Process data section analysis method and system

Abstract

The purpose of the present invention is to provide a yield prediction simulation system and method that can simulate yield prediction by predicting a yield during a second cycle and reflecting a label variation amount on the basis of process operation data during a first cycle. Moreover, the invention also provides a yield prediction simulation system and a yield prediction simulation method, which can divide the first period into a plurality of sections based on the service life of the catalyst and the corresponding yield change, and perform key factor analysis, yield prediction and label variation analysis according to the different sections, thereby improving the yield prediction simulation efficiency. Yield prediction simulation is carried out in the second period, and the prediction accuracy is further improved in the whole period.

Description

Technical Field

[0001] This invention relates to a yield prediction method for a chemical process, and more specifically, to a yield prediction method and a process data processing method applicable thereto, wherein the yield during a second cycle can be predicted based on process operation data during a first cycle, reflecting label variations, and yield prediction simulation is performed. Background Technology

[0002] When implementing a series of processes consisting of multiple steps, ensuring the integrity and reliability of the organic connection between each process is crucial. To achieve this integrity, an effective process management system needs to be developed so that the input values ​​of the main equipment in the process can be used to determine whether there are any abnormalities in each process and to diagnose the causes.

[0003] Most historical data generated by industrial processes typically consists of a few variables and a linear data structure, allowing existing algorithms to provide sufficient prediction / classification results. However, advancements in ICT and sensor technologies have led to the generation of data with hundreds or even thousands of variables in the manufacturing sector. In particular, modern industrial processes, such as chemical and manufacturing processes, and power plants, strive to reduce costs and maximize profits while complying with safety, health, and environmental regulations, resulting in increasingly larger and more complex data sets.

[0004] Therefore, it is crucial to select and manage the data that has the greatest impact on profit generation from the aforementioned complex and massive datasets. For example, in commercial chemical processes, process operating conditions affect catalyst activity in the short and long term, and catalyst activity is directly related to product yield. Therefore, predicting catalyst activity from both short-term and long-term perspectives is extremely important.

[0005] From a short-term perspective, it's necessary to monitor changes in catalyst activity based on process operating conditions and optimize these conditions to enhance catalyst activity and increase product yield. From a long-term perspective, in commercial chemical processes using catalysts, the longer the process runs, the less active the catalyst becomes, leading to reduced activity. Therefore, replacement is necessary after a period of time. Since catalyst replacement is time-consuming and costly, predicting future catalyst activity and determining catalyst lifetime / replacement timing is crucial.

[0006] Therefore, new technologies need to be developed to understand and reflect the process operating conditions (key factors) that have a significant impact on catalyst activity, thereby improving the accuracy of catalyst reaction activity prediction.

[0007] (Patent Document 1) Patent Document 1: Korean Patent Publication No. 10-2018-0029114 (published on March 20, 2018) (Patent Document 2) Patent Document 2: Korean Patent Registration No. 10-2222125 (published on March 3, 2021) (Patent Document 3) Patent Document 3: Korean Patent Publication No. 10-2018-0061769 (published on June 8, 2018) Summary of the Invention

[0008] The problem the invention aims to solve The purpose of this invention is to provide a yield prediction simulation system and method, which predicts the yield during the second cycle based on process operation data during the first cycle, reflects the label variation, and simulates the yield prediction.

[0009] The purpose of this invention is to provide a yield prediction simulation system and method, which divides the first cycle into multiple segments based on catalyst lifetime and corresponding yield changes, and performs key factor analysis, yield prediction and label change analysis for each segment. Thus, yield prediction simulation is performed for the second cycle, and the prediction accuracy is further improved throughout the entire cycle.

[0010] Another object of the present invention is not limited to the above-mentioned object, and another object not mentioned can be clearly understood by those skilled in the art from the following.

[0011] Problem Solving Methods An embodiment of the present invention discloses a process data segmentation method, which is a segmentation analysis unit operating as a computer device to divide chemical process data for a given period into multiple time segments, including the following steps: calculating the inflection points of one or more first principal factors in the process data to determine the segment in one step; and integrating or splitting the segment determined in one step from one or more second principal factors in the process data to determine the segment in a second step.

[0012] In one embodiment, the first primary factor and the second primary factor may each include at least one of yield, air temperature during catalyst regeneration, selectivity, and conversion rate.

[0013] In one embodiment, the step of determining a segment at one time may include the following steps: calculating multiple inflection points of the principal factors; performing clustering on the calculated multiple inflection points; and selecting segment candidate sites from the clustered inflection points.

[0014] In one embodiment, the step of determining the segment a second time may include the following steps: calculating the average and deviation of the different segments determined in the first time, and integrating or splitting the segments determined in the first time based on the calculated average and deviation, and determining the segment a second time.

[0015] One embodiment of the present invention discloses a computer-readable recording medium containing a computer program for implementing the process data segmentation method.

[0016] Invention Effects The technical advantage of this invention is that it can predict the yield during the second cycle based on the process operation data during the first cycle, reflect the label variation, and simulate the yield prediction.

[0017] The technical advantage of this invention is that it can divide the first cycle into multiple segments based on catalyst lifetime and corresponding yield changes, and perform key factor analysis, yield prediction and label change analysis for each different segment. As a result, yield prediction simulation can be performed for the second cycle, thereby improving the prediction accuracy throughout the entire cycle.

[0018] Furthermore, this invention can screen out key factors based on process operating conditions and apply them to the prediction model, thereby improving the accuracy of predicting catalyst activity changes. Through accurate catalyst activity prediction, it can determine product sales plans and catalyst replacement timing, effectively manage product production and the time and cost required for catalyst replacement, and also grasp the changes in catalyst activity based on process operating conditions. By optimizing operating conditions, catalyst activity can be improved and product yield increased. Attached Figure Description

[0019] Figure 1 This is a block diagram of a yield prediction simulation system according to an embodiment of the present invention.

[0020] Figure 2 This is a flowchart of a yield prediction simulation method as described in one embodiment.

[0021] Figure 3 This is a flowchart of a data preprocessing method described in one embodiment.

[0022] Figure 4 This is a flowchart of a segment analysis method described in one embodiment.

[0023] Figures 5 to 10 This is a schematic diagram of the segment analysis process described in one embodiment.

[0024] Figure 11 and Figure 12 This is a schematic diagram of a data realization method described in one embodiment.

[0025] Figure 13This is a schematic diagram of the lifetime factor described in one embodiment.

[0026] Figure 14 This is a schematic flowchart of a lifetime factor calculation method according to one embodiment.

[0027] Figure 15 The yield prediction results of one embodiment are shown.

[0028] Figure 16 This is a schematic diagram of a user interface (UI) for analyzing label variation as described in one embodiment.

[0029] Figure 17 The yield prediction simulation results of one embodiment are shown.

[0030] Figure 18 This is a schematic diagram reflecting the yield prediction simulation results of the lifetime factor described in one embodiment. Detailed Implementation

[0031] An embodiment of the present invention discloses a process data segmentation method, which is a segmentation analysis unit operating as a computer device to divide chemical process data for a given period into multiple time segments, including the following steps: calculating the inflection points of one or more first principal factors in the process data to determine the segment in one step; and integrating or splitting the segment determined in one step from one or more second principal factors in the process data to determine the segment in a second step.

[0032] The above-mentioned objects, other objects, features, and advantages of the present invention can be readily understood through the following preferred embodiments related to the accompanying drawings. However, the present invention is not limited to the embodiments described herein, but is embodied in other forms. Rather, the embodiments described herein are provided to make the disclosure complete and to fully convey the spirit of the invention to those skilled in the art.

[0033] In this specification, when terms such as "first" and "second" are used to describe components, those components should not be limited to those terms. These terms are used only to distinguish one component from others. The embodiments described and illustrated herein include complementary embodiments.

[0034] Unless otherwise specified in the context of this specification, the singular form also includes the plural form. The expressions used in this specification, such as "including," "composed of," and "present as," do not exclude the presence or addition of more than one other component besides the mentioned component.

[0035] In this specification, the term "software" means the technology of running hardware on a computer; the term "hardware" means a device or machine that constitutes a computer (CPU, memory, input device, output device, peripheral device, etc.); the term "step" means a series of processes or operations connected in a time sequence to achieve a given purpose; the terms "computer program," "program," or "algorithm" mean a set of instructions suitable for processing by a computer; and the term "program recording medium" means a computer-readable recording medium that sets up and executes or records a general program.

[0036] In this specification, terms such as “~part,” “~module,” “~unit,” “~block,” and “~board” used to indicate components of the invention can mean a physical, functional, or logical unit for processing at least one function or action, which can be presented as one or more hardware or software or firmware, or as a combination of hardware, software, and / or firmware.

[0037] In this specification, "processing device," "computer," "computing device," "server device," and "server" can refer to a system having an operating system like Windows, Mac, or Linux, computer programs, memory, applications, and memory devices (e.g., HDD, SSD). For example, a computer can be a desktop computer or laptop computer, a mobile terminal, etc., but these are illustrative and not limited to. A mobile terminal can be one of the mobile wireless communication devices, such as a smartphone, tablet computer, or PDA.

[0038] The present invention will now be described in detail with reference to the accompanying drawings. Various specific details have been prepared in the process of describing the following particular embodiments to provide a more detailed description and aid in understanding the invention. However, it will be understood that those skilled in the art will readily recognize the invention even without these specific details. Furthermore, it is to be disclosed in advance that, in order to avoid confusion in the description of the invention, parts that are of common or well-known conventional art and are not closely related to the invention will not be described.

[0039] Figure 1 This is a schematic block diagram of a yield prediction simulation system according to an embodiment of the present invention. In the following description, it is assumed that the yield prediction simulation system (hereinafter referred to as the "yield prediction system" or "simulation system") of the present invention is applied to an olefin production process. For example, the simulation system of the present invention can be applied in a PDH (Propane DeHydration) process that uses propane as a feedstock to produce propylene, through which hydrogen can be extracted from propane to generate one of the olefins, namely propylene.

[0040] In one embodiment of the present invention, the yield prediction simulation system can predict the yield during a second cycle based on process operation data collected during a first cycle. In this case, the first and second cycles can be of the same duration or of different durations; in the embodiments of this specification, it is schematically assumed to be 4 years. Preferably, a cycle can be related to the lifetime of the catalyst used in the chemical process; for example, when the catalyst lifetime is 4 years, a cycle can be set to 4 years.

[0041] like Figure 1 As shown, the yield prediction simulation system 100 described in one embodiment may include a data preprocessing unit 110, a segment analysis unit 120, a data realization processing unit 130, a lifetime factor analysis unit 140, a key factor analysis unit 150, a yield prediction unit 160, a tag variation analysis unit 170, and a yield prediction simulation device 180. Each of these components (110 to 180) can be implemented as programmable software that can run in a computer device, and can be combined with firmware and software implementation parts as needed.

[0042] The data preprocessing unit 110 is a functional unit that collects and extracts data from the data storage unit 200 and performs preprocessing. (See below for reference.) Figure 3 Describe the operation of the data preprocessing unit 110.

[0043] The segment analysis unit 120 can divide a cycle into multiple segments based on preprocessed data. For example, for a cycle (e.g., 4 years), based on the changes in given factors such as process temperature and yield caused by the lifetime (aging) of the catalyst used in the process, the cycle can be divided into multiple periods. See below for reference. Figures 4 to 10 Describes the schematic operation of the section analysis unit 120.

[0044] The data realization processing unit 130 is a functional unit that realizes the data used for yield prediction, generating it into a form suitable for input into the yield prediction model. The yield prediction simulation system of the present invention uses data from a previous period (first period) to predict the next period (second period). At this time, data for the second period can be generated based on the data from the first period and input into the yield prediction model. (Refer to the following...) Figure 11 and Figure 12 Describes the illustrative operation of the data realization processing unit 130.

[0045] The aging factor analysis unit 140 is a functional unit used to reflect the aging (lifespan) of the catalyst used in the process, in order to more accurately predict the process yield. Generally, catalysts have different lifespans depending on their type, and the aging process varies throughout their lifespan. In particular, when the catalyst ages faster in the later stages, it may be difficult to accurately reflect this in the yield prediction model. Therefore, one embodiment of the present invention further considers the catalyst aging factor. For example, it can be reflected as a weighted value in the process data, which is calculated based on the catalyst aging factor over time, and its value can be input into the yield prediction model. This process can improve the yield prediction performance. The following refers to... Figures 13 to 15 Describes the schematic operation of the life factor analysis unit 150.

[0046] The key factor analysis unit 150 extracts key process factors from the preprocessed data in the data preprocessing unit 110. For example, the key factor analysis unit 150 can be implemented using a known machine learning algorithm, such as a machine learning algorithm using a feature selection method.

[0047] The yield prediction unit 160 is a functional unit that predicts process yield using preprocessed data and extracted key factors. In one embodiment, the yield prediction unit 160 is implemented as a machine learning-based learning model, which can learn a yield prediction model using preprocessed data and key factors, and then use the learned yield prediction model to predict the yield. For example, when preprocessing data from the first cycle and the first half of the second cycle, the yield prediction model can be learned using the preprocessed data and key factors, and then the yield prediction results for the remaining periods of the second cycle can be output.

[0048] The label variation analysis unit 170 is a functional unit that calculates the variation of labels (input variables of the yield prediction model). In this invention, "label" refers to the input variables of the yield prediction model, such as various operating conditions in a chemical process, such as temperature, pressure, and flow rate. Some labels (hereinafter, "control labels") can be controlled and adjusted by users (e.g., process operators or staff at the plant). By changing the value of such a control label, the label values ​​of other labels can be changed. The label variation analysis unit 170 can calculate the variation of at least a portion of the other labels caused by changes in one or more control labels. For example, it can be implemented using a machine learning algorithm that learns from process operating data stored in the data storage unit 200.

[0049] The yield prediction simulation device 180 is a functional unit that simulates yield changes based on the yield predicted by the yield prediction unit 160 and the label variation analysis results calculated by the label variation analysis unit 170. While the yield prediction unit 160 statically predicts the yield during the second period based on data stored at the current time point (data from the first period or data from the first period combined with the first half of the second period), the yield prediction simulation device 180 can simulate future yield changes when process conditions (temperature, pressure, flow rate, etc.) change. To this end, the yield prediction simulation device 180 receives both the predicted yield calculated by the yield prediction unit 160 and the analysis results calculated by the label variation analysis unit 170, and simulates yield changes accordingly.

[0050] The following is for reference. Figure 2 Describe the schematic operation of the yield prediction simulation system. Figure 2 This is a flowchart of a yield prediction simulation method as described in one embodiment. Figure 2 As shown, the yield prediction simulation method described in one embodiment includes: a step of preprocessing the yield prediction data including the data of the first cycle (S10); a segment analysis step of dividing the cycle into multiple segments based on the preprocessed data (S20); a data realization step of generating the data of the second cycle based on the data of the first cycle and the data of the first half of the second cycle and realizing the data (S30); and a lifetime factor analysis step of calculating the lifetime factor of the catalyst used in the process (S40).

[0051] Furthermore, the yield prediction simulation method includes: the step of analyzing key factors based on the segment analysis results (S50) and the step of modeling the yield prediction model based on different segments to predict the yield of the second cycle (S60).

[0052] Furthermore, the yield prediction simulation method may further include: a label change analysis step (S70) that calculates the change in the remaining labels when the user changes the label value of the control label, and a yield prediction simulation step (S80) that reflects the changed label value in the yield prediction step (S60) to predict the predicted yield and simulates the change in yield.

[0053] The following will describe it in more detail. Figure 2 The steps are shown. Figure 3 A schematic method of the data preprocessing step (S10) described in one embodiment is shown. For example... Figure 3As shown, the data preprocessing step (S10) may include: a step of preprocessing the yield prediction data in minutes (S110), a step of selecting the analysis object label (S120), a step of extracting time and daily unit data from the selected label data in the data preprocessed in minutes (S130), and a step of performing outlier processing and missing value interpolation on the daily unit data (S140).

[0054] In order to preprocess the data in step (S110), the data storage unit 200 collects and extracts the data required for yield prediction. For example, the data storage unit 200 can be implemented as a database, but the data format is not particularly limited. In one embodiment, the yield prediction data extracted by the data storage unit 200 may include: (i) process operation data of the olefin producer; (ii) laboratory data including LIMS data; (iii) plant event data including data related to periods of plant downtime; and (iv) historical yield, conversion rate, and selectivity data related to olefin production.

[0055] Process operation data for olefin producers (PDH plants) can be sensor data collected from sensors installed in various equipment (e.g., reactors, flow paths, etc.) throughout the plant. For example, these sensors can be variable sensors that monitor process operation conditions such as temperature, pressure, flow rate, and composition, and data can be collected from each sensor on a minute-by-minute basis.

[0056] Process operation data can be stored in the data storage unit 200 according to different sections, units, and tags. Here, a unit is a mid-size collection of tags within the plant, a section is a large-size collection of units, and multiple sections are grouped together to form the entire PDH plant. Additionally, tags can serve as identifiers to identify the various sensors installed in the plant. That is, each sensor is assigned a unique tag; for example, when the PDH plant has more than 9,000 sensors, an equivalent number of tags can exist. In this specification below, unless there is a particular concern about confusion, the data output by the sensor corresponding to each tag will be referred to as "tag" or "tag data".

[0057] Laboratory data may include data from a Laboratory Information Management System (LIMS). In one embodiment, for accurate yield prediction, both actual observation data (label data) and laboratory data can be used. Furthermore, when outliers or missing values ​​occur in the label data, laboratory data can be used for processing, i.e., interpolation. Alternatively, laboratory data may be omitted in other embodiments.

[0058] For example, plant event data may include data related to periods of downtime (shut-down history), major / minor overhauls, etc., which can be used to analyze and process outliers or missing values ​​in the tagged data. Historical yield values ​​include historical yield data related to olefin production. Furthermore, in addition to historical yields, conversion and selectivity values ​​may be further included. In the following description, unless there is a concern of confusion, yield, conversion, and selectivity will be collectively referred to as "yield," and from a machine learning perspective, yield (i.e., yield, conversion, and selectivity) may sometimes be referred to as the "target."

[0059] As described above, the data stored in the data storage unit 200 is continuously accumulated in units of a given set period, which can be data spanning a period or more prior to the current time. In this case, the set period unit can be seconds or minutes. For example, data can be collected in 30-second units and then converted to minutes and / or time units for analysis and storage. However, this set period unit is illustrative and not limited to a specific period. Furthermore, a period can be set in relation to the catalyst lifespan; it should be understood that in one embodiment, it could be set to 4 years, but this is also illustrative.

[0060] In step S110, the yield prediction data extracted by the data storage unit 200 is preprocessed into minute-unit data. For example, when receiving second-unit data from the data storage unit 200, it is converted into minute-unit data. When outliers or missing values ​​occur, outlier handling and missing value interpolation are performed.

[0061] Next, in step (S120), target labels for analysis are selected. For example, all label data received from all sensors installed in the factory can be used to perform key factor analysis and yield prediction, as described below. However, preferably, a portion of the label data from the entire dataset can be selected, and this selected label data can be used to perform subsequent (e.g., after step S130) time / day unit data extraction, key factor analysis, and yield prediction. In this case, target labels for analysis are selected in step (S120). For example, label selection can be based on previous research and the knowledge and experience of field technicians, selecting labels deemed helpful for the analysis.

[0062] In one embodiment, this analysis target label selection step (S120) can also be performed in advance before the step of preprocessing minute-unit data (S110), in which case minute-unit data preprocessing (S110) can be performed only for the labels selected as analysis targets.

[0063] When the target label for analysis is selected in step (S120), in the subsequent step (S130), time unit data is extracted and processed again to extract daily unit data. At this point, for data integration, factory data (label data) and LIMS data can also be applied for data integration. Furthermore, as an alternative embodiment, the target label selection step (S120) can be performed after extracting the time unit data. In this case, after extracting time unit data for all process data, only the target label for analysis can be extracted as daily unit data.

[0064] After extracting daily unit data in step (S130), step (S140) performs data preprocessing. For example, data preprocessing includes outlier handling and missing value interpolation. Outlier handling involves filtering outliers and removing or correcting them, applying only the corrected values ​​as valid input values. Furthermore, for segments that were filtered as outliers and removed, or segments lacking process data due to factory production interruptions, missing value interpolation is performed. For example, missing value interpolation is performed by generating new data through linear regression and distribution-based random number generation.

[0065] After the yield prediction data has been preprocessed through the above steps, it is organized and transformed into a learning model data format for machine learning, and then stored in data storage unit 200 or other arbitrary storage unit.

[0066] Figures 4 to 10 This is a segment analysis step described in one embodiment ( Figure 2 A schematic diagram of the illustrative method shown in S20). Figure 4 This is a schematic flowchart illustrating the steps of segment analysis. Figures 5 to 10 This is a schematic diagram of the segment analysis process described in one embodiment.

[0067] The segment analysis step (S20) divides a cycle into multiple segments based on the preprocessed data. In one embodiment, segments can be divided based on inflection points where the yield trend changes drastically or the degree of yield variation. These segments are then divided into those exhibiting similar yield increase / decrease trends during a cycle of process operation. Different segments can be applied to various modeling tasks such as key factor extraction, yield prediction, and yield prediction simulation, thereby improving the accuracy of yield prediction.

[0068] However, in the alternative embodiment, the segment analysis step (S20) can be omitted. In this case, the period does not need to be divided into multiple segments. Each period can use the key factor extraction, yield prediction, and yield prediction simulation models one by one.

[0069] like Figure 4 As shown, the segment analysis step (S20) in one embodiment may further include: selecting the main factors required for segment analysis (S210), selecting the inflection points of the main factors (S220), and determining the segments by integrating or splitting segments through variability analysis of the main factors (S230). Furthermore, in one embodiment, after step (S230), a step of confirming or modifying the segments based on catalyst design may be further included (S240).

[0070] The steps are briefly described below; first, in step (S210), the key factors required for the segment analysis are selected. For example, key factors may be selected that include at least one of the following as target values: yield, conversion rate, and selectivity, as well as the key labels that influence these target values. For this purpose, Figure 5 The main factors selected in step (S210) are illustrated schematically.

[0071] Figure 5 In this context, the baseline factor is a factor that shares a common value and exhibits certain variations across all periods, serving as a benchmark for segment classification. The cross factor, even if its values ​​differ across periods, demonstrates the same variation and can be used to define the characteristics (inflection points, variability) of different segments. The test factor possesses the characteristic of dividing periods within a cycle and can be used to test results after inflection point exploration. Of course, Figure 5 The specific types or classification criteria of the main factors shown are illustrative and may vary depending on the specific implementation of the invention.

[0072] Next, step (S220) calculates the inflection points of each primary factor, thus determining the segment in one step. Here, the "primary factor" refers to the primary factor used in the inflection point calculation step (S220), which can be... Figure 5At least some of the key factors shown in the cross factors, in one embodiment, are yield, selectivity, conversion rate, and air temperature during catalyst regeneration (Regen Air temperature).

[0073] Figure 6 This is a flowchart illustrating the specific method for calculating the inflection points of the principal factors for each primary principal factor. (Reference) Figure 6 As can be seen, the first step is to remove noise from the data of each major factor (S221). For example, the data can be smoothed to avoid identifying noise as an inflection point. For example, a 14-day moving average method can be used to achieve smoothing.

[0074] Secondly, inflection points are calculated from the main factors for noise removal. An inflection point is a point where the data value stream changes (e.g., a point where it changes from a bump to a depression). In conventional inflection point calculation methods, the point where the second derivative of the data is 0 is called an inflection point.

[0075] However, it is difficult to calculate representative inflection points using this method. Therefore, one embodiment of the present invention employs a modified plateau detection method. That is, the traditional plateau detection method is a mathematical method that finds plateau regions in a graph. In contrast, the modified plateau detection method modifies the plateau detection method that finds plateau points to find non-plateau regions and select inflection points.

[0076] In one example, Figure 7 (a) shows the results of detecting inflection points using the deformation stability detection method. Figure 6 The graph in (a) shows the air temperature during catalyst regeneration (Regen Air temperature). Here, the X-axis represents the cumulative propylene production over one cycle (or it could be another variable corresponding to time), and the Y-axis represents the temperature.

[0077] Then, step (S223) performs clustering on the calculated multiple inflection points. The purpose of clustering is to find representative inflection points by clustering multiple inflection points. The specific method is as follows: set the number of target clusters (e.g., 4) and the number of representative inflection points within each cluster (e.g., 2 in each cluster) to appropriate values. At this time, for example, representative inflection points can be selected as the inflection point with the fewest first derivative values ​​(gradients) and the largest inflection point among the inflection points in the cluster. Thus, the minimum and maximum values ​​(starting value and ending value) of the cluster characteristics can be found.

[0078] For example, clustering algorithms can employ the Density-Based Spatial Clustering of Applications with Noise (DBSCAN) method. However, DBSCAN is one of the well-known clustering methods, and of course, this invention is not limited to this method.

[0079] Figure 7 (b) illustrates the method described above for... Figure 7 The inflection point in (a) represents the clustering result. Figure 6 In (b), four clusters are formed by blue dots, yellow dots, light green dots and red dots. Two representative inflection points (white dots) are selected from each cluster, and a total of eight representative inflection points are calculated.

[0080] Secondly, step (S224) selects the final inflection point. For example, based on cumulative propylene production, one of the inflection points clustered within 100,000 tons is selected, thus ultimately choosing the most important inflection point among the representative inflection points located at the candidate site of the segment. For example, such as... Figure 8 As shown, Figure 8 (a) Represented with red dots Figure 7 The eight representative inflection points selected in (b) Figure 8 (b) shows the result of selecting four final inflection points after step (S224).

[0081] in addition, Figure 7 and Figure 8 An example of the process for selecting the inflection point of Regen Air temperature, a key factor, is shown. Multiple final inflection points are also selected from other key factors such as yield, selectivity, and conversion rate using the same or similar methods. For example, Figure 9 (a) shows the results for the four final inflection points (red dots) of the selected yield. Figure 9 (b) shows the results of selecting four final inflection points (red dots) for each selection.

[0082] In this way, the final inflection points selected from each major factor become candidate sites for the segment, thus determining the segment in one step.

[0083] Refer again Figure 4 It can be seen that step (S230) integrates or splits the segments determined once from each of the second main factors through variability analysis, thereby determining the segments a second time.

[0084] At this point, "second principal factor" refers to the principal factor used in the segment quadratic determination step (S230) based on variability analysis. Alternatively, all or part of the principal factor used in the inflection point calculation step (S220) (i.e., the first principal factor) can be used as the second principal factor. Figure 5 Other principal factors shown are used as secondary principal factors. For example, in one embodiment, yield, selectivity, and Regen Air temperature from the first principal factors can be used as secondary principal factors for variability analysis, while Figure 5 Among the main factors, conversion rate, injection gas, and reactor back-end pressure can be used to confirm the section.

[0085] Step (S230) calculates the variability (e.g., average and deviation) of different segments from the inflection point selected from the second main factor in step (S220), and integrates or splits the segments determined in the first step based on the calculated variability, thereby determining the final segments in the second step.

[0086] To consolidate and segment regions, for example, to consolidate regions with low variability and segment regions with high variability. For example, a region with low yield variability that would not be segmented, but which contains sites with high Regen Air temperature or selectivity variability, is segmented based on these sites.

[0087] Thus, based on other major factors, intervals that are not divided into segments based on any one major factor can be divided into two or more segments. Alternatively, even if any interval is divided into two or more segments based on any one major factor, segments can be integrated based on other major factors.

[0088] Users can pre-set the number of segments, that is, how many segments a cycle is divided into. For example, one cycle of the present invention can be divided into three to five segments. In one embodiment, considering catalyst design and variability within segments, one cycle can be divided into four segments.

[0089] The segmentation results determined in the above two steps can be used to proceed to the next step (e.g., Figure 2 Instead of the key factor analysis step (S50), step (S230) can be followed by a step (S240) to identify and modify the segment based on catalyst design. Step (S240) can take into account catalyst design factors such as catalyst characteristics or lifetime, identify whether the segment identified in step (S230) is suitable, and modify the segment site as needed.

[0090] Figure 10 The segmentation result finally determined through this segment analysis step (S20) is shown. For example... Figure 10 As shown, when the catalyst lifetime is considered as one cycle (e.g., 4 years), one cycle is divided into 4 segments (SG1 to SG4). Figure 10In the diagram, the red chart shows the air temperature during catalyst regeneration (Regen Air temperature), while the gray and light green charts show the yield and selectivity, respectively.

[0091] like Figure 10 As shown, the first section (SG1) involves the addition of catalyst, causing the Regen Air temperature to gradually increase and stabilize. The second section (SG2) represents the steady-state period, where the Regen Air temperature remains stable, and the yield and selectivity remain steady. The third section (SG3) shows that the yield gradually decreases with catalyst aging. That is, although this step increases the Regen Air temperature, the yield cannot be maintained or increased due to catalyst aging; instead, it gradually decreases. The fourth section (SG4) represents a step where the yield decreases sharply further; even if the Regen Air temperature is further increased, the yield and selectivity will not increase further but will decrease sharply.

[0092] As described above, the present invention can consider the yield increase or decrease trend caused by catalyst lifetime within a cycle period, and can divide the intervals showing similar trends into segments, and based on each segment, in subsequent steps (e.g., Figure 2 The key factor analysis steps (S50), yield prediction steps (S60), and yield prediction simulation steps (S80) shown are implemented separately for each different segment, and the overall analysis / prediction results are derived for one cycle, thereby improving the accuracy of yield prediction.

[0093] Now for reference Figure 11 and Figure 12 Describe the steps of data realization ( Figure 2 The illustrated method is shown in S30. The data realization step (S30) generates data for the remaining periods of the second period based on the data from the first period and the first half of the second period, and then realizes it. The yield prediction step (S60) of this invention uses data from the previous period (the first period) to predict the next period (the second period). More specifically, after generating the data for the second period based on the data from the first period, the data for the second period is input into the yield prediction model to predict the yield for the second period. At this time, if data exists for a portion of the second period (hereinafter, "the first half of the second period"), the data realization step (S30) uses the first half of the second period data to generate data for the remaining periods of the second period.

[0094] For example, consider the following premise: a cycle period is 4 years, the first cycle is from January 2017 to December 2020, and the second cycle is from January 2021 to December 2024. Assuming the current time is August 2022, the data storage unit 200 stores the process operation data of the first cycle and the data for the first half of the second cycle (i.e., from January 2021 to July 2022).

[0095] At this time, the data realization processing unit 130 generates data for the remaining period of the second cycle (i.e., from August 2022 to December 2024) based on the process operation data of the first cycle and the data of the first half of the second cycle. At this time, taking into account the trend or average characteristics of the data in the first cycle and the characteristics of the data in the first half of the second cycle, the data realization processing unit 130 can generate data for the remaining period of the second cycle.

[0096] However, due to differences in catalyst addition and initial operating conditions, the process operating data (e.g., data from various labels) for the first and second cycles tend to show different trends or values ​​between the two cycles. Therefore, it is difficult to generate data for the second cycle using the operating conditions data from the first cycle. Therefore, the data realization step (S30) of this invention corrects the data from the first cycle to make it conform to the trend of the second cycle, thereby generating the data for the second cycle.

[0097] In one embodiment, the method for realizing each tag data may include at least the average difference reflection method and the random number generation method.

[0098] When there is an average difference between the first and second periods, the average difference reflection method can be applied. In one embodiment, when the data trends in the first and second periods are similar within a given period, but the average values ​​are different, the average point of the first period's data is shifted to generate the data for the second period. For example, Figure 11 The diagram illustrates labeled data to which the mean difference reflection method can be applied. Figure 11 In the chart, the X-axis represents the time corresponding to one period, and the Y-axis represents the data value of that label. Furthermore, the black chart represents the data for the first period (CY1) of that label, and the red chart represents the first half of the second period (CY21). It should be understood that the end point of the first half of the second period (CY2) is the current time point.

[0099] Comparing the data from the first period (CY1) with the first half of the second period (CY21), we can see that the trends of the two data are similar, and the average value of the second period's data is larger. Therefore, the average difference method can be applied to increase the data for the corresponding period of the first period to the average difference, thereby generating data for the remaining periods of the second period. In one embodiment, the data for the corresponding period of the first period is increased to the average and used directly. Alternatively, at least for a portion of the intervals, the data can be modified by generating random numbers or other methods to generate the second data.

[0100] When data for a given period is incomplete or contains outliers, random number generation can be applied. In one embodiment, if incomplete data exists in the first period, random numbers are generated to produce data for the second period. For example, Figure 12 The illustration shows label data to which random number generation methods can be applied.

[0101] Figure 12 In the chart, the X-axis represents the time period, and the Y-axis shows the data values ​​for that label. The black chart represents the data for the first period (CY1) of that label, and the orange-yellow chart represents the first half of the second period (CY21). The end time of the first half of the second period (CY2) signifies the current time.

[0102] like Figure 12 As shown, the first period of data (CY1) is generated starting from a specific point in time. That is, this tag data could mean that no sensor was set up or the sensor was not working before the specific point in time. However, the second period has a first half of data (CY21), and the data for the remaining intervals (CY22) is generated by generating random numbers based on the data of the first period. For example, at the current point in time, the average and dispersion of a given interval in the past (e.g., the past 30 days (D30) of the current point in time) are calculated, and while maintaining this average and dispersion, the second half of the data (CY22) is generated by generating random numbers from the data at individual moments.

[0103] Now for reference Figures 13 to 16 Describe the steps of life factor analysis ( Figure 2The illustrated method is shown in S40. The purpose of the lifetime factor analysis step (S40) is to reflect the aging of the catalyst used in the process, and the purpose of using the catalyst in the process is to make the yield prediction simulation step (S80) more accurate in predicting the simulated yield. For example, in the PDH process for producing propylene, as the cumulative propylene production increases, the catalyst lifetime shortens, leading to a sharp drop in catalyst yield in the latter half of the process. Therefore, in order to accurately reflect the yield changes caused by changes in process conditions and the yield reduction caused by catalyst lifetime in the simulated yield prediction, it is preferable to apply the catalyst lifetime factor (aging factor) to perform the simulated yield prediction.

[0104] For example, such as Figure 13 As shown, the yellow chart represents the yield (Y1) of the first period within a cycle, while the red chart represents the lifetime factor (AF1) of the first period. It can be seen that the lifetime factor (AF1) increases or decreases in a similar trend to the yield (Y1). Furthermore, as referenced... Figure 10 The description states that in the fourth section, which is the last section of the cycle, the yield will decrease no matter how much the temperature is increased. At this point, the yield decreases relatively linearly until the third section, while in the fourth section, the yield decreases non-linearly.

[0105] Additionally, the light green chart represents the yield (Y2) up to the current time point in the second cycle, while the blue chart represents the lifetime factor (AF2) up to the current time point in the second cycle. Following a trend similar to the yield (Y2) and lifetime factor (AF1) of each first cycle, it can be predicted that the yield (Y2) corresponding to the fourth segment of the second cycle will continue to decrease non-linearly. In order to more accurately predict the simulated yield during this period, the simulated yield is calculated by reflecting the lifetime factor (AF) corresponding to the second cycle.

[0106] In one embodiment, even if the lifetime factor (AF) is not applied to the first and second segments throughout the entire second cycle, the yield prediction model can still predict the yield relatively accurately. Therefore, the lifetime factor may not be applied to the first and second segments, while it can be applied to the third and fourth segments. In another embodiment, the yield in the third segment decreases more linearly, and the yield prediction model can accurately predict the yield to a certain extent. Therefore, the lifetime factor (AF) can be applied only to the fourth segment. In yet another example, the application time point of the lifetime factor can be determined based on the cumulative propylene yield and / or the lifetime factor value. That is, the lifetime factor can be applied from the time point when the cumulative propylene yield reaches or exceeds a preset given yield, or it can be applied from the time point when the lifetime factor value falls below a preset given value.

[0107] Figure 14 This is a schematic flowchart of a lifetime factor calculation method according to one embodiment.

[0108] like Figure 14 As shown, step (S410) calculates the aging factor of the catalyst. One embodiment calculates the aging factor based on the ratio of heat applied to the catalyst to the yield. For example, as in Equation 1, the aging factor is the daily propylene yield divided by the heat applied to the catalyst.

[0109] AF (Aging Factor) = (Daily propylene production) / (Heat applied to catalyst) --- Equation 1 For example, the "heat applied to the catalyst" can be calculated by multiplying the air temperature during catalyst regeneration (Regen Air temperature) by the flow rate.

[0110] Then, step (S420) divides the calculated lifetime factor into multiple time intervals. For example, the lifetime factor can be divided into segments based on the cumulative propylene production over a typical temperature change cycle of 2 weeks.

[0111] Secondly, for each interval of the segmentation, a weighted value applicable to the simulated yield is calculated from the lifetime factor. For example, the specific method for calculating the weighted value is as follows: Figure 14 As shown, the lifetime factor for each interval can be normalized (step S430), and a weighted value can be calculated using the normalized lifetime factor (step S440). In this embodiment, step (S430) selects a representative value for the lifetime factor for each interval and normalizes it. For example, the representative value could be the average value for each interval, but this is illustrative, and any other method can be used to select a representative value for each interval. For example, after calculating the representative value for each interval, normalization is performed using the Min-Max normalization method so that the lifetime factor has a value between 0 and 1.

[0112] Then, step (S440) scales the normalized lifetime factor, calculating the lifetime factor weighting value applicable to the simulated yield. The lifetime factor weighting value is multiplied by the yield prediction simulation ( Figure 2 The simulated yield value output by the result of S80) is a value that scales the lifetime factor, taking into account the proportion by which the simulated yield is further reduced. For example, if it is considered appropriate to reduce the simulated yield by 1%, the lifetime factor weighted value will have a value between 0.99 and 1.00. That is, through step (S440), the normalized lifetime factor is converted into a value scaled between 0.99 and 1.00 (lifetime factor weighted value). However, this weighted value range is illustrative, and of course, different weighted value ranges can be set according to the specific implementation of the invention.

[0113] In one embodiment, the lifetime factor weighting value (C) is calculated according to the following formula 2.

[0114] Lifetime factor weighted value (C) = Normalized lifetime factor (A) * Response ratio (B) + (1 - Response ratio (B)) --- Formula 2 In Equation 2 above, the normalized lifetime factor (A) is the lifetime factor calculated in step (S430). The response ratio (B) is a value determined based on the scaling range of the lifetime factor weighted value (C). For example, as mentioned above, when the final simulated yield value is reduced by 1% of the simulated yield, the response ratio (B) will become 0.01. Therefore, the normalized lifetime factor (A) and response ratio (B) calculated according to each segment are input into Equation 2 to calculate the scaled lifetime factor in each segment, i.e., the lifetime factor weighted value (C).

[0115] As described above, after calculating the lifetime factor weighting value (C) through steps (S410 to S440), as described in step (S810), when performing a yield prediction simulation, the weighting value (C) is applied to the applicable range of the lifetime factor to obtain the final simulated yield. That is, the simulation is performed... Figure 2 In step (S80), starting from the time point when the given baseline is met to apply the lifetime factor, the lifetime factor is reflected in the yield. The following will refer to... Figure 18 Describe it.

[0116] Now refer to it again Figure 2 A brief description of the key factor analysis step (S50). The key factor analysis step (S50) uses data preprocessed by the data preprocessing unit (110) to extract key process factors affecting yield. For example, key factors of the second-cycle data generated in the data realization step (S30) can be extracted based on different segments split by the segment analysis step (S20). For example, the key factor extraction method can be implemented using well-known machine learning algorithms, such as the Boruta algorithm, which utilizes feature selection methods.

[0117] In the subsequent yield prediction step (S60), the key factors extracted in step (S50) can be applied to the yield prediction model. For example, step (S50) selects more than 100 key factors from the plant's more than 9,000 labels, and then, in the yield prediction step (S60), yield prediction is performed based on the label data values ​​of the selected more than 100 key factors. Furthermore, in one embodiment, these key factors can also be applied in the label variation analysis step (S70).

[0118] like Figure 2As shown, the yield prediction step (S60) uses preprocessed data and key factors to predict the process target (at least one of yield, conversion, and selectivity). For example, based on the different segments divided in the segment analysis step (S20), the yield of the second cycle can be predicted using the second cycle data generated in the data realization step (S30).

[0119] For example, yield prediction models can integrate bagging algorithms such as random forests and boosting algorithms such as XGBoost and LightGBM (LGBM) to adopt yield prediction models with high predictive power, using the first period's data as learning data for the model to learn.

[0120] The bagging series algorithms employ parallel, iterative random sampling and statistical analysis, which enhances the learning data. Therefore, even with insufficient learning data, sufficient learning can be achieved, preventing underfitting and overfitting. The boosting series algorithms also perform multiple random samplings, but sequentially rather than in parallel. This allows for learning by adjusting the weighting of the next learning data based on previous learning outcomes. Specifically, incorrect answers are assigned higher weights, resulting in higher accuracy. Preferably, the bagging series algorithms, which are widely used to prevent overfitting and underfitting, employ the random forest algorithm or error model widely applied in them, while the boosting series algorithms, used to improve accuracy, employ the LGBM model. However, this yield prediction model is illustrative and, of course, does not utilize well-known and suitable machine learning methods according to the specific embodiments of the invention. Figure 15 The diagram illustrates the configuration of a screen representing the yield prediction results using the aforementioned ensemble model. Figure 15 The predicted yield (Y) during the second period is shown in an orange-yellow chart.

[0121] refer to Figure 16 Describe the steps of label variation analysis ( Figure 2 As shown in S70). The label variation analysis step (S70) uses a label variation analysis model to analyze the label variation. This label variation analysis model calculates the variation of at least a portion of the remaining labels based on the variation of one or more control labels in the yield prediction data.

[0122] In one embodiment, at least one operable control tag is selected from the tags used for yield prediction. Here, "control tag" means a tag in the process conditions of the chemical process that can be operated by the user; for example, it may include at least one tag among catalyst regeneration air temperature, charge heater temperature, catalyst regeneration air flow rate, and reactor feed flow rate. Therefore, for example, it can be understood that when yield prediction uses 100 factors (tags), the four tags become control tags, and at least a portion of the remaining 96 tags become tags that change according to changes in the four control tags.

[0123] The label variation analysis unit 170 can calculate the variation of at least a portion of the remaining labels based on the variation of one or more control labels. For example, the segment analysis step can calculate the variation of at least a portion of the remaining labels based on the variation of one or more control labels. For example, based on the different segments split by the segment analysis step (S20), the second period data generated by the dataset realization step (S30) and the key factor analysis (S50) are used to perform label variation analysis (S70) on the selected labels.

[0124] In response, Figure 16 A schematic user interface (UI) for analyzing label variations is shown. Figure 16 In the illustrated embodiment, it should be understood that four labels—Regen Air temperature, ChargeHeater temperature, Regen Air flow rate, and ReactorFeed flow rate—are used as control labels. For example, an arrow button (10) is displayed to the right of the Regen Air temperature label, allowing the operator to adjust its value to increase or decrease. The other three control labels are also adjusted using the same arrow button. Furthermore, as described above, when the operator adjusts at least one of the four control labels to any value and then presses the "Start Analysis" button (20) (e.g., by clicking with a mouse), the label change analysis model can calculate and output the changes in the remaining labels based on the changes in the control labels.

[0125] Furthermore, as described above, the label variation analysis results obtained in the label variation analysis step (S70) can be used as input variables in the subsequent yield prediction simulation step (S80). The yield prediction simulation step (S80) simulates changes in the yield based on the yield predicted by the yield prediction unit (160) and the label variation analysis results calculated by the label variation analysis unit (170), and outputs a simulated yield. That is, based on the second-cycle data generated by data realization (S30) in the yield prediction step (S60), the future yield is statically predicted. Conversely, when the user changes the control label to an arbitrary value through the label variation analysis (S70), thereby changing the values ​​of the other labels, the yield prediction simulation step (S80) predicts how the future yield will change. Therefore, in the yield prediction simulation step (S80), the yield prediction simulation (180) inputs the predicted yield calculated by the yield prediction unit (160) and the analysis results calculated by the label variation analysis unit (170) into the yield prediction simulation model to calculate the simulated yield.

[0126] A yield prediction simulation model can be implemented through machine learning algorithms. For example, it can integrate bagging algorithms such as random forest and boosting algorithms such as XGBoost and LGBM to adopt a yield prediction model with high predictive power.

[0127] Figure 17 An illustrative yield prediction simulation result is shown in one embodiment. Figure 17 In the chart, the orange-yellow graph represents the predicted yield (Y) from the yield prediction step (S60), which is consistent with... Figure 15 The yield chart shown is the same. Figure 17 In the diagram, the pink chart represents the simulated yield (“simulated yield”) (Ys) implemented in the yield prediction simulation step (S80). That is, the yield that the staff simulated in the process... Figure 16 The screen changes at least one control tag value, and thus, with the remaining tag values ​​also changing, the simulated yield is displayed. In this way, the present invention not only simply uses second-cycle data to perform static yield prediction, but also dynamically simulates and displays whether the yield changes based on the changes in the main control tag when the main control tag is changed. This allows for more accurate and precise yield analysis and prediction based on changes in the main control tag.

[0128] Furthermore, in one embodiment, when calculating the simulated yield, the lifetime factor can be applied as a weighting value to a given period of the second cycle (e.g., the third and fourth segments of the second cycle, or the fourth segment, etc.).

[0129] like Figure 14The step shown (S810) calculates the lifetime factor weighting value (C) through steps (S410 to S440). Then, when performing the yield prediction simulation, the weighting value (C) is applied to the lifetime factor applicable range to obtain the final simulated yield.

[0130] For example, in this step (S810), the final simulated yield (Ysf) is calculated according to the following formula 3.

[0131] Final simulated yield (Ysf) = Simulated yield (Ys) * Weighting value (C) --- Equation 3 In the above formulas, the simulated yield (Ys) refers to the simulated yield calculated based on the predicted yield of the yield prediction step (S60) and the label variation analysis results of the label variation analysis (S70). The weighting value is the lifetime factor weighting value (C) calculated according to formula 2.

[0132] Figure 18 A simulated yield graph reflecting the lifetime factor is shown. Figure 18 In the chart, the horizontal axis represents the cumulative propylene yield, and the vertical axis represents the simulated yield. In the view, the black chart represents the yield without simulation, i.e., the predicted yield output by the yield prediction step (S60). The red solid line chart represents the simulated yield output by the yield prediction simulation step (S80) when the temperature input in the control label is increased, and the blue solid line chart represents the simulated yield output when the temperature input is decreased; this chart does not reflect the various lifetime factors.

[0133] In one embodiment, the lifetime factor can be applied from the point in time when both the condition that the cumulative propylene production reaches a given production (P1) or higher and the condition that the lifetime factor value is below a given value (AF1) are met. For example Figure 18 In the simulation, the heating simulation starts from the point when the cumulative yield of the simulated yield (red chart) reaches P1. As shown in Equation 3, the simulated yield (Ys) is multiplied by the weighting value (C) to output the final simulated yield (Ysf) (red dashed line). The cooling simulation starts from the point when the lifetime factor reaches P2, and the weighting value (C) is applied to output the final simulated yield (Ysf) (blue dashed line). In this way, the catalyst lifetime can be taken into account, and a simulation prediction that is closer to the actual yield can be achieved.

[0134] As described above, those skilled in the art can make various modifications and variations based on the description in this specification. Therefore, the scope of this invention should not be limited to the described embodiments, but should be defined by the following claims and their equivalents.

[0135] Figure Labels 100: Yield prediction simulation system; 110: Data preprocessing unit 120: Section Analysis Department; 130: Data Realization Processing Department 140: Life Factor Analysis Department; 150: Key Factor Analysis Department 160: Yield Prediction Department; 170: Label Variation Analysis Department 180: Yield prediction simulation device; 200: Data storage unit [Potential for Industrial Applications] This invention relates to a yield prediction method for a chemical process, and more specifically, to a yield prediction simulation method that can predict the yield during a second cycle based on process operation data during a first cycle, reflecting label variations, and simulating the yield prediction.

Claims

1. A method for segmenting process data, characterized in that: A method for a segment analysis unit (120) operating as a computer device to divide chemical process data for a given period into multiple time segments includes the following steps: Step S220: Calculate the inflection points of one or more first principal factors in the process data, and determine the segment in one step; and Step S230: From one or more second primary factors in the process data, integrate or split the segments determined in the first step, and determine the segments again.

2. The process data segmentation method according to claim 1, characterized in that: The first and second major factors respectively include at least one of yield, air temperature during catalyst regeneration, selectivity, and conversion rate.

3. The process data segmentation method according to claim 1, characterized in that: Step S220 of determining a segment at one time includes: Step S222: Calculate multiple inflection points of the principal factors; and Step S223: Perform clustering on the calculated multiple inflection points.

4. The process data segmentation method according to claim 3, characterized in that: Step S220, which determines the segment at one time, includes the following steps: Step S224: Select candidate sites for segments from the inflection points of the cluster.

5. The process data segmentation method according to claim 3, characterized in that: Step S222, which calculates the multiple inflection points, uses the deformation stability detection method to detect the inflection points of each major factor.

6. The process data segmentation method according to claim 1, characterized in that: In step S230 of the secondary segment determination, the average and deviation of the different segments determined in the first determination are calculated, and the segments determined in the first determination are integrated or split based on the calculated average and deviation, and the segments are determined in the second determination.

7. The process data segmentation method according to claim 1, characterized in that: The process further includes step S240: this step modifies the secondary determined segment based on the characteristics or lifetime of the catalyst.

8. A computer-readable recording medium containing a computer program for implementing the process data segmentation method according to any one of claims 1 to 7.

Images