Design condition quality risk pre-detection method

Abstract

This disclosure relates to quality risk detection technology, and provides a quality risk detection device and method, which acquires quality performance information of products manufactured based on design conditions that have a predetermined level or higher similarity to the design conditions from a first database; calculates a first quality risk index or a second quality risk index by judging whether the number of acquired quality performance information is N (N is an integer greater than or equal to 1) or more, based on a judgment result; and outputs the quality risk detection result of the design conditions based on either the first quality risk index or the second quality risk index.

Description

Technical Field

[0001] This disclosure relates to quality risk detection technology for design conditions. Background Technology

[0002] In the manufacturing sector, production follows a customer-order-based model. Therefore, the design conditions corresponding to the customer's requirements are pre-registered, and when an order is received, the product is manufactured by matching the appropriate design conditions.

[0003] To minimize design errors and product defects, it is necessary to verify the existence of risks in pre-registered design conditions. However, manually collecting and analyzing data based on multiple design conditions to determine their validity is time-consuming. Furthermore, for new design conditions without prior production history, the lack of existing data makes it difficult to assess potential risks.

[0004] Therefore, the need for a method to assess and manage the risks of pre-registered design conditions is proposed. Summary of the Invention

[0005] (a) Technical problems to be solved This disclosure aims to provide a technique for detecting quality risks in design conditions.

[0006] (II) Technical Solution According to one aspect, this embodiment provides a quality risk detection method, which is a method for detecting quality risks of design conditions in an apparatus, including: an acquisition step, acquiring quality performance information of products produced based on design conditions that have a predetermined level or higher similarity to the design conditions from a first database; a calculation step, calculating a first quality risk index or calculating a second quality risk index through an artificial intelligence model based on a judgment result that the number of acquired quality performance information is N (N is an integer greater than or equal to 1); and an output step, outputting the quality risk detection result of the design conditions based on either the calculated first quality risk index or the second quality risk index.

[0007] According to another aspect, this embodiment provides a quality risk detection device, which is a device for detecting quality risks of design conditions, including: an information acquisition unit, which acquires quality performance information of products produced based on design conditions that have a predetermined level or higher similarity to the design conditions from a first database; and a calculation unit, which calculates a first quality risk index or calculates a second quality risk index by means of an artificial intelligence model, based on a judgment result that determines whether the number of acquired quality performance information is N (N is an integer greater than or equal to 1), and outputs a quality risk detection result of the design conditions based on either the first quality risk index or the second quality risk index.

[0008] (III) Beneficial Effects This disclosure provides a quality risk detection technology for design conditions. Attached Figure Description

[0009] Figure 1 This is a diagram illustrating the process of deriving design condition quality risk detection results according to one embodiment.

[0010] Figure 2 This is a diagram illustrating the overall process of detecting quality risks according to an embodiment of the design conditions.

[0011] Figure 3 This is a diagram illustrating in detail the process of deriving the quality risk detection results of design conditions according to one embodiment.

[0012] Figure 4 This is a diagram illustrating the process of deriving quality risk detection results according to an embodiment.

[0013] Figure 5 This is a diagram illustrating the clustering process of pre-registered design conditions according to one embodiment.

[0014] Figure 6 This is a diagram illustrating the process of arranging quality risk detection objects according to one embodiment.

[0015] Figure 7 This is a diagram used to illustrate a similarity basis for order conditions according to one embodiment.

[0016] Figure 8 This is a diagram used to illustrate a similarity benchmark for the required quality assurance conditions according to one embodiment.

[0017] Figure 9 This is a diagram used to illustrate a similarity reference for manufacturing conditions according to one embodiment.

[0018] Figure 10 This is a diagram illustrating the arrangement of quality unit inspection objects according to one embodiment.

[0019] Figure 11 This is a diagram illustrating the inspection results of an object inspected according to a quality unit in one embodiment.

[0020] Figure 12 This is a diagram illustrating the inspection results of similar order unit inspection objects according to one embodiment.

[0021] Figure 13 This is a diagram illustrating the inspection results of an object inspected according to design conditions of one embodiment.

[0022] Figure 14 This is a diagram illustrating a design condition quality risk pre-detection device according to one embodiment. Detailed Implementation

[0023] Hereinafter, some embodiments of the present disclosure will be described in detail with reference to the illustrative accompanying drawings. When adding reference numerals to components in the various drawings, the same components will be represented by the same symbol as much as possible, even if they are shown in different drawings. Furthermore, when describing this embodiment, if it is determined that a detailed description of a related well-known structure or function would obscure the main idea of ​​the present invention, its detailed description may be omitted. Where terms such as "comprising," "having," and "consisting of" are used in this specification, other parts may be added unless "only" is used. When components are expressed as a singular, a plural number may be included unless specifically and explicitly stated otherwise.

[0024] Furthermore, when describing the components of this disclosure, terms such as first, second, A, B, (a), and (b) may be used. These terms are merely for distinguishing the component from other components and do not limit the nature, order, sequence, or number of the corresponding components.

[0025] In the description of the positional relationship of components, when two or more components are described as "connected," "combined," or "accessed," the two or more components can be directly "connected," "combined," or "accessed." However, it should be understood that further "intervention" can occur between the two or more components and other components to "connect," "combined," or "access" them. Here, other components can be included in one or more of the two or more components that are mutually "connected," "combined," or "accessed."

[0026] In descriptions of time-series relationships related to components, action methods, or production methods, for example, when describing temporal or procedural sequences using phrases such as "after," "following," "next to," or "before," discontinuous situations may be included unless "immediately following" or "directly" is used.

[0027] On the other hand, when referring to the numerical value or its corresponding information (e.g., level, etc.) of a component, even without other explicit documentation, the numerical value or its corresponding information can be interpreted as including the range of errors that may be caused by various factors (e.g., process factors, internal or external shocks, noise, etc.).

[0028] The embodiments are described in detail below with reference to the accompanying drawings.

[0029] While the manufacturing sector can prioritize producing products for sale to customers, in many cases, it follows a model of producing products only after confirming customer orders. Furthermore, the products manufactured are based on diverse design conditions tailored to customer requirements and the specific circumstances of the production facility.

[0030] Therefore, from the producer's perspective, design conditions are pre-registered based on the conditions that the customer may require. When an order is received, the product is manufactured by matching the design conditions that meet those conditions. However, in this case, it may be possible that there are no pre-registered design conditions that meet the customer's requirements, thus necessitating the registration of new design conditions.

[0031] In this regard, it is necessary to confirm in real time or in advance whether the pre-registered design conditions are suitable for producing normal products, or whether the newly registered design conditions are appropriate. However, manually verifying the large number of diverse design conditions for errors is time-consuming.

[0032] This disclosure aims to provide a scheme for identifying the quality risks of pre-registered design conditions and newly registered design conditions, as a scheme to minimize the error rate of design conditions and the defect rate of manufactured products.

[0033] Figure 1 This is a flowchart illustrating the process of deriving design condition quality risk detection results according to an embodiment.

[0034] Reference Figure 1 The method for detecting quality risks in design conditions may include an acquisition step (S100) for obtaining quality performance information.

[0035] As an example, the acquisition step can retrieve quality performance information of products manufactured based on design conditions that have a predetermined level or higher similarity to the design conditions from a database. The quality of products previously manufactured according to registered design conditions can be measured, and the suitability of the design conditions can be determined based on the measured numerical information. This disclosure can quantify the quality of products manufactured under design conditions similar to the design conditions to be tested, thereby determining whether the design conditions to be tested are suitable.

[0036] Methods for detecting quality risks in design conditions may include a calculation step (S110) for calculating quality risk indicators.

[0037] As an example, in the calculation step, a first quality risk indicator can be calculated based on whether the number of quality performance information pieces obtained in the acquisition step is N (N is an integer greater than or equal to 1), or a second quality risk indicator can be calculated using an artificial intelligence model. The number of quality performance information pieces used as the benchmark for calculating the quality risk indicator can be set by the user for convenience. If the number of quality performance information pieces obtained from the database is N or more, it means that the past quality performance data is reliable. In this case, the first quality risk indicator is calculated using the quality performance data retrieved from the database. When the retrieved data is non-existent or less than N pieces and the quantity is small, the second quality risk indicator can be calculated based on the data output by a trained and reliable artificial intelligence model.

[0038] As an example, the first quality risk indicator calculated in the calculation step may include a first process capability index and a first defect rate, and the second quality risk indicator may include a second process capability index and a second defect rate. The process capability index is a numerical value used to assess the level of process capability. Alternatively, it is a numerical value used to determine whether the current production process is suitable by comparing the required level for improving the production process with the actual production results. The defect rate is a percentage value representing the proportion of products produced under specific design conditions that do not meet a user-defined benchmark. This disclosure can calculate the process capability index and defect rate using information obtained from a database storing past production history or information output by an artificial intelligence model, thereby determining whether the design conditions to be inspected are good or bad.

[0039] For example, the first process capability index and the first defect rate calculated in the calculation step can be calculated based on the average and deviation of quality values ​​contained in the quality performance information. Furthermore, the second process capability index and the second defect rate can also be calculated based on the average and deviation of values ​​output by an artificial intelligence model; alternatively, the process capability index or defect rate can be calculated from the artificial intelligence model, thus outputting only the results.

[0040] For example, in the calculation step, when the number of quality performance information obtained from the database in the acquisition step is more than N, the first quality risk indicator can be calculated based on the quality performance information obtained from the database. When the number of quality performance information is less than N, the second quality risk indicator can be calculated based on the information output by the artificial intelligence model.

[0041] For example, the artificial intelligence model used to calculate the second quality risk indicator in the calculation step may include a Bayesian neural network model.

[0042] The method for detecting quality risks in design conditions may include an output step (S120) that outputs the quality risk detection results.

[0043] For example, the output step can output the quality risk detection results of the design conditions based on either the first quality risk index or the second quality risk index calculated in the above calculation steps.

[0044] For example, in the output step, the quality risk detection result can be output as "Good Performance" based on the first process capability index calculated in the calculation step being higher than the first benchmark value and the first defect rate being lower than the second benchmark value; the quality risk detection result can be output as "Potential Performance Risk" based on the first process capability index being lower than the first benchmark value or the first defect rate being higher than the second benchmark value; the quality risk detection result can be output as "Expected Good" based on the second process capability index being higher than the third benchmark value and the second defect rate being lower than the fourth benchmark value; and the quality risk detection result can be output as "Expected Risk" based on the second process capability index being lower than the third benchmark value or the second defect rate being higher than the fourth benchmark value. The output being "Good Performance" or "Expected Risk" is determined based on the first quality risk indicator, meaning that the number of quality performance information items retrieved from the database is more than N as set by the user; the output being "Expected Good" or "Expected Risk" is determined based on the second quality risk indicator, meaning that the number of quality performance information items retrieved is less than N.

[0045] By employing different methods for calculating quality risk indicators based on the quantity of quality performance information retrieved, the reliability of calculation results in cases of insufficient samples can be addressed, and the suitability of the design conditions to be tested can be accurately determined, thereby reducing design errors and product defect rates.

[0046] The following text provides a more detailed and varied explanation of the overall process for detecting quality risks under design conditions, with reference to the accompanying diagrams.

[0047] Figure 2 This is a diagram illustrating the overall process of detecting quality risks according to an embodiment of the design conditions.

[0048] Reference Figure 2 In passing Figure 1 The acquisition, calculation, and output steps have been detailed in detail. Adding extraction and arrangement steps can further enable more detailed detection of quality risks in design conditions.

[0049] A method for detecting quality risks in design conditions may include an extraction step, which involves extracting similar design conditions from a database storing past design history based on user-defined benchmarks (S200). The step of extracting similar design conditions can extract similar design conditions from the database storing past design history based on user-defined similarity benchmarks. Furthermore, the extraction of the aforementioned similar design conditions can be categorized according to the type of design conditions, namely, order conditions, quality assurance conditions, and manufacturing conditions.

[0050] For example, the extraction of similar design conditions based on order conditions can be based on factors such as product variety, specifications, thickness, width, and intended use. The extraction of similar design conditions based on quality assurance conditions can be based on factors such as the smelting composition and finished product composition required by the customer, material quality such as mechanical properties, and appearance quality such as shape, surface, and dimensions. Furthermore, each quality factor can include sample type, test conditions, and warranty information. The extraction of similar design conditions based on manufacturing conditions can be based on information about the core design factors designed within the plant area, factory, steel output composition, steelmaking, continuous casting, and rolling processes.

[0051] The individual design conditions extracted based on similar benchmarks can be stored in a separate database.

[0052] Methods for detecting quality risks in design conditions may include a scheduling step (S210) of arranging the inspection items to be inspected.

[0053] As an example, the items to be inspected can be determined based on at least one of the similar design conditions extracted in the extraction step and the design conditions to be inspected. An item to be inspected means, among multiple items set as design conditions, that serves as the benchmark used to check the appropriateness of the design conditions. The aforementioned items to be inspected can be arranged based on the similar design conditions extracted in the aforementioned step of extracting similar design conditions, or based on the actual design conditions to be inspected. Once the items to be inspected are determined, the relevant information for these items can be stored in a separate database.

[0054] As an example, the inspection target items may include at least one of quality unit items, order unit items, and design unit items. Based on the determined inspection target items, a quality risk detection result can be output, including at least one of a first quality risk detection result for the quality unit item, a second quality risk detection result for the order unit item, and a third quality risk detection result for the design unit item.

[0055] In order to examine the design conditions to be tested against multiple benchmarks and determine whether they are suitable, this disclosure uses at least one of the following items to judge the test results: quality unit items, order unit items, and design unit items. When at least one of the items to be tested is judged as risky, the design conditions to be tested can be output as unsuitable.

[0056] As an example, the quality risk detection result may include a fourth quality risk detection result related to the suitability of the design conditions. This fourth quality risk detection result may include any one of the following: good performance, risky performance, good expected performance, risky expected performance, or indeterminate. For instance, if at least one of the first quality risk detection result, the second quality risk detection result, and the third quality risk detection result is a risky performance or a risky expected performance, the fourth quality risk detection result may be output as either a risky performance or a risky expected performance.

[0057] This disclosure improves the accuracy of suitability assessment by allowing the use of two or more benchmarks to determine the suitability of design conditions, rather than relying on a single criterion when assessing quality risks.

[0058] Figure 3 The following sections, with reference to the accompanying drawings, provide a more detailed and varied explanation of each process involved in assessing quality risks under design conditions.

[0059] Figure 3 This is a diagram illustrating in detail the process of deriving the quality risk detection results of design conditions according to one embodiment.

[0060] The quality risk detection results of the design conditions are determined based on the calculated quality risk indicators. The quality risk detection results can be any one of the following: good performance, risky performance, good expected performance, risky expected performance, or indeterminate performance.

[0061] The process of deriving the detection results of design condition quality risk is as follows.

[0062] 1) Retrieve the quality inspection target (S300) 2) Retrieve actual performance data to determine the amount of data to be obtained (S310) (S320) 3) Calculate the mean and standard deviation of the quality performance (S330) 4) Calculate quality risk indicators based on the quantity of acquired quality performance data (S340) 5) Query the benchmark value for judging the quality risk detection results (S350) 6) Output the quality risk detection results based on the calculated quality risk indicators and judgment criteria (S360).

[0063] According to this disclosure, the steps for outputting the detection results of design condition quality risks can also be summarized as data acquisition steps, index calculation steps, and result output steps.

[0064] Specifically, the design conditions for the quality risk to be detected can be queried (S300). These design conditions can also be referred to as the quality risk inspection objects. Furthermore, the above design conditions can be used to perform inspections only on design conditions for which no quality risk inspection has been performed before, and therefore no inspection results exist in the database.

[0065] After selecting the inspection target by querying the design conditions of the quality risk to be detected, it is confirmed whether there is a past product production history based on the design conditions (S310). According to this disclosure, the existence of a past product production history can be confirmed from a database storing product production history based on design conditions. The database storing product production history can also be referred to as an operational quality performance database.

[0066] The benchmark is not limited to design conditions that are exactly the same as the design conditions; product production history can also be confirmed for design conditions that are similar to the design conditions to be tested. Similar design conditions can be selected based on user-defined benchmarks, or through artificial intelligence models or algorithms that calculate the similarity between design conditions. Furthermore, the query for the existence of past product production history can be performed within a user-defined time period.

[0067] When querying quality performance data for the design conditions to be tested from a database that stores past product production history, confirm the number of quality performance data retrieved (S320).

[0068] The reason for confirming the quantity of quality performance data is that quality risk assessment is based on statistical calculations of averages and deviations, which raises reliability concerns. If the data volume is large, quality risk is assessed based on the average and standard deviation of the retrieved quality performance data; if the data volume is small, quality risk is assessed based on data output by an artificial intelligence model. The user can set the baseline for the quantity of quality performance data, which can be set to N (N being an integer greater than or equal to 1).

[0069] After confirming the quantity of quality performance data, a quality risk indicator is calculated based on the quantity of quality performance data (S330). When the number of quality performance data retrieved is N or more, the average and deviation of the quality performance stored in the database are calculated. When the number of quality performance data retrieved is less than N, the average and deviation of the quality are calculated using a quality prediction artificial intelligence model.

[0070] As an example, when only 10 quality performance data points are retrieved and the user-defined baseline is 30, quality risk assessment cannot be performed based solely on these 10 data points. Therefore, the 10 retrieved quality performance data points can be input into a quality prediction AI model, which will output 20 additional predicted data points. The quality risk index can then be calculated using both the 10 retrieved quality performance data points and the 20 predicted data points output by the AI ​​model. Alternatively, the quality prediction AI model can be used as input data to output 30 predicted data points, and the quality risk index can be calculated based on the output predicted data.

[0071] Figure 3 For ease of explanation, the user-defined benchmarks are set to 30, but this disclosure is not limited to this. Users can set the number of benchmarks for quality performance data using the quality prediction artificial intelligence model in various ways as needed.

[0072] When the average and deviation of quality are calculated using quality data stored in the database or information output by an artificial intelligence model, the process capability index and defect rate are calculated based on the aforementioned average and deviation information (S340). As mentioned above, the process capability index is a value used to determine whether the current production process is suitable by comparing the required level for improving the production process with the actual production results. The defect rate is the percentage of products produced under specific design conditions that do not meet the user-set benchmark.

[0073] After calculating the process capability index and defect rate, the judgment benchmark (S350) for the customer's required quality is queried. The aforementioned required quality includes any one of the following: quality unit, order unit, or design unit, representing the benchmark value for judging good or risky conditions. This value is pre-stored in the database by the user and can be changed as needed.

[0074] After obtaining the judgment benchmark for the required quality, the quality risk detection result is judged and output. The quality risk detection result is judged based on whether the process capability index and the defect rate are above or below the benchmark. According to this disclosure, when the number of quality performance data is N or more and is calculated based on the values ​​stored in the database, it can be referred to as the first process capability index and the first defect rate, the first quality risk indicator, and the first quality risk detection result; while when the detection result is calculated and output through an artificial intelligence model, it can be referred to as the second process capability index and the second defect rate, the second quality risk indicator, and the second quality risk detection result.

[0075] Specifically, when the capability index of the first process is higher than the first benchmark value and the first defect rate is lower than the second benchmark value, the quality risk detection result is output as "good performance"; when the capability index of the first process is lower than the first benchmark value or the first defect rate is higher than the second benchmark value, the quality risk detection result can be output as "risk performance".

[0076] When the capability index of the second process calculated by the artificial intelligence model is higher than the third benchmark value and the second defect rate is lower than the fourth benchmark value, the quality risk detection result is output as expected good; when the capability index of the second process is lower than the third benchmark value or the second defect rate is higher than the fourth benchmark value, the quality risk detection result can be output as expected risk.

[0077] The terminology used for the above benchmark values ​​and quality risk detection results is not limited to this; users can also make different settings as needed.

[0078] Even if fewer than N data points of quality performance are found, if no other quality prediction AI model is specified, the output can be "cannot be determined".

[0079] Figure 4 This is a diagram illustrating the process of deriving quality risk detection results according to an embodiment.

[0080] If based on Figure 3 The quality risk assessment results were derived, and the assessment results for each assessment item can also be stored in a separate database.

[0081] As an example, when deriving quality risk inspection results for a quality unit item, user-specified items can be stored in a database. Furthermore, instead of directly storing the inspection results in the database, they can be categorized into separate databases for items related to component quality, material quality, and appearance quality. According to this disclosure, these databases can be referred to as a component quality inspection results database, a material quality inspection results database, and an appearance quality inspection results database, respectively.

[0082] When deriving quality risk inspection results for items in an order unit, user-specified items can also be stored in the database. Furthermore, the quality risk inspection results for items in an order unit, derived from data categorized in the database for the aforementioned component quality, material quality, and appearance quality, and storing the inspection results therein, can be summarized according to user-defined benchmarks and stored in a separate database. According to this disclosure, this database can be referred to as a similar order inspection results database.

[0083] When deriving quality risk assessment results for a design unit project, user-specified items can also be stored in a database. Furthermore, data stored in a similar order inspection results database and the derived quality risk assessment results for the design unit project can be aggregated according to user-defined benchmarks and stored in a separate database. According to this disclosure, this database can be referred to as the design condition inspection results database.

[0084] Figure 5 This is a diagram illustrating the clustering process of pre-registered design conditions according to one embodiment.

[0085] While it's possible to deduce test results by querying a database to see if there's a history of products manufactured under identical design conditions, the sheer number of design conditions—tens or even hundreds of thousands—might be insufficient to derive the results. Furthermore, design conditions for products never manufactured before are novel; without a production history, it might be impossible to determine the suitability of the design conditions under test. Therefore, this disclosure, by pre-defining what design conditions are similar to the design conditions under test and querying historical quality data based on design conditions within a similar range, can more accurately deduce the quality risks of the design conditions.

[0086] Specifically, design conditions can be categorized into order conditions, quality assurance conditions, and manufacturing conditions. An operation can be performed in advance to classify order conditions, quality assurance conditions, and manufacturing conditions according to similar conditions from a database storing the history of products manufactured based on each design condition. According to this disclosure, the database storing the history of manufactured products can be referred to as an order design result history information database.

[0087] As an example, similarity criteria for order conditions can be extracted from the historical information database of order design results (S500). The benchmark for similarity criteria can be set by the user as needed. Examples of order conditions may include product type, specifications, order thickness, order width, and order purpose.

[0088] Category-based items such as product type, specifications, and order purpose can be defined as a user-defined list of similar values, while continuous items such as order thickness and order width can be defined as a user-defined range of similar values. Once the extraction is complete, this information can be stored in a separate database. According to this disclosure, this database can be referred to as a similar order condition database.

[0089] As another example, similar quality assurance conditions can be extracted from the historical information database of order design results (S510). The benchmark for similar quality assurance conditions can be set by the user as needed. As an example, quality assurance conditions can be categorized into composition conditions, material quality, and appearance quality. In composition conditions, extraction can be based on benchmarks with the same smelting composition and finished product composition guarantee range for each alloy composition; in material quality conditions, extraction can be based on benchmarks with the same sample type, test conditions, and guarantee range for each mechanical property; and in appearance quality conditions, extraction can be based on benchmarks with guarantee ranges for the same shape, surface, and dimensions. Once extraction is complete, this information can be stored in a separate database. According to this disclosure, this database can be referred to as a similarity assurance condition database.

[0090] As another example, similar manufacturing conditions can be extracted from the historical information database of order design results (S520). Core design factors for steel output composition and process conditions such as steelmaking, continuous casting, and rolling, designed to meet customer order conditions and quality assurance conditions, can be extracted and managed according to similar clustering criteria. The criteria for similar manufacturing conditions can be set by the user as needed. For example, similar steel output composition can be extracted based on steel output target number criteria with the same deoxidation method, dephosphorization method, and alloy target composition; process conditions can be extracted based on the same criteria for steelmaking (steel plant, secondary refining method, etc.), continuous casting (light reduction conditions, flame cleaning conditions, etc.), heating furnace (tapping temperature, furnace time), rolling (start temperature, end temperature, reduction), and cooling (start temperature, end temperature, speed). Once the extraction is complete, this information can be stored in a separate database. According to this disclosure, this database can be referred to as a similar manufacturing condition database.

[0091] Figure 6 This is a diagram illustrating the process of arranging quality risk inspection objects according to one embodiment.

[0092] According to this disclosure, when detecting quality risks in design conditions, instead of simply arranging a single item for inspection, two or more items are arranged to check whether the design conditions are good or risky, thereby improving the reliability of the quality risk detection results.

[0093] As an example, inspection objects used to detect quality risks can include quality unit inspection objects, order unit inspection objects, and design unit inspection objects.

[0094] Specifically, arranging quality unit inspection objects involves connecting similar order conditions, quality assurance conditions, and manufacturing conditions using quality units that specify the requirements for composition, material, and appearance quality, and then arranging them in the quality inspection object data table of the quality risk analysis database. The basis for arranging quality unit inspection objects can be based on core design factors that influence quality. Quality unit inspection objects can be further categorized into component units, material units, and appearance unit inspection objects.

[0095] For example, tensile strength in material quality can be arranged as similar order conditions by variety, order thickness range, and width range; quality assurance conditions can be arranged by specimen type, length direction, specimen direction, and specimen number; test conditions can be arranged by YP code and guarantee range. As manufacturing conditions, plant area, factory, steel composition, steelmaking, continuous casting, and rolling conditions can be arranged.

[0096] The order unit inspection objects can be arranged from the order list, using similar order units that have the same design conditions, similar order conditions and manufacturing conditions, and the same required quality items. By arranging the order unit inspection objects, risky orders with quality risks can be detected in advance, and the same quality risk assessment indicators can be avoided when the same order occurs.

[0097] The inspection targets for design units can be those managed by the design department, arranging similar order conditions, quality assurance requirements, and manufacturing conditions as inspection targets. By arranging the inspection targets for design units, risky design conditions with potential quality risks can be detected in advance.

[0098] When the above-mentioned quality unit inspection objects, order unit inspection objects, and design unit inspection objects are arranged, the arranged data can be stored in the required quality inspection object database, the similar order inspection object database, and the design condition inspection object database, respectively.

[0099] Figure 7 This is a diagram used to illustrate a similarity basis for order conditions according to one embodiment.

[0100] Reference Figure 5 and Figure 7 In the process of clustering pre-registered design conditions, the design conditions can be classified into order conditions, quality conditions, and manufacturing conditions. Figure 7This is a diagram illustrating an example of order conditions after clustering. As mentioned above, order conditions can include product type, product specifications, thickness and width of the ordered products, order purpose, order size, etc. Order conditions are not limited to these; if pre-defined conditions are available, they can be additionally included in the order conditions. Furthermore, the similarity benchmark for conditions can be set according to a user-defined range.

[0101] As an example, categorical items such as product type can be defined as a user-defined list of similar values. For instance, among the product types listed as order conditions are hot-rolled, thick plate, and wire rod, hot-rolled can be further divided into FH and FD. Continuous items such as order thickness can also be defined as a user-defined range of similar values. As another example, in the order thickness of a product, for hot-rolled products, thicknesses between 0 and 1.5 mm can be classified into a similar group, and thicknesses between 1.5 and 6 mm can also be classified into a similar group.

[0102] Therefore, two order conditions—one for hot-rolled products with an FH code and an order thickness of 1.7, and the other for hot-rolled products with an FH code and an order thickness of 5.5—are classified as similar order conditions and can be stored in a separate database for management.

[0103] Figure 8 This is a diagram used to illustrate a similarity benchmark for the required quality assurance conditions according to one embodiment.

[0104] Reference Figure 8 Customer-required quality conditions can be categorized into composition quality, material quality, and appearance quality; composition quality can be categorized into smelting composition and finished product composition; material quality can be categorized into material properties, material internal structure, material surface, material magnetism, and material coating; appearance quality can be categorized into appearance surface, appearance shape, and appearance dimensions; and smelting composition can be categorized into single component and composite component, etc.

[0105] In addition, other conditions required by customers can be categorized and managed for component quality, material quality, and appearance quality.

[0106] As an example, in the composition quality conditions, the conditions of a single component of the smelting composition with SI as the condition are classified as similar conditions and stored in a separate database for management; More specifically, in the composition quality conditions, the single component of the smelting composition is MN, and the composite component is PCM; the single component of the finished product composition is SI, and the composite component is CEQ; conditions based on this can be classified as similar conditions and stored in a separate database for management.

[0107] The classification items and conditions for the above quality conditions are not limited to these; users can make various settings as needed.

[0108] Figure 9 This is a diagram used to illustrate a similarity reference for manufacturing conditions according to one embodiment.

[0109] Reference Figure 9 Manufacturing conditions can be classified into manufacturing plant, steel composition, and process conditions; manufacturing plant can be divided into plant area and plant; steel composition can be divided into deoxidation method, dephosphorization method, and target composition; process conditions can be divided into steelmaking, continuous casting, and rolling.

[0110] Similar to order conditions and quality conditions, non-continuous items can be categorized according to user-defined categories, while continuous items can be categorized according to user-defined ranges. Based on the categorization items, the condition items stored in the past production history information data table can be diversified and stored in the database.

[0111] As an example, design conditions with the following factory conditions can be regarded as similar manufacturing conditions and classified as such: the target composition of the steel output is C, and the furnace exit temperature of the rolling process is between 100 and 200 degrees.

[0112] Figure 10 This is a diagram illustrating the arrangement of quality unit inspection objects according to one embodiment.

[0113] As mentioned above, the quality unit can require the quality unit to connect the same order conditions, quality assurance conditions, and manufacturing conditions, thereby arranging them in the quality inspection object data table of the quality risk analysis database.

[0114] Accordingly, the quality ID and SEQ ID of each detailed item can be used as the key of the database. Data related to each item of the quality inspection object and each item of the order conditions, guarantee conditions and manufacturing conditions can be summarized and stored in the database for querying and use when necessary.

[0115] Figure 11 This is a diagram illustrating the inspection results of an object inspected according to a quality unit in one embodiment.

[0116] Reference Figure 11 This is an illustrative diagram showing the calculation of various quality risk indicators for the quality unit's inspection objects, and the resulting graph from which the quality risk detection results are derived.

[0117] As an example, for a design condition with quality ID AB1 and SEQ number 01, when it is determined that there are more than N quality performance information records found, a performance-based judgment index including process capability index and defect rate is calculated, and the judgment result of good performance or performance risk can be derived from it.

[0118] The process capability index is calculated based on average and deviation information according to user-defined benchmarks; the defect rate can be expressed as a percentage representing the proportion of defective products relative to the total number of data. When the process capability index is above the first benchmark value and the defect rate is below the second benchmark value, the quality risk detection result can be output as "good," and the overall judgment result can be output as "good performance."

[0119] As another example, for a design condition with quality ID AB2 and SEQ number 02, when it is determined that there are fewer than N pieces of queried quality performance information, the AI ​​model further outputs predicted data, and a judgment index is calculated based on the output predicted data to derive the judgment result. Alternatively, the judgment index can be calculated based on the queried quality performance information and the output predicted data, and the judgment result can be derived. In this case, the judgment result based on performance is "cannot be judged," while the judgment result based on prediction can be output as "expected risk" or "expected good." The comprehensive judgment result can be output as "expected risk" or "expected good" based on the judgment result based on prediction.

[0120] As another example, for the design condition with quality ID AB3 and SEQ number 01, when it is determined that there are fewer than N quality performance records found, the process capability index and defect rate should be calculated based on the artificial intelligence model. However, in the absence of an applicable artificial intelligence model, the judgment result based on performance and the judgment result based on prediction can be output as "cannot be judged" respectively, and the comprehensive judgment result can also be output as "cannot be judged".

[0121] Figure 12 This is a diagram illustrating the inspection results of similar order unit inspection objects according to one embodiment.

[0122] Reference Figure 12 This example demonstrates how, for each similar order condition, based on the conditions for the order condition, quality assurance condition, manufacturing condition, and quality performance information for each condition, the results of the quality risk inspection for the quality assurance condition and the derivation of the comprehensive judgment result are output.

[0123] In this disclosure, for each condition regarding the inspection objects of component units, material units, and appearance units, if even one object outputs an unsuitable quality risk test result, the overall judgment result can be determined as an unsuitable design condition. In other words, only when all inspection objects output "good" can the overall judgment result for the inspection objects of similar order units be determined as a "good" condition.

[0124] As an example, in Figure 12 If the similar order identification ID is AC1 and the design condition identification ID is BD1, and the quality risk inspection results for the composition, material, and appearance quality are all output as "good performance", then the overall quality risk detection result of the order unit design conditions can be judged as "good performance".

[0125] Figure 13 This is a diagram illustrating the inspection results of an object inspected according to design conditions of one embodiment.

[0126] Reference Figure 13 This example demonstrates how, for each design condition, based on conditions related to order conditions, quality assurance conditions, manufacturing conditions, and quality performance information for each condition, the results of the quality risk inspection for quality assurance conditions are output, and the comprehensive judgment results are derived.

[0127] As an example, if the design condition identification ID is BD1, and the quality risk inspection results for the composition, material, and appearance quality are all output as "good", then the overall quality risk detection result of the design condition can be judged as "good".

[0128] As another example, when the design condition identification ID is BD2, although the quality risk check result for the composition and material quality is output as good performance, if the quality risk check result for the appearance quality is output as risk performance, then the overall quality risk detection result of the design condition can be judged as risk performance.

[0129] Figure 14 This is a diagram illustrating a design condition quality risk pre-detection device according to one embodiment.

[0130] Reference Figure 14 The design condition quality risk pre-detection device 1400 may include an information acquisition unit 1430 for acquiring quality performance information.

[0131] As an example, the information acquisition unit 1430 can retrieve quality performance information of products manufactured based on design conditions that have a predetermined level or higher similarity to the design conditions from a database. The quality of products previously manufactured according to registered design conditions can be measured, and the suitability of the design conditions can be determined based on the measured numerical information. This disclosure can quantify the quality of products manufactured under design conditions similar to the design conditions to be tested, thereby determining whether the design conditions to be tested are suitable.

[0132] The design condition quality risk pre-detection device 1400 may include an index calculation unit 1440 for calculating quality risk indicators.

[0133] As an example, the indicator calculation unit 1440 can calculate a first quality risk indicator based on whether the number of quality performance information pieces obtained from the information acquisition unit is N (N is an integer greater than or equal to 1), or calculate a second quality risk indicator using an artificial intelligence model. The number of quality performance information pieces used as the basis for calculating the quality risk indicator can be set by the user for convenience. If the number of quality performance information pieces obtained from the database is N or more, it means that the past quality performance data is reliable. In this case, the first quality risk indicator is calculated using the quality performance data retrieved from the database. When the retrieved data is non-existent or less than N pieces and the number is small, the second quality risk indicator can be calculated based on the data output by a trained and reliable artificial intelligence model.

[0134] As an example, the first quality risk indicator calculated in the indicator calculation unit 1440 may include a first process capability index and a first defect rate, and the second quality risk indicator may include a second process capability index and a second defect rate. The process capability index is a value used to determine whether the current production process is suitable by comparing the required level for improving the production process with the actual production results. The defect rate is a percentage representing the proportion of products produced under specific design conditions that do not meet a user-defined benchmark. This disclosure can calculate the process capability index and defect rate using information obtained from a database storing past production history or information output by an artificial intelligence model, thereby determining whether the design conditions to be inspected are good or risky.

[0135] For example, the first process capability index and the first defect rate calculated in the index calculation unit 1440 can be calculated based on the average and deviation of quality values ​​contained in the quality performance information. Furthermore, the second process capability index and the second defect rate can also be calculated based on the average and deviation of values ​​output by an artificial intelligence model, and the process capability index or defect rate can be calculated from the artificial intelligence model, thus outputting only the results.

[0136] For example, in the indicator calculation unit 1440, when the number of quality performance information obtained from the database in the information acquisition unit 1430 is N or more, a first quality risk indicator can be calculated based on the quality performance information obtained from the database; when the number of quality performance information is less than N, the above-mentioned second quality risk indicator can be calculated based on the information output by the artificial intelligence model.

[0137] For example, the artificial intelligence model used to calculate the second quality risk indicator in the indicator calculation unit 1440 may include a Bayesian neural network model.

[0138] The design condition quality risk pre-detection device 1400 may include a result output unit 1450 that outputs the quality risk detection results.

[0139] For example, the result output unit 1450 can output the quality risk detection result of the design conditions based on either the first quality risk index or the second quality risk index calculated in the index calculation unit 1440.

[0140] For example, the result output unit 1450 can output a quality risk detection result as "good performance" based on the first process capability index calculated in the indicator calculation unit 1440 being higher than the first benchmark value and the first defect rate being lower than the second benchmark value; it can output a quality risk detection result as "performance risk" based on the first process capability index being lower than the first benchmark value or the first defect rate being higher than the second benchmark value; it can output a quality risk detection result as "expected good" based on the second process capability index being higher than the third benchmark value and the second defect rate being lower than the fourth benchmark value; and it can output a quality risk detection result as "expected risk" based on the second process capability index being lower than the third benchmark value or the second defect rate being higher than the fourth benchmark value. The output being "good performance" or "performance risk" is determined based on the first quality risk indicator, meaning that the number of quality performance information items retrieved from the database is more than N as set by the user; the output being "expected good" or "expected risk" is determined based on the second quality risk indicator, meaning that the number of quality performance information items retrieved is less than N.

[0141] The design condition quality risk pre-detection device 1400 disclosed herein may consist of an information acquisition unit 1430, an index calculation unit 1440, and a result output unit 1450, and may further include a design condition extraction unit 1410 and a programming unit 1420.

[0142] Specifically, the design condition quality risk pre-detection device 1400 may include a design condition extraction unit 1410, which extracts similar design conditions from a database storing past design history based on user-defined criteria. The design condition extraction unit 1410 can extract similar design conditions from the database storing past design history according to user-defined similarity criteria. Furthermore, the extraction of similar design conditions can be categorized according to the type of design conditions, namely, order conditions, quality assurance conditions, and manufacturing conditions.

[0143] For example, the extraction of similar design conditions based on order conditions can be based on factors such as product variety, product specifications, thickness, order width, and order purpose. The extraction of similar design conditions based on quality assurance conditions can be based on factors such as the smelting composition and finished product composition required by the customer, material quality such as mechanical properties, and appearance quality such as shape, surface, and dimensions. Furthermore, each quality aspect can include sample type, test conditions, and warranty information. The extraction of similar design conditions based on manufacturing conditions can be based on information about the core design factors designed within the plant area, factory, steel output composition, steelmaking, continuous casting, and rolling processes.

[0144] The individual design conditions extracted based on similar benchmarks can be stored in a separate database.

[0145] The design condition quality risk pre-detection device 1400 may include a sorting unit 1420 (S210) for sorting out the inspection items to be detected.

[0146] As an example, the inspection target items can be determined based on at least one of the similar design conditions extracted in the design condition extraction unit 1410 and the design conditions to be inspected. Inspection target items mean that among the multiple items set as design conditions, they serve as the benchmark used to check whether the design conditions are appropriate. The inspection target items can be arranged based on the similar design conditions extracted in the design condition extraction unit 1410 or based on the actual design conditions to be inspected. When the inspection target items are determined, the relevant information for these inspection target items can be stored in a separate database.

[0147] As an example, the inspection target item may include at least one of a quality unit item, an order unit item, and a design unit item. Alternatively, depending on the inspection target item, it may include at least one of a first quality risk detection result, a second quality risk detection result for the aforementioned order unit item, and a third quality risk detection result for the aforementioned design unit item. To examine the design conditions to be inspected against multiple benchmarks to determine their suitability, this disclosure uses at least one of the quality unit item, order unit item, and design unit item to determine the inspection result. When at least one of the inspection target items is determined to be risky, the design conditions as the inspection target can be output as unsuitable.

[0148] As an example, the quality risk detection result may include a fourth quality risk detection result related to the suitability of the design conditions. This fourth quality risk detection result may include any one of the following: good performance, risky performance, good expected performance, risky expected performance, or indeterminate. For instance, if at least one of the first quality risk detection result, the second quality risk detection result, and the third quality risk detection result is a risky performance or a risky expected performance, the fourth quality risk detection result may be output as either a risky performance or a risky expected performance.

[0149] Through the aforementioned actions, accurate quality risk monitoring of the design conditions to be tested can be performed. This reduces the defect rate of products manufactured based on the design conditions to be tested.

[0150] The above description is merely illustrative of the technical concept of this disclosure. Those skilled in the art can make various modifications and variations without departing from the essential characteristics of this technical concept. Furthermore, these embodiments are not intended to limit the technical concept of this disclosure but are for illustrative purposes; therefore, the scope of this technical concept is not limited by these embodiments. The scope of protection of this disclosure should be interpreted by the claims, and all technical concepts within the same scope should be interpreted as included within the scope of the rights of this disclosure.

[0151] Cross-reference related applications This patent application claims priority to U.S. Patent Application No. 10-2023-0184030, filed in Korea on December 18, 2023, the entire contents of which are incorporated herein by reference. Similarly, any other country that claims priority for the same reasons stated above in this patent application is also incorporated herein by reference.

Claims

1. A quality risk detection method, which is a method for detecting quality risks in the design conditions of an apparatus, characterized in that, include: The acquisition step involves obtaining quality performance information of products manufactured based on design conditions that have a predetermined level or higher similarity to the design conditions from a first database. The calculation steps involve determining whether the number of quality performance information obtained is more than N, and then calculating a first quality risk indicator or a second quality risk indicator through an artificial intelligence model, where N is an integer greater than or equal to 1. as well as The output step involves outputting the quality risk detection results of the design conditions based on either the calculated first quality risk index or the second quality risk index.

2. The quality risk detection method according to claim 1, characterized in that, Further includes: The extraction step involves extracting similar design conditions from a second database that stores past design history, based on user-defined benchmarks. as well as Arrange the steps, and arrange the items to be inspected. The inspection target item is determined based on at least one of the extracted similar design conditions and the design conditions to be inspected.

3. The quality risk detection method according to claim 2, characterized in that, The inspection targets include at least one of the following: quality unit items, order unit items, and design unit items. The quality risk detection results include at least one of the following: a first quality risk detection result for the quality unit project, a second quality risk detection result for the order unit project, and a third quality risk detection result for the design unit project.

4. The quality risk detection method according to claim 3, characterized in that, The quality risk detection results include a fourth quality risk detection result related to the suitability of the design conditions. The fourth quality risk detection result includes any one of the following: good performance, risky performance, good expected performance, risky expected performance, or indeterminate. Based on the fact that at least one of the first quality risk detection result, the second quality risk detection result, and the third quality risk detection result is an actual risk or a predicted risk, the fourth quality risk detection result output is either the actual risk or the predicted risk.

5. The quality risk detection method according to claim 4, characterized in that, If at least one of the first quality risk detection result, the second quality risk detection result, and the third quality risk detection result is deemed undeterminable, the fourth quality risk detection result will output as undeterminable.

6. The quality risk detection method according to claim 2, characterized in that, The similar design conditions include order conditions, quality assurance conditions, and manufacturing conditions.

7. The quality risk detection method according to claim 1, characterized in that, The primary quality risk indicators include the first process capability index and the first defect rate. The second quality risk indicator includes the second process capability index and the second defect rate. Based on the fact that the first process capability index is higher than the first benchmark value and the first defect rate is lower than the second benchmark value, the quality risk detection result is output as "good performance". If the first process capability index is lower than the first benchmark value or the first defect rate is higher than the second benchmark value, the quality risk detection result output is the actual performance risk. Based on the fact that the second process capability index is higher than the third benchmark value and the second defect rate is lower than the fourth benchmark value, the quality risk detection result output is "expected to be good". If the second process capability index is lower than the third benchmark value or the second defect rate is higher than the fourth benchmark value, the quality risk detection result output is the expected risk.

8. The quality risk detection method according to claim 7, characterized in that, The first process capability index and the first defect rate are calculated based on the average and deviation of the quality values ​​contained in the quality performance information.

9. The quality risk detection method according to claim 1, characterized in that, In the calculation step, When there are more than N judgment results, the first quality risk indicator is calculated based on the quality performance information obtained from the first database. When there are fewer than N judgment results, the second quality risk indicator is calculated based on the information output by the artificial intelligence model.

10. The quality risk detection method according to claim 1, characterized in that, The artificial intelligence model includes a Bayesian neural network model.

11. A quality risk detection device, which is a device for detecting quality risks under design conditions, characterized in that, include: The information acquisition unit acquires quality performance information of products manufactured based on design conditions that have a predetermined level or higher similarity to the design conditions from the first database. as well as The indicator calculation unit calculates a first quality risk indicator based on whether the number of quality performance information obtained is more than N, or calculates a second quality risk indicator through an artificial intelligence model, where N is an integer greater than or equal to 1. as well as The results output unit outputs the quality risk detection results of the design conditions based on either the first quality risk index or the second quality risk index calculated.

12. The quality risk detection device according to claim 11, characterized in that, Further includes: The design condition extraction unit extracts similar design conditions from a second database that stores past design history, based on user-defined benchmarks. as well as The scheduling department schedules the items to be inspected. The inspection target item is determined based on at least one of the extracted similar design conditions and the design conditions to be inspected.

13. The quality risk detection device according to claim 12, characterized in that, The inspection targets include at least one of the following: quality unit items, order unit items, and design unit items. The quality risk detection results include at least one of the following: a first quality risk detection result for the quality unit project, a second quality risk detection result for the order unit project, and a third quality risk detection result for the design unit project.

14. The quality risk detection device according to claim 13, characterized in that, The quality risk detection results include a fourth quality risk detection result related to the suitability of the design conditions. The fourth quality risk detection result includes any one of the following: good performance, risky performance, good expected performance, risky expected performance, or indeterminate. Based on the fact that at least one of the first quality risk detection result, the second quality risk detection result, and the third quality risk detection result is an actual risk or a predicted risk, the fourth quality risk detection result output is either the actual risk or the predicted risk.

15. The quality risk detection device according to claim 14, characterized in that, If at least one of the first quality risk detection result, the second quality risk detection result, and the third quality risk detection result is deemed undeterminable, the fourth quality risk detection result will output as undeterminable.

16. The quality risk detection device according to claim 12, characterized in that, The similar design conditions include order conditions, quality assurance conditions, and manufacturing conditions.

17. The quality risk detection device according to claim 11, characterized in that, The primary quality risk indicators include the first process capability index and the first defect rate. The second quality risk indicator includes the second process capability index and the second defect rate. Based on the fact that the first process capability index is higher than the first benchmark value and the first defect rate is lower than the second benchmark value, the quality risk detection result is output as "good performance". If the first process capability index is lower than the first benchmark value or the first defect rate is higher than the second benchmark value, the quality risk detection result output is the actual performance risk. Based on the fact that the second process capability index is higher than the third benchmark value and the second defect rate is lower than the fourth benchmark value, the quality risk detection result output is "expected to be good". If the second process capability index is lower than the third benchmark value or the second defect rate is higher than the fourth benchmark value, the quality risk detection result output is the expected risk.

18. The quality risk detection device according to claim 17, characterized in that, The first process capability index and the first defect rate are calculated based on the average and deviation of the quality values ​​contained in the quality performance information.

19. The quality risk detection device according to claim 11, characterized in that, The index calculation unit is configured as follows: When there are more than N judgment results, the first quality risk indicator is calculated based on the quality performance information obtained from the first database. When there are fewer than N judgment results, the second quality risk indicator is calculated based on the information output by the artificial intelligence model.

20. The quality risk detection device according to claim 11, characterized in that, The artificial intelligence model includes a Bayesian neural network model.

Images