An optimization system for sample delivery on a factory and product basis

Abstract

The invention is related to an artificial intelligence-enabled system optimizing the delivery frequency and quantity of a sample required for quality tests on the basis of a product (U) and factory (F) by analyzing the quality measurement data and quality test result data collected from the quality unit (K) and R&D units (A) in factories (F).

Description

[0001] DESCRIPTION

[0002] AN OPTIMIZATION SYSTEM FOR SAMPLE DELIVERY ON A FACTORY AND PRODUCT BASIS

[0003] Technological Field:

[0004] The invention relates to a system optimizing the delivery frequency and quantity of sample required for quality tests by analyzing quality data collected from quality and research and development (R&D) units in factories.

[0005] State of the Art:

[0006] Quality measurements are carried out for the quality control of the products in the factories, and detailed quality tests are carried out in the R&D units. For this, a certain number of samples are taken from each batch of products produced in the factory, and measurements and tests are carried out on these samples.

[0007] In factories, the number and frequency of samples taken during production are important. If the number of samples taken is small, then the quality control tests will not be carried out efficiently, and as a result, the probability of defects in the products transferred to a customer will increase. In today's world, thanks to the global communication tools such as social media platforms and the Internet, customer satisfaction or dissatisfaction may be easily heard by everyone. Therefore, businesses frequently collect samples for the quality control tests. The high number of samples taken causes waste of quality control resources, loss of labor and extra costs. Therefore, in the state of the art, there is a need for an optimization system for the number of samples to be taken from the products.

[0008] The U.S. Patent Document No. US6636818, which was encountered in the literature search, discloses a system to determine whether the products are produced at a desired quality level. The main object of said U.S. patent is to determine whether the products are produced in a desired quality. However, the invention of the specification calculates the optimum number of samples for the quality tests and measurements.

[0009] The European Patent Document No. EP3553612, which was encountered in the literature search, discloses a method and system for determining the sample size. Although the European patent document discloses the determination of the sample size, it is not obvious what the cluster from which that sample was taken and what data or calculations were used to obtain that sample. According to the patent, a user selects a type of sampling plan, and the system outputs the sampling size to the user via a device and interface. In the invention of the description, the quality measurement and quality test data are analyzed, and then the optimum number of samples to be taken is calculated by calculating the quality variation and the correlation coefficient data.

[0010] The European Patent Document No. EP0086059A2, which was encountered in the literature search, describes computer-aided improvements for use in sampling procedures. The European patent document does not describe exactly how the sample size is calculated. When an operator enters the number of products in the batch into a computer together with the codes such as LII code and a K code, the number of samples is obtained as a result.

[0011] The Chinese Patent Document No. CN1 15906399A, which was encountered in the literature search, discloses a system for estimating the quality processes of the products. Production quality is estimated in the D4 document. There is no effort to calculate the optimal number of samples.

[0012] As can be seen, a system calculating the optimum number of samples by converting the quality tests and quality measurements into a score data does not exist in none of the systems in the state of the art. Consequently, there is a need for a sample optimization system that goes beyond the state of the art and eliminates the disadvantages thereof.

[0013] Brief Description of the Invention:

[0014] The invention is a sample optimization system that goes beyond the state of the art, eliminates the disadvantages thereof, and contains additional advantages.

[0015] The object of the invention is to provide a system which performs a correlation analysis for the quality measurements and quality tests conducted in quality and R&D units for products manufactured in factories, and consequently, calculates the optimum number of samples delivered to the quality tests by a calculation process. Another object of the invention is to prevent fewer samples than required from being delivered to the quality tests.

[0016] Another object of the invention is to prevent more samples than required from being delivered to the quality tests.

[0017] In the invention, the optimum number of samples delivered to the quality control tests carried out by R&D and similar units is automatically determined by the system. In this way, the sale and shipment of the defective products caused by delivering few samples to the quality test is prevented, thereby increasing the customer satisfaction. Likewise, problems such as loss of labor, inefficient use of quality resources, increased cost and waste caused by delivering extra samples to the quality testing are avoided.

[0018] The advantages of the invention are described below:

[0019] • Resource Efficiency: Time and cost savings are achieved by optimizing the sample delivery processes.

[0020] • Improvement of Quality Processes: The accuracy of the factories' own quality control processes is checked more effectively.

[0021] • Scalability: The system may be applied to a wide range of products and different production processes.

[0022] Description of the Figures:

[0023] The invention will be explained with reference to the attached figures, so that the features of the invention will be understood more clearly. However, this is not intended to limit the invention to these particular embodiments. On the contrary, it is intended to cover all alternatives, modifications and equivalents of the invention which may be included within the scope of the invention as defined by the appended claims. It should be understood that the details shown are only for illustration of the preferred embodiments of the present invention, and are presented for the purpose of providing the most useful and easy-to- understand description of both the embodiment of the methods and the rules and conceptual features of the invention. In the figures:

[0024] Fig. - 1 is a schematic view of the system of the invention. The figures which will help to understand this invention are numbered as indicated in the attached drawings and are given below with their names.

[0025] Description of the References:

[0026] 10. Calculation module

[0027] 20. Server

[0028] 30. Database

[0029] 40. First data collection module

[0030] 50. Second data collection module

[0031] 60. Artificial intelligence module

[0032] A. R&D unit

[0033] F. Factory

[0034] K. Quality unit

[0035] U. Product

[0036] Description of the Invention:

[0037] In this detailed description the sample optimization system of the invention is described with examples only for a better understanding of the subject in a non-limiting sense.

[0038] As it is known, products (U) are produced in factories (F), quality measurements are carried out in quality units (K), and quality tests are carried out in R&D units (A). While the quality measurements include size, color and similar measurements made for each product (U) produced, the quality tests include more detailed and complex examinations requiring manpower which vary from product to product, such as lifetime, tensile or impact strength, corrosion resistance and the like. Therefore, it is not possible to apply quality control tests to all products (U) produced in a factory (F). The system of the invention optimizes how many of a batch of products (U) produced in factories (F) should be delivered to the R&D unit (A) as the samples for a quality test. Said product (U) herein includes a range of the product (U) which may be measured and tested for all kinds of quality measurements and quality tests such as white goods, machinery and similar products produced in factories. While an R&D unit (A) is a unit where the detailed quality control tests are performed, a quality unit (K) defines a unit where simpler quality measurements are carried out on products (U) than the quality control tests. A factory (F), on the other hand, means all facilities where the production is carried out in an enterprise, including different branches.

[0039] In the invention, a large number of data is processed during the determination of the optimum number of samples to be delivered to the R&D unit (A). The most important one of these data is the quality measurement data (K) from the quality unit (K) of a factory (F) or one of the branches of the factory (F) and the quality test result data from the R&D unit (A). In the invention, the quality measurement data and quality test result data are processed by a correlation analysis, resulting in a correlation coefficient. A correlation analysis is a statistical method which provides information on the relationship between the variables, the orientation of that relationship, and the severity of that relationship. The higher the correlation in the analysis, the less often that factory (F) will deliver samples; if the correlation is low, the delivery frequency and quantity of a sample will be increased. If the quality measurements performed in the factory (F) and the results of the quality tests are compatible with each other, that is, in case that the correlation is high, it is possible to apply fewer quality tests by taking fewer samples. Currently, the quality measurements provide sufficient results. However, if the quality measurements performed in the factory (F) and the results of the quality tests are incompatible with each other, that is, in case that the correlation is low, more samples should be taken for the quality testing and the correlation should be corrected. The invention obtains a correlation coefficient ranging from 0 - 1 by processing the quality measurement data obtained from the quality unit (K) of the factory (F) or one of the branches of the factory (F) and the quality tests result data obtained from the R&D unit (A) with the correlation analysis, and uses this score during optimizing the number of samples to be delivered to the R&D unit (A) for the quality testing. The most important aspect in the invention is the acquisition of the correlation coefficient. Said correlation coefficient inherently differs for each product (U) produced in the factory (F) and for each factory (F) owned by an enterprise, or for each branch of a factory (F). In this way, the invention provides sample optimization on the basis of a factory (F) and product (U). Here, the product (U) may be itself, or the sub-parts which make up the product (U).

[0040] In the invention during obtaining the correlation coefficient, the quality test results data performed in the R&D unit (A) for the samples of a product (U) for sample optimization of the factory (F) are collected by a first data collection module (40). Likewise, the quality measurement data performed in the quality unit (K) for a product (U) for sample optimization of the factory (F) are collected by a second data collection module (50). The data collected by the first and second data collection modules (40, 50) are transferred to a server (20) by wired or wireless data transferring methods and stored in a database

[0041] (30). In the invention, a calculation module (10) running on the server (20) processes the measurement data and quality test results data, which are stored in the database

[0042] (30), with the correlation analysis, and determines the similarities and differences therebetween by comparing the data. As a result of this comparison, the calculation module (10) assigns a correlation coefficient ranging from 0 to 1 with the data of a product (U) and the factory where it is produced (F). After this stage, an artificial intelligence module (60) evaluates the correlation coefficient, outputs an increase in the number of samples delivered for the quality test in the factory (F) if the correlation coefficient is below the specified limit value, and outputs a reduction in the number of samples delivered for the quality test in the factory (F) if the correlation coefficient is below the specified limit value.

[0043] Alternatively, in the invention, the correlation coefficient, the primary number of samples, the quality variation of the product, the number of critical parameters for the product, and the past performance score of the factory (F) are also used as determining factors during the determination of the sample optimization of the product (U).

[0044] In the invention, the primary number of samples (B) indicates the minimum number of samples, that is, the minimum number of samples which should be delivered by the factory (F) for a particular product. For example, at least 2 samples should be taken for each batch of product (U). This value is determined by the factory (F) managers.

[0045] The higher the correlation coefficient (K), which was mentioned earlier in the invention, the lower the number of samples will be. If K = 1 , the number of samples is minimal.

[0046] In the invention, the quality variation (V) of the product is calculated according to the data of the R&D unit (A) center. If the quality variation of the product is high, more samples need to be delivered. In other words, if the value (V) is high, the number of samples required increases. Variation is a value which is directly proportional to the number of tests required to measure the quality of the product. In the invention, the number of the critical parameters (P) are the critical quality requirements followed regarding the product (U). The number of the critical parameters is obtained from the quality unit (K). As the number of the critical parameters (P) increases, the number of samples which need to be delivered for quality test inherently increases.

[0047] In the invention, the past performance (G) of the factory is a value ranging from 0 to 1 , and if it is good, the number of samples reduces. If G is low, the number of samples increases. The historical score of the factory (G), and the incoming complaint data about the factory (F) and the product (U) are taken into consideration.

[0048] In the invention, the number of samples (N) will be calculated as a result and will be output by the calculation module (10) in the system.

[0049] In the invention, the calculation module (10) uses the formula N=B*(1-K)*V*P(1 -G). Herein, as the correlation coefficient (K) increases, the number of samples (N) which need to be delivered to the quality test decreases, so the quality coefficient (K) is placed in the formula so as to be inversely proportional to the result. Likewise, if the past performance of the factory (G) is low, the number of samples (N) increases. Therefore, the past performance of the factory (G) is placed in the formula so as to be inversely proportional to the result. Other parameters are directly proportional to the number of samples (N).

[0050] Exemplary Calculation:

[0051] Let's say for a combination of factory (F) and product (U):

[0052] • B = 5 (Primary number of samples)

[0053] • K = 0.7 (Correlation coefficient)

[0054] • V = 2 (Quality variation)

[0055] • P = 5 (Number of critical parameters)

[0056] • G = 0.6 (Factory's performance score)

[0057] In this case, the number of samples is calculated as follows:

[0058] N=B*(1-K)*V*P(1 -G) = 5x(1-0.7)x2x5(1-0.6) = 5x0.3x2x5x0.4 =6 N=6 samples required. In this case, in the invention, the artificial intelligence module (60) outputs an increase in the number of samples (N) by one. If the result is fractional, the correct solution is to round it to a higher integer.

[0059] It is possible that many additional features may be found in the invention as an alternative. For example, machine learning algorithms may be used to better optimize the delivery number of samples. By analyzing the historical data, the system may predict the future sample requirements. In particular, it may learn how the correlation coefficient changes over time and which parameters affect the quality more. This feature indicates that the system is not only based on a static formula, but also has a dynamic and learning structure. Thus, more accurate sample delivery plans may be provided over time.

[0060] Alternatively, the system may monitor sudden changes in quality parameters in real time and give instant alerts if the limits are exceeded, as it monitors the quality measurement and quality test results data in both the quality unit (K) and the R&D unit (A). For example, when a factory (F) experiences an abnormal deviation, the system may automatically give a warning and request the factory (F) to deliver samples more frequently. Here, anomaly detection is provided by the artificial intelligence module (60) in the system. This feature indicates that the system is proactive and that problems may be detected and intervened early. This minimizes costs and quality errors.

[0061] Alternatively, in the invention, the system uses a blockchain technology to ensure the security of the quality data used in sample delivery processes and to increase the traceability of data. Thus, the data shared between factories and R&D centers are guaranteed to be unchangeable and reliable. In the invention, the server (20) processes the quality measurement and quality test results data in both the quality unit (K) and the R&D unit (A) with the blockchain technology while recording them in the database (30). The lockchain technology provides differentiation in the industry by increasing the reliability and transparency of the system. In particular, the secure storage of the precise quality data may create a great competitive advantage.

[0062] In another alternative of the invention, the system may be integrated with the autonomous sensors and loT devices over the production lines to collect the quality data directly. These devices may work integrated with the system to instantly determine the sample requirements and start the delivery process. This feature allows the system to have a wider industrial usage area. Instead of manually collecting the sample data, an integration with the smart devices both reduces the margin of error and accelerates the processes.

[0063] In another alternative to the invention, the system includes an estimation and simulation module which may simulate different sample delivery scenarios. For example, it may simulate in advance the consequences of changes in the correlation coefficient or quality variations in the factory and create optimized strategies accordingly. Said estimation and simulation module provides the ability to evaluate possible future scenarios and determine the optimal sample strategy in advance. Thus, the system becomes a tool for the strategic decisions.

[0064] In a different alternative of the invention, a feedback loop may be formed based on the test results for the samples delivered by the factories (A). If the sample test results meet the factory's expectations (A), the system automatically gives a feedback and optimizes the sample frequency. This ensures the system to operate in an active feedback loop with the factories and allows for continuous improvement of the quality control processes.

[0065] In a different alternative of the invention, it is possible to customize algorithms specifically for different sectors and product types. For example, an algorithm used for an automotive factory (F) may differ from an algorithm in a food factory (F) and may be optimized for specific quality parameters. This feature allows the system to be applied flexibly in every sector, so that the system may appeal to a wide range of markets.

Claims

CLAIMS1 . A system optimizing the number of samples which need to be delivered for a quality test per the manufactured product (U) in factories (F) or branches thereof, having a quality unit (K) which performs a quality measurement for each product (U) and an R&D unit (A) which performs a quality test for the products (U) for which the samples are delivered, in which various products (U) are produced, characterized in that it comprises:• at least one first data collection module (40) which collects the quality measurement data from the quality unit (K) in the factory (F) in which a product (U) is manufactured and transfers it to a database (30),• at least one second data collection module (50) which collects the quality test result data from the R&D unit (A) in the factory (F) in which a product (U) is manufactured and transfers it to a database (30),• at least one database (30) which stores said quality measurement data and quality test result data,• at least one calculation module (10) which calculates a correlation coefficient ranging from 0 - 1 by processing the quality measurement data and quality test result data in said database (30) with a correlation analysis,• at least one artificial intelligence module (60) which outputs a reduction in the number of samples which need to be delivered to the quality test as the correlation coefficient calculated with said calculation module (10) converges to 1 , or to increase the number of samples which need to be delivered to the quality test as the correlation coefficient divergences from 1 ,• at least one server (20) running said artificial intelligence module (60) and the calculation module (10).

2. A system according to Claim 1 , characterized in that it comprises:• at least one calculation module (10) which calculates a quality variation value based on the quality test result data obtained from the R&D unit (A) center,• at least one artificial intelligence module (60) which increases the number of samples which need to be delivered to the quality test as the qualityvariation value increases, and reduces the number of samples which need to be delivered to the quality test as the quality variation value reduces,• at least one server (20) running the artificial intelligence module (60) and the calculation module (10).

3. A system according to Claim 1 , characterized in that it comprises:• at least one calculation module (10) which calculates the number of critical parameters required for the quality measurement of the product (U) based on the quality test result data obtained from the R&D unit (A) center,• at least one artificial intelligence module (60) which increases the number of samples which need to be delivered to the quality test as the number of critical parameters increases, and reduces the number of samples which need to be delivered to the quality test as the number of critical parameters reduces,• at least one server (20) running the artificial intelligence module (60) and the calculation module (10).

4. A system according to Claim 1 , characterized in that it comprises:• at least one calculation module (10) which determines the ahistorical score of the factory so as to vary between 0 and 1 based on the incoming complaint data about the factory (F) and the product (U),• at least one artificial intelligence module (60) which outputs a reduction in the number of samples which need to be delivered to the quality test as the historical score of the factory converges to 1 , or increase the number of samples which need to be delivered to the quality test as the historical score of the factory divergences from 1 ,• at least one server (20) running the artificial intelligence module (60) and the calculation module (10).

5. A system according to one of the preceding claims, characterized in that it comprises:• at least one calculation module (10) which calculates the number of samplesat least one server (20) running said calculation module (10).

6. A system according to Claim 1 , characterized in that it comprises at least one server (20) which records the quality measurement and quality test results data collected from both the quality unit (K) and the R&D unit (A) in the database (30) while processing it with the blockchain technology.

7. A system according to Claim 1 , characterized in that it comprises autonomous sensors and loT devices to be able to directly collect the quality measurement data and quality test results data of the factory (F).

8. A system according to Claim 1 , characterized in that it comprises an estimation and simulation module which may simulate different sample delivery scenarios by analyzing the quality test result data from the R&D unit (A) center, the quality test result data from the R&D unit (A) center, and the sample number data calculated by the calculation module.

9. An operation method for a system optimizing the number of samples which need to be delivered for a quality test per the manufactured product (U) in factories (F) or branches thereof, having a quality unit (K) which performs a quality measurement for each product (U) and an R&D unit (A) which performs a quality test for the products (U) for which the samples are delivered, in which various products (U) are produced, characterized in that it comprises the steps of:• collecting the quality measurement data from the quality unit (K) in the factory (F) in which a product (U) is manufactured, by a first data collection module (40),• collecting• the quality test result data from the R&D unit (A) in the factory (F) in which a product (U) is manufactured, by a second data collection module (50),• processing of said quality measurement data and quality test result data with a correlation analysis by a calculation module (10),• assigning a value ranging from 0 - 1 to the correlation coefficient by the calculation module (10),• outputting a reduction in the number of samples which need to be delivered to the quality test as the correlation coefficient calculated by an artificial intelligence module (60) converges to 1 ,• outputting an increase in the number of samples which need to be delivered to the quality test as the correlation coefficient calculated by an artificial intelligence module (10) divergences from 1 ,10. A method according to Claim 9, characterized in that it comprises the steps of:• calculating quality variation value by a calculation module (10), based on the quality test result data obtained from the R&D unit (A) center,• by an artificial intelligence module (60), increasing the number of samples which need to be delivered to the quality test, as the quality variation value increases and reducing the number of samples which need to be delivered to the quality test as the quality variation value reduces.11 . A method according to Claim 9, characterized in that it comprises the steps of:• calculating the number of critical parameters required for the quality measurement of the product (U) based on the quality measurement data obtained from the quality unit (K) center, by a calculation module (10),• by an artificial intelligence module (60), increasing the number of samples which need to be delivered to the quality test, as the number of the critical parameters increases and reducing the number of samples which need to be delivered to the quality test as the number of the critical parameters reduces.

12. A method according to Claim 9, characterized in that it comprises the steps of:• determining the historical score of the factory by a calculation module (10), so as to vary between 0 and 1 based on the incoming complaint data about the factory (F) and the product (U),• by an artificial intelligence module (60), reducing the number of samples which need to be delivered to the quality test as the historical score of the factory converges to 1 , or increasing the number of samples which need to be delivered to the quality test as the historical score of the factory divergences from 1 .

Images