limited thereto). In another embodiment, the user can manually input the multiple parameter combinations and the repeated detection number to the machine 2, so that the machine 2 produces a corresponding number of products.

[0080] It is worth mentioning that after the machine 2 produces the products, the machine 2 itself can perform the stage-by-stage detection on the products and generate corresponding detection results, or other detection devices (including the optimization system 1) can perform the stage-by-stage detection on the products and generate corresponding detection results. Finally, the optimization system 1 receives the detection results.

[0081] In the following description, P0 refers to all K groups of parameter combinations, one or more groups of parameter combinations passing the first stage detection are represented by P1, one or more groups of parameter combinations passing the second stage detection are represented by P2, and so on. In other words, K groups of parameter combinations P0 are used in the first stage detection, one or more groups of parameter combinations P1 passing the first stage detection are used in the second stage detection, one or more groups of parameter combinations P2 passing the second stage detection are used in the third stage detection, and so on. i-1 wherein i represents the stage number, that is, all K groups of parameter combinations (P0) are used in the first stage detection, and one or more groups of parameter combinations (P1) passing the first stage detection are used in the second stage detection. i-1 i ​=P1). Similarly, the second-stage detection will use one or more parameter combinations P1(P) that passed the first-stage detection. i-1 =P1) is used for detection, and the combination of one or more parameters (P) passes the second stage of detection. i =P2). As shown in Figure 2, in the i-th stage of detection, the K sets of parameter combinations P are tested by machine 2 or other detection equipment. i-1 The corresponding product undergoes one or more testing items 123 (hereinafter referred to as testing items S) in the i-th stage of testing. i The optimization system 1 obtains P based on the K sets of parameter combinations. i-1 At least one product produced separately underwent one or more testing items S in the i-th stage of testing. i The detection results are then generated (step S24). In one embodiment, i is a positive integer and 1≤i≤L, where L is the number of detection stages set by the user.

[0082] It is worth mentioning that if the number of repeated tests M is greater than or equal to 1, then in step S24, the processor 11 needs to obtain all the test items S corresponding to at least one product that has undergone the i-th stage test for each set of parameters. i The system then generates M test results. In one embodiment, repeated testing refers to testing multiple products corresponding to the same parameter combination separately. In another embodiment, repeated testing refers to testing the same product multiple times.

[0083] Please refer to Figures 5 and 6. In the embodiments shown in Figures 5 and 6, the number of detection stages set by the user is 2 (stages #1 and #2 are used as examples in Figure 5), where stage #1 includes three detection items S. i (Figure 6 uses items 1, 2, and 3 as examples). Furthermore, the user sets the number of repeated tests M to be 2. In this embodiment, for the first set of parameter combinations, the user or machine 2 needs to report the test results of product 1 and product 2 to the optimization system 1 respectively.

[0084] In one embodiment, after generating multiple sets of recommended parameter combinations for this round, the optimization system 1 enters the detection result report page shown in Figure 5, where the user manually or the machine 2 (or detection equipment) automatically fills in the detection results (e.g., the detection results of the first item, the detection results of the second item, and the detection results of the third item).

[0085] After obtaining all the test results of the product in the i-th stage of testing, the processor 11 of the optimization system 1 optimizes the system according to each test item S. i The quality standard 124 determines whether the test results of each product meet the standard, thereby determining which products meet all the test items S of the i-th stage of testing.i quality standards 124 of the i-th stage detection, i.e., determining one or more products and their corresponding one or more parameter combinations P to be subjected to the next stage detection i (step S25). For example, the i-th stage detection includes three detection items S i The detection value specifications of the three detection items S i are all 0-5 points, the optimization targets are all to be as small as possible, and the specification requirements are all 0 points. In this embodiment, a product must have detection results of 0 points in all of the three detection items S i to be considered as meeting the quality standards 124 of all of the detection items S i of the i-th stage detection.

[0086] In the embodiment of FIG. 7, the detection results of the two products corresponding to the first parameter combination do not meet the standards. In this case, the optimization system 1 determines that the first parameter combination does not meet all of the detection items S i of the i-th stage detection. At this time, the optimization system 1 excludes the first parameter combination from the next stage detection.

[0087] In the embodiment of FIG. 8, the detection results of the two products corresponding to the second parameter combination are determined by the optimization system 1 to meet the standards. In this case, the optimization system 1 determines that the second parameter combination meets the quality standards 124 of all of the detection items S i of the i-th stage detection. At this time, the optimization system 1 can require the machine 2 or other detection equipment to perform the next stage detection on the two products corresponding to the second parameter combination, and wait to receive the detection results of the next stage detection.

[0088] It is worth mentioning that, in the present application, the processor 11 only triggers the feedback mechanism of the next stage detection (e.g., the reward field of stage #2 in FIG. 8) when all of the products corresponding to any parameter combination pass all of the detection items S i of the previous stage detection. Furthermore, only after the feedback mechanism of the next stage detection is triggered (e.g., after the reward button in FIG. 8 is displayed), the user or the machine 2 can input the detection results of the products corresponding to the parameter combination in the next stage detection into the optimization system 1. That is, only after the feedback mechanism of the next stage detection is triggered, the processor 11 can receive the detection results of the next stage detection through the triggered feedback mechanism (i.e., the reward field). Specifically, in the present application, the feedback mechanism includes indicating the parameter combination to be used in the next stage detection, and waiting to receive the stage detection results of at least one product corresponding to the parameter combination after the next stage detection is completed.

[0089] More specifically, the phased detection of the present application is to divide the multiple detection items 124 according to types, and to classify the detection items 124 of the same or similar types into the same detection phase. For example, the first detection phase only contains appearance quality detection items (such as scorching, ragged, or underfilling, etc.), and the second detection phase only contains size quality detection items (such as unbalanced angle, unbalanced amount, inner circle perpendicularity, or outer circle perpendicularity, etc.). If a parameter combination fails the first phase detection, it means that the product produced according to the parameter combination cannot meet the quality requirements, and thus there is no need to perform the second phase detection on the product corresponding to the parameter combination.

[0090] The phased detection used in the present application can effectively avoid the waste of detection time, save the time cost of repeatedly performing unnecessary detection items, and improve the efficiency of the parameter optimization process. For example, the first phase detection can contain appearance detection items that are less time-consuming, and the second phase detection can contain detailed detection items that need to be performed by precision instruments and are more time-consuming. When a product fails the appearance detection items, the optimization system 1 does not need to perform the detailed detection items that take longer time. Therefore, by using the technical means of phased detection, the present application can effectively shorten the detection time required for traditional complete detection.

[0091] In step S25 of FIG. 2, the processor 11 determines one or more parameter combinations P i Before that, the processor 11 first determines whether the number of the current phase detection (i.e., i) reaches the number of detection phases L set by the user. When i < L (i.e., the parameter combination recommended in this round has not completed all detection phases), the processor 11 sets i = i + 1 (step S26), and executes steps S23 to S25 again. Thus, in the next phase detection (i.e., the i + 1 phase detection), the processor 11 performs detection on one or more products that meet the quality standards 124 of all detection items S i of the previous phase detection (i.e., the i phase detection) by the machine 2 or other detection devices, and obtains the detection results of one or more detection items S i+1 generated by the optimization system 1 after the i + 1 phase detection of these products, respectively. In other words, when the processor 11 determines that the number of the current phase detection does not reach the number of detection phases set by the user, it obtains one or more parameter combinations corresponding to the phase detection results that meet the phase standards of the current detection, and performs the next phase detection on at least one product produced according to the one or more parameter combinations.

[0092] Likewise, after the end of the i+1th stage detection and the user or the machine 2 returns the detection results of the product to the optimization system 1, the processor 11 determines one or more products and their corresponding one or more parameter combinations P to be detected in the next stage detection (i.e., the i+2th stage detection) according to the detection results i+1 .

[0093] When i=L, it means that the recommended parameter combinations in this round have completed all stage detections, that is, the K parameter combinations recommended by the processor 11 in this round have been detected. At this time, the processor 11 further determines whether the preset stopping condition set by the user is met (step S27). In one embodiment, the preset stopping condition is the number of execution rounds of the K parameter combinations recommended by the processor 11. In another embodiment, the preset stopping condition is that the detection results of the product corresponding to any parameter combination meet the quality standards 124 of all detection items 123 of all detection stages.

[0094] In yet another embodiment, the preset stopping condition can be the total number of experiments. The total number of experiments refers to the total number of parameter combinations that need to be detected. In this embodiment, when the processor 11 determines that the number of stages of the current stage detection reaches the number of detection stages, but the total number of parameter combinations that have been recommended does not reach the total number of experiments, the processor 11 executes steps S22 to S26 again to recommend one or more parameter combinations (not necessarily K) in the next round and perform detection until the total number of parameter combinations that have been recommended is equal to the total number of experiments.

[0095] For example, if the total number of experiments of the preset stopping condition is set to 10, the batch recommendation group number K is set to 3, and the repeated detection number M is set to 1, because the optimization system 1 will batch recommend three parameter combinations (i.e., three experiments) in each round, in the fourth round, the optimization system 1 will only recommend one parameter combination and make the machine 2 perform the last experiment. However, the above is only one specific implementation example of the present application, but is not limited thereto.

[0096] If i=L (i.e., the number of current stages reaches the number of detection stages) but the preset stopping condition is not met, the optimization system 1 returns to step S22. At this time, the processor 11 will refer to the detection results of one or more rounds previously obtained to automatically generate and recommend new K parameter combinations.

[0097] As shown in FIG. 9, when the three sets of parameter combinations recommended in the first round of recommendation are all detected, since the preset stopping condition has not been met (e.g., the user sets the number of execution rounds to be 6), the processor 11 will refer to all the detection results previously obtained (and can selectively refer to the historical data of the optimization system 1 and / or the machine 2) and again recommend new K sets of parameter combinations (FIG. 9 takes the first set of parameter combinations, the second set of parameter combinations, and the third set of parameter combinations recommended in the second round of recommendation as an example). In an embodiment, the processor 11 will score each parameter combination after receiving the detection results of each parameter combination in the previous round. Then, the processor 11 can generate new K sets of parameter combinations to be recommended in the next round according to the scores in combination with the historical data (i.e., the experimental data accumulated in each execution round or the experimental data of the machine 2 in the past). In this way, the machine 2 can produce one or more new products according to the new K sets of parameter combinations and perform phased detection on the one or more products, and the optimization system 1 can also obtain the detection results of the products in phases.

[0098] If i = L (i.e., the current stage number reaches the detection stage number) and the preset stopping condition is met, the processor 11 can generate and recommend a set of optimized parameter combinations based on all the detection results obtained in all the stages (step S28). In an embodiment, the processor 11 uses a statistical method, a machine learning method, or a Bayesian optimization method to calculate and generate the optimized parameter combinations based on all the previous detection results. When the machine 2 produces products using the optimized parameter combinations, the production process can have the best yield, or the products produced can have the best quality.

[0099] Compared with the experimental planning method for machine parameter tuning in the past, the present application batch-recommends multiple sets of parameter combinations by the optimization system 1, and through an interactive method, the parameter combinations recommended in one round are first detected, and then multiple sets of parameter combinations recommended in the next round are batch-recommended according to the detection results. Moreover, during detection, a phased detection method is used, and only the parameter combinations that pass the detection of the previous stage need to be detected in the next stage. In this way, the best parameter combinations recommended by the system can be effectively obtained in the shortest time.

Claims

1. An interactive machine parameter optimization system for batch and stage testing, the interactive machine parameter optimization system being connected to an external machine, comprising: a database recording a plurality of parameter categories; and a processor coupled to the database and performing the following procedures: (a) obtaining a batch recommended group number, a testing stage number, and a preset stopping condition; (b) generating at least one group of parameter combinations based on the plurality of parameter categories and the batch recommended group number; (c) instructing the machine to perform at least one stage testing on at least one product produced according to each group of the parameter combinations, and obtaining a stage testing result corresponding to each stage testing from the machine; (d) determining whether a stage number of the stage testing currently performed reaches the testing stage number; (e) when the stage number reaches the testing stage number, determining whether the preset stopping condition is reached; (f) when it is determined that the preset stopping condition is reached, generating an optimized group of parameter combinations based on all the stage testing results obtained in the past; and (g) when it is determined that the preset stopping condition is not reached, repeating the procedures (b)-(f); wherein the procedure (c) comprises: when the stage number does not reach the testing stage number, obtaining a parameter combination corresponding to the stage testing result meeting a stage standard, and performing a next stage testing on at least one product produced according to the corresponding parameter combination.

2. The interactive machine parameter optimization system for batch and stage testing according to claim 1, wherein the database further records one or more adjustment factors of each of the plurality of parameter categories, and in the procedure (b), the processor is configured to generate the at least one group of parameter combinations based on the one or more adjustment factors of each of the plurality of parameter categories and the batch recommended group number.

3. The interactive machine parameter optimization system for batch and stage testing according to claim 1, wherein each of the testing stage numbers comprises at least one testing item.

4. The interactive machine parameter optimization system for batch and stage testing according to claim 3, wherein in the procedure (c), the processor is configured to determine the at least one product meeting the stage standard of all the testing items of the stage testing currently performed, and take the at least one group of parameter combinations corresponding to the at least one product as the at least one group of parameter combinations in the next stage testing.

5. The interactive machine parameter optimization system for batch and stage testing according to claim 1, wherein each of the groups of parameter combinations comprises recommended values of the plurality of parameter categories, and the recommended values of each of the parameter categories of each of the groups of parameter combinations are not repeated.

6. The interactive machine parameter optimization system for batch and stage testing according to claim 1, further comprising a human-machine interface coupled to the processor, the human-machine interface receiving the batch recommended group number, the testing stage number, and the preset stopping condition, wherein the preset stopping condition is a recommended number of rounds of execution of the at least one group of parameter combinations. ​ ​ ​ ​ ​ ​ ​ ​ ​ ​ ​ 7. The interactive machine parameter optimization system with batch and stage testing of claim 1, wherein the processor further executes the following procedures: obtaining a number of repeated tests, wherein the number of repeated tests represents a number of times each set of the parameter combinations needs to be tested; wherein in the procedure (c), the processor is configured to repeat the stage testing for each set of the parameter combinations corresponding to the at least one product according to the number of repeated tests to generate a plurality of stage testing results.

8. The interactive machine parameter optimization system with batch and stage testing of claim 3, wherein in the procedure (c), the processor is configured to trigger a feedback mechanism of the next stage testing when any of the at least one product corresponding to each set of the parameter combinations passes all testing items of the current stage testing, wherein the feedback mechanism comprises indicating the parameter combination used for the next stage testing, and waiting for receiving the stage testing result of the at least one product corresponding to the parameter combination in the next stage testing.

9. The interactive machine parameter optimization system with batch and stage testing of claim 1, wherein the database further records quality standards of each testing item, and the preset stopping condition is that the stage testing result of any of the at least one product corresponding to each set of the parameter combinations meets all quality standards of the stage testing, wherein the quality standards comprise testing value specifications, optimization targets and specification requirements of each stage testing.

10. A method of interactive machine parameter optimization with batch and stage testing, comprising: a) obtaining a plurality of parameter categories in a database; b) setting a number of batch recommended groups, a number of testing stages and a preset stopping condition by a processor; c) generating at least one set of parameter combinations by the processor based on the plurality of parameter categories and the number of batch recommended groups; d) performing at least one stage testing by the processor for at least one product produced according to each set of the parameter combinations to generate a stage testing result corresponding to each of the stage testing; e) determining whether a number of the stage testing currently performed reaches the number of testing stages by the processor; f) when the number of the stage testing reaches the number of testing stages, determining whether the preset stopping condition is reached; g) when it is determined that the preset stopping condition is reached, generating an optimized set of parameter combinations by the processor according to all the stage testing results obtained in the past; h) when it is determined that the preset stopping condition is not reached, repeating the step c) to the step g); and wherein the step d) comprises: when the number of the stage testing does not reach the number of testing stages, obtaining a parameter combination corresponding to the stage testing result meeting a stage standard, and performing a next stage testing for at least one product produced according to the corresponding parameter combination. ​ 11. The interactive machine parameter optimization method for batch and stage testing according to claim 10, wherein the step a) comprises obtaining one or more adjustment factors for each of the plurality of parameter categories in the database, and the step c) comprises generating the at least one set of parameter combinations based on the one or more adjustment factors for each of the plurality of parameter categories and the batch recommendation group number.

12. The batch-and-stage detected interactive machine parameter optimization method of claim 10, wherein each of the stage detections in the stage detection number respectively comprises at least one detection item, and the step d) comprises: determining the at least one product that meets the stage criteria of all the testing items of the current stage testing, and taking the at least one set of parameter combinations corresponding to the at least one product as the at least one set of parameter combinations in the next stage testing.

13. The interactive machine parameter optimization method for batch and stage testing according to claim 10, wherein each of the set of parameter combinations comprises recommended values for each of the plurality of parameter categories, and the recommended values for each of the plurality of parameter categories in each of the set of parameter combinations are not repeated.

14. The interactive machine parameter optimization method for batch and stage testing according to claim 10, wherein the preset stopping condition is a number of execution rounds for recommending the at least one set of parameter combinations, and the step h) comprises repeating the steps c) to g) to recommend new at least one set of parameter combinations and perform testing in the next round when it is determined that the stage number reaches the testing stage number but the number of execution times of the step d) has not reached the number of execution rounds.

15. The interactive machine parameter optimization method for batch and stage testing according to claim 10, wherein the step a) further comprises obtaining quality criteria for each of the testing items in the database, wherein the preset stopping condition is that the stage testing result of the at least one product corresponding to any of the parameter combinations meets the quality criteria of all the stage testing, and the quality criteria comprises testing value specifications, optimization targets, and specification requirements of each of the stage testing.

16. The interactive machine parameter optimization method for batch and stage testing according to claim 10, wherein the preset stopping condition is a total number of experiments, the total number of experiments is a total number of the parameter combinations that need to be tested, and the step h) comprises repeating the steps c) to g) to recommend new at least one set of parameter combinations and perform testing in the next round until the total number of the at least one set of parameter combinations that have been recommended equals the total number of experiments when it is determined that the stage number reaches the testing stage number but the total number of the at least one set of parameter combinations that have been recommended has not reached the total number of experiments.

17. The interactive machine parameter optimization method for batch and stage testing according to claim 10, further comprising: b1) obtaining, by the processor, a number of repeated testing, wherein the number of repeated testing represents a number of times each of the set of parameter combinations needs to be tested; wherein, in the step d), the processor repeatedly performs the stage testing on the at least one product corresponding to each of the set of parameter combinations according to the number of repeated testing to generate a plurality of stage testing results.

18. The method of claim 17, wherein in step d), the processor triggers a feedback mechanism for a next stage of detection when any of the product corresponding to the parameter combination passes all the items of the current stage of detection, wherein the feedback mechanism comprises indicating the parameter combination used for the next stage of detection, and waiting for receiving the stage detection results of the product corresponding to the parameter combination in the next stage of detection.

19. The method of claim 11, wherein the step b) is performed by the processor accepting external setting of the batch recommended group number, the detection stage number, and the preset stop condition, or retrieving the batch recommended group number, the detection stage number, and the preset stop condition from the database.

Images