Process data processing device and process processing method

Abstract

The process data processing device of the present invention includes: a sensor unit that detects the physical magnitude value relative to the time during process progression; a process data recording unit that, in conjunction with the sensor unit, receives input of the physical magnitude value relative to time each time the process schedule progresses, records the input physical magnitude value relative to time as process data, and accumulates and records the process data for each process schedule; and a calculation unit that selects at least one excellent process data from the process data by a pre-set excellent process selection calculation, and sets pre-set range values ​​for the selected excellent process data using the upper and lower critical values ​​of the process data.

Description

Technical Field

[0001] The present invention relates to a process data processing apparatus and a process processing method, and more particularly to a process data processing apparatus and a process processing method for processing process data input from various sensors.

Background Art

[0002] In semiconductor manufacturing processes, various sensors are installed to receive input of process data, and the semiconductor process is monitored and controlled based on the input process data.

[0003] Such process data receives input of temperature, pressure, voltage, current, electromagnetic field, distance, etc. from respective sensors. However, such process data is mixed with impulses and noise, and such impulses and noise tend to generate frequent warning alarms for operators, inducing fatigue.

[0004] When frequent warning alarms are generated due to such noise, for example, assuming there is process data regarding time-contrast pressure in the process chamber during the semiconductor manufacturing process as shown in FIG. 1, no warning alarm is generated in the process section (1a) where the pressure in the chamber hardly fluctuates.

[0005] However, in a section (1b) where noise and impulses are instantaneously generated, such as when gas is injected into the process chamber or the power switch for generating plasma is turned on, a large input value is instantaneously input together with noise, causing the internal pressure value to fluctuate greatly.

[0006] During this process, if the pressure increases above a certain level, the pressure monitoring system generates a warning alarm. However, because these warning alarms occur at every stage of the process, they cause significant fatigue among workers. In particular, frequent warning alarms can lead to distrust among workers, causing them to ignore the alarms and fail to take corrective action when actual problems occur.

[0007] However, because there is no standard procedure for setting the upper and lower critical limits to generate warning alarms on the process data, these limits are not determined based on the process data itself, but rather depend on the operator's knowledge. This leads to qualitative differences in statistical process control techniques between skilled and unskilled operators, making it impossible to manage the system stably and uniformly.

[0008] Furthermore, the process of setting upper and lower critical limits for process data warning alarms must be carried out through consultation and obtaining the consent of all workers responsible for managing the process, ensuring that everyone understands and agrees to the process.

[0009] As a result, after setting the upper and lower critical values ​​for process data through consultation, the upper and lower critical values ​​for warning alarms are set to be very broad in order to avoid having to go through the cumbersome consultation process again. Because the upper and lower critical values ​​for process data are set to be very broad in this way, many meaningful warning alarms are ignored.

[0010] Consequently, even if upper and lower critical values ​​for process data are set through consultation, as mentioned above, in sections (1b) where noise or impulses are instantaneously generated and fluctuate greatly, such as when gas is injected into the chamber or when the power switch for generating plasma is turned on, it is not possible to set meaningful upper and lower critical values ​​for warning alarms in the process data. Therefore, a technical solution is needed to address this issue. [Overview of the Initiative] [Problems that the invention aims to solve]

[0011] The technical objective of the present invention to solve the aforementioned problems is to provide a process data processing device and a process processing method that enable the uniform setting of upper and lower critical values ​​without requiring the operator to directly adjust the process data.

[0012] Another technical objective of the present invention is to provide a process data processing device and a process processing method that can set meaningful upper and lower critical values ​​even in sections where impulses fluctuate instantaneously like noise.

[0013] The problems that this invention aims to solve are not limited to those described above, and other problems not mentioned can be clearly understood by a person with ordinary skill in the art to which this invention pertains from this specification and the accompanying drawings. [Means for solving the problem]

[0014] The process data processing device of the present invention for achieving the aforementioned technical problems includes: a sensor unit that detects a physical magnitude value relative to the time during process progression; a process data recording unit that, in conjunction with the sensor unit, receives input of a physical magnitude value relative to time each time the process schedule progresses, records the input physical magnitude value relative to time as process data, and accumulates and records the process data for each process schedule; and a calculation unit that selects at least one excellent process data from the process data by a pre-set excellent process selection calculation, and sets pre-set range values ​​for the selected excellent process data using the upper and lower critical values ​​of the process data.

[0015] According to one embodiment, the upper and lower critical values ​​can be formed to include upper and lower limits of large fluctuations, where the physical magnitude value changes rapidly and changes significantly over a certain period of time, similar to pulses and noise, when viewed in terms of the amount of change in the relative physical magnitude value over a certain period of time.

[0016] According to one embodiment, when the calculation unit performs the excellent process selection calculation, the process data can be divided into sections and calculated accordingly.

[0017] According to one embodiment, when the calculation unit performs the excellent process selection calculation, the intervals can be divided based on whether the differential data generated by differentiating the process data exceeds a previously set reference value or does not.

[0018] According to one embodiment, when the calculation unit performs the excellent process selection calculation, the process data can be processed time-series by section to extract characteristic information.

[0019] According to one embodiment, the characteristic information can be extracted for each interval, including the start time, end time, duration, start value, end value, minimum value, and maximum value.

[0020] According to one embodiment, the characteristic information can be extracted for each interval, further including at least one of the following: Range, Mean, Autocorrelation, Standard deviation, Slope, R², Kurtosis, Concavity, Convexity, Skewness, and Crossing ratio.

[0021] According to one embodiment, when the calculation unit performs the excellent process selection calculation, it clusters the characteristic information of the process data to generate cluster cluster data, and extracts at least one cluster within a certain range at the cluster center of the generated cluster cluster data as an excellent process.

[0022] According to one embodiment, when the arithmetic unit generates the cluster group data, it can generate a plurality of cluster group data and select the number of clusters from the plurality of cluster group data.

[0023] According to one embodiment, when the arithmetic unit selects the number of clusters from the plurality of cluster group data, it can be selected by any one of the Silhouette method, the Elbow method, and the prior information classification method.

[0024] According to one embodiment, the arithmetic unit can extract the excellent process by using any one of the K-Means model, the GMM (Gaussian Mixture Model) model, and the Agglomerative Hierarchical Clustering model.

[0025] According to one embodiment, the upper critical value and the lower critical value of the process data are set by applying the upper and lower limit rates and the upper and lower limit deviations to the average value of one process data associated with the excellent process or a plurality of process data extracted from the excellent process, or can be set by the average value of the maximum value and the minimum value of a plurality of process data extracted from the excellent process.

[0026] On the other hand, the process data recording unit receives the input of the physical magnitude value with respect to time each time the process schedule progresses in conjunction with the sensor unit, records the physical magnitude value with respect to the received time as process data, and accumulates and records the process data for each process schedule; the arithmetic unit selects at least one excellent process data by the preset excellent process selection operation from the process data and sets the preset range value with the upper critical value and the lower critical value of the process data for the selected excellent process data; and the critical value monitoring unit generates a warning alarm when the physical magnitude value with respect to the time detected by the sensor unit exceeds the upper critical value and the lower critical value; including.

[0027] According to an embodiment, when the upper critical value and the lower critical value in the calculation stage are viewed by classifying based on the amount of change in the physical magnitude value compared over a certain period of time, they can be formed to include a large fluctuation upper and lower limit region that instantaneously changes sharply like a pulse and noise and greatly fluctuates the physical magnitude value compared over a certain period of time.

[0028] According to an embodiment, when the calculation unit performs the excellent process selection calculation in the calculation stage,

[0029] the process data can be calculated by being divided into sections.

[0030] According to an embodiment, when the calculation unit performs the excellent process selection calculation in the calculation stage, the sections can be classified based on whether the differential data generated by differentiating the process data exceeds a preset reference value or not.

[0031] According to an embodiment, when the calculation unit performs the excellent process selection calculation in the calculation stage,

[0032] the process data can be processed in time series for each section to extract feature information.

[0033] According to an embodiment, the feature information can be extracted to include the start time, end time, duration, start value, end value, minimum value, and maximum value for each section.

[0034] According to one embodiment, the characteristic information can be extracted for each interval, further including at least one of the following: Range, Mean, Autocorrelation, Standard deviation, Slope, R², Kurtosis, Concavity, Convexity, Skewness, and Crossing ratio.

[0035] According to one embodiment, during the calculation stage, when the calculation unit performs the excellent process selection calculation, it clusters the characteristic information of the process data to generate cluster cluster data, and extracts at least one cluster within a certain range at the cluster center of the generated cluster cluster data as an excellent process.

[0036] According to one embodiment, during the calculation stage, when the calculation unit generates the cluster group data, it can generate multiple cluster group data and select the number of clusters from the multiple cluster group data.

[0037] According to one embodiment, when the calculation unit selects the number of clusters during the calculation stage, it can select one of the following: the silhouette method, the elbow method, and the pre-information classification method.

[0038] According to one embodiment, in the calculation stage, the superior process can be extracted using any one of the following: the K-Means model, the GMM (Gaussian Mixture Model) model, and the Agglomerative Hierarchical Clustering model.

[0039] According to one embodiment, when setting the upper and lower critical values ​​of the process data in the calculation stage, they can be set by applying upper and lower limit rates or upper and lower limit deviations to the average value of one process data associated with the excellent process or multiple process data extracted in the excellent process, or by setting them to the average value of the maximum value and the average value of the minimum values ​​of multiple process data extracted in the excellent process. [Effects of the Invention]

[0040] This invention has the effect of enabling the setting of upper and lower critical values ​​in a uniform and stable manner, because the upper and lower critical values ​​are set based on process data generated during the process, without the operator having to directly set the upper and lower critical values.

[0041] Furthermore, when the upper and lower critical values ​​of process data are set and the amount of change in the relative physical magnitude value over a certain period of time is used as the basis for classification, the present invention has the effect of being able to uniformly and stably judge process data even in the upper and lower limits of large fluctuations where the relative physical magnitude value changes rapidly instantaneously, like pulses and noise, and fluctuates greatly over a certain period of time.

[0042] The effects of the present invention are not limited to those described above, and other effects not mentioned can be clearly understood by a person with ordinary skill in the art to which the present invention pertains from this specification and the accompanying drawings. [Brief explanation of the drawing]

[0043] [Figure 1] Figure 1 is a graph showing the relationship between the pressure value relative to time as one process schedule is carried out. [Figure 2] Figure 2 is a diagram showing the configuration of a process data processing device according to one embodiment of the present invention. [Figure 3] Figure 3 is a graph showing the process data obtained by differentiating the process data, where the process data represents the physical magnitude value relative to the pressure-to-time ratio in the process chamber, as shown in Figure 2. [Figure 4] Figure 4 is a table listing the feature information extracted by the processing unit. [Figure 5] Figure 5 is a dimensional graph showing the state in which the calculation unit clusters the feature information extracted for each interval. [Figure 6] Figure 6 shows graphs of the effective silhouette values ​​for each cluster of data when the calculation unit performs the silhouette method. [Figure 7] Figure 7 is a graph of the results obtained when the calculation unit performs the elbow method. [Figure 8] Figure 8 is a graph showing the center of each cluster of data. [Figure 9] Figure 9 is a graph showing the process data for each of the cluster group data selected for the superior process, plotted on a single graph. [Figure 10] Figure 10 is a graph showing the upper and lower critical values ​​of the process data. [Figure 11] Figure 11 is a flowchart of a process processing method using a process data processing device according to one embodiment of the present invention. [Figure 12] Figure 12 is a graph showing the state of monitoring whether the process data, as the process progresses according to the process schedule, has moved beyond the upper and lower critical limits stored in the critical value monitoring unit. [Figure 13] Figure 13 is a graph showing the state of monitoring whether the process data, as the process progresses according to the process schedule, has passed the upper and lower critical limits stored in the critical value monitoring unit. [Modes for carrying out the invention]

[0044] Hereafter, embodiments for carrying out the present invention will be described with reference to the attached drawings. In this case, when the entire specification is described as "encompassing" a certain component, this is considered not to control other components unless specifically contradicted, but rather to mean that other components may be further included. Furthermore, terms such as "...part" in the specification are considered to mean a unit that processes at least one function or operation when describing electronic hardware or electronic software, and when describing mechanical devices, they are considered to mean a single part, function, application, pivot point, or drive element. In addition, identical or similar components will be described using the same reference numerals in the drawings, and redundant descriptions of identical components will be omitted.

[0045] Furthermore, when an element or layer is referred to in the present invention as “on,” “connected,” “joined,” “attached,” “adjacent,” or “covering” another element or layer, this means that it is directly on, connected to, joined to, attached to, touching, or covering the other element or layer, or that intermediate elements or layers may be present. Conversely, when an element is referred to as “directly on,” “directly connected,” or “directly joined” another element or layer, it should be understood that there are no intermediate elements or layers. Throughout the specification, the same reference numeral refers to the same element. The term “and / or” as used in the present invention includes all combinations and subcombinations of one or more items from the enumerated items.

[0046] Figure 2 is a diagram of the configuration of a process data processing device according to one embodiment of the present invention. Figure 3 is a graph showing the process data differentiated by the process data, where the physical magnitude value against pressure relative to time in the process chamber is shown as process data in the process data processing device shown in Figure 2. Figure 4 is a table showing each of the feature information extracted by the calculation unit. Figure 5 is a dimensional graph showing the state in which the feature information extracted by the calculation unit for each interval is clustered on a base. Figure 6 is a graph of the silhouette effective values ​​for each of the cluster group data when the calculation unit performs the silhouette method. Figure 7 is a graph of the result values ​​when the calculation unit performs the elbow method. Figure 8 is a graph showing the center of each of the cluster group data. Figure 9 is a graph showing the process data for each of the cluster group data selected as excellent processes on a single graph. Figure 10 is a graph showing the upper and lower critical values ​​of the process data.

[0047] As shown in Figures 1 to 10, a process data processing device according to one embodiment of the present invention includes a sensor unit 10, a process data recording unit 20, and a calculation unit 30.

[0048] The sensor unit 10 is a device that detects physical magnitude values ​​in relation to the time taken for a process to proceed in a Processing Unit. Here, the process includes all processes for processing a substrate such as a wafer, such as oxidation, photo, exposure, heating, photoresist removal, physical etching, chemical etching, cleaning, deposition, and packaging, and can also include all processes related to the transfer and storage of the substrate. The sensor unit 10 used in such a process can detect physical magnitude values ​​such as temperature, humidity, density, voltage, current, power, pressure, displacement, angle, and light intensity, and generate analog or digital signal values ​​corresponding to the magnitude. For example, the sensor unit 10 can be configured by selectively using a thermistor, RTD sensor, potential sensor, humidity sensor, eddy current sensor, CMOS sensor, photodiode, strain gauge, piezoelectric sensor, lidar sensor, or laser sensor, and may further include a signal converter that converts these sensors into analog or digital signals. As the sensor unit 10 can be modified and implemented in various ways, this embodiment exemplifies that, for the sake of explanation, the sensor unit 10 includes a pressure sensor for measuring the pressure inside the process chamber and a signal converter that converts the signal value detected by the pressure sensor into a digital signal. As a result, when the sensor unit 10 measures the pressure inside the process chamber, process data is detected as a physical magnitude value relative to the pressure over time, as shown in Figure 3.

[0049] The process data recording unit 20 is an input / output arithmetic device such as a PC, ROM (Read Only Memory), PLC, or programmable logic. This process data recording unit 20 works in conjunction with the sensor unit 10 to receive input of physical magnitude values ​​relative to time each time the process schedule progresses. It records the received physical magnitude values ​​relative to time as process data, accumulating and recording the process data for each process schedule. For example, as shown in Figure 3, the process data recording unit 20 records process data for one process schedule, and then records new process data in a similar form to Figure 3 when the next process schedule progresses, thereby recording process data for each process schedule. In this case, as shown in Figure 3, when the process data is divided based on the amount of change in the physical magnitude value relative to a certain time, it can be formed to include a small fluctuation region 2a where the physical magnitude value relative to a certain time does not change significantly, and a large fluctuation region 2b where the physical magnitude value relative to a certain time changes significantly.

[0050] The calculation unit 30 is an input / output calculation device such as a PC, ROM, PLC, or programmable logic. In this case, the calculation unit 30 can be configured to work in conjunction with the process data recording unit 20 as a single input / output calculation device, or they can be configured to work in conjunction separately.

[0051] The calculation unit 30 then uses a pre-configured excellent process selection calculation to select at least one excellent process data from the process data, and sets the pre-configured range values ​​for the selected excellent process data using the upper limit critical value 6a and the lower limit critical value 6b of the process data.

[0052] More specifically, to give an example of how the calculation unit 30 performs the superior process selection calculation, first, when the calculation unit 30 is performing the superior process selection calculation, it divides the process data into sections 3a and performs the calculation. The reason for dividing the process data into sections 3a and performing the calculation is to prevent the characteristics of the data for each detailed schedule from being mixed and used together when analyzing the data for many detailed schedules that proceed within a single process schedule. To give an example of a detailed schedule, when a detailed schedule proceeds within a process chamber and process data related to pressure is generated, a process proceeds where process gas is injected and the pressure inside the process chamber changes rapidly, a process proceeds while the pressure inside the process chamber is maintained at a constant level, and a process proceeds where process gas is discharged and the pressure inside the process chamber is discharged.

[0053] In this case, the criterion by which the calculation unit 30 divides the process data into sections 3a can be set by the respective time points already set on the process schedule.

[0054] Furthermore, the arithmetic unit 30 divides the process data into sections 3a based on whether the differential data generated by the differentiation of the process data exceeds or does not exceed a pre-set criterion value 3b, as shown in Figure 3. In this way, when a single process schedule is in progress, the process data is divided into sections 3a for each detailed schedule within the process schedule, so that during data analysis, sections 3a are not mixed with other detailed schedules during analysis. The process data shown in Figure 3 shows the changes in the process data slowly to make it easier to understand the process for forming the differential data, but the process data graphs exemplified later will be illustrated and explained using graphs with large changes in process data.

[0055] Furthermore, as described above, the calculation unit 30 divides the process data into intervals 3a and then performs time-series processing on each interval 3a to extract feature information 4. Here, the feature information 4 can be used to compare each interval 3a, to apply a prediction model to each interval 3a and score it, or to detect whether there are any anomalies in each interval 3a.

[0056] As an example of such feature information 4, as shown in Figure 4, feature information 4 can be extracted for each interval 3a of process data, including start time, end time, duration, start value, end value, minimum value, and maximum value. Furthermore, feature information 4 can be formed by extracting at least one of the following for each interval 3a based on the extracted values: range, mean, autocorrelation, standard deviation, slope, coefficient of determination (R2), kurtosis, concavity, convexity, skewness, and crossing ratio. As a result, the calculation unit 30 extracts a total of 18 feature information pieces 4 for each interval 3a as described above, and can then compare them for each interval 3a, apply a prediction model to each interval 3a and score them, or use them as detection information to determine whether there are anomalies in each interval 3a.

[0057] Furthermore, after the calculation unit 30 extracts the feature information 4 as described above for each section 3a, it clusters the feature information 4 from the process data as shown in Figure 5, and points are grouped together to generate cluster data 5. In this embodiment, five cluster data 5 can be generated for each section 3a.

[0058] To explain why the calculation unit 30 clusters the characteristic information 4 of the process data, even if each piece of process data is extracted by the same sensor unit 10, environmental changes, recipe changes, deviations in the equipment itself, etc., can cause many cases where each piece of process data generates other forms of process data. In this way, even if all other forms of process data fall within the normal range, the difference in form between them can cause normal process data to be considered defective data. Therefore, the calculation unit 30 proceeds with clustering to determine whether each piece of process data, even if its form differs from that of the others, falls within the normal range of the upper critical value 6a and the lower critical value 6b.

[0059] To proceed with clustering, the calculation unit 30 generates multiple cluster data sets 5 as shown in Figure 5, and selects the number of clusters from the multiple cluster data sets 5 so that only clusters close to the superior process are set.

[0060] Here, the calculation unit 30 can select the number of clusters among the multiple cluster data 5 using one of the silhouette method, the elbow method, or the prior information classification method.

[0061] First, when the calculation unit 30 decides to use the silhouette method, the calculation unit 30 calculates the number of clusters using the equation (1) below and executes the method. As illustrated in Figure 6, different result values ​​for the number of silhouette points are derived, and the five clusters with the highest number of silhouette points are set as a reliable number of clusters.

[0062] JPEG2026509849000002.jpg1341--Formula (1), a (i) , b (i) The mean distance from data point (i) within each cluster (a, b) is S (i)(This is the silhouette coefficient of data point (i))

[0063] Next, when the calculation unit 30 uses the elbow method, the elbow method is a method in which the calculation unit 30 squares and sums the difference in distance between samples belonging to a specific cluster at the center of the clustered data 5, i.e., the error, and then specifies a pivot point by the number of clusters where the improvement in the corresponding value is not large when the number of clusters is increased. Here, the criterion for the degree to which the improvement is not large is that the sum of squared errors becomes small. When the calculation unit 30 executes the elbow method in this way, as shown in Figure 7, the decrease in the value of the sum of squared errors becomes small up to the number of clusters of 5, and at the moment the number of clusters becomes 6, it can be confirmed that there is almost no decrease in the value of the sum of squared errors. By setting the interval 3a of 5, where the decrease in the value of the sum of squared errors becomes gradual, by the number of clusters, it becomes possible to optimally select the number of clusters in the clustered data 5.

[0064] Next, if the arithmetic unit 30 decides to use a pre-classified information method, the pre-classified information method will set the number of clusters by referring to the number of clusters previously used and the number of clusters already set by related engineers.

[0065] Once the number of clusters in the cluster data 5 is determined by selecting the method in this way, the calculation unit 30 starts the clustering process. At this time, the clustering process proceeds using one of the following models: the K-Means model, the GMM (Gaussian Mixture Model), and the Agglomerative Hierarchical Clustering model, to extract the best processes.

[0066] Here, in order to extract the best process, the calculation unit 30 extracts cluster data 5 that are within a certain range at the cluster center of the generated cluster data 5, as shown in Figure 8. Here, the cluster center of the cluster data 5 can be calculated by averaging the plane coordinates of the clusters that make up the cluster data 5. Generally, clusters that occur most frequently are densely concentrated at the center of such cluster data 5. In other words, the closer the cluster data 5 are to the center, the more stable the process corresponds to, and therefore they can be selected as the best process. In this case, when selecting the center of the clustering data as the best process, cluster data 5 that are within a certain rate range at the center, for example, 5%, are extracted as the best process, and the most reliable best process can be selected.

[0067] In this way, the calculation unit 30, which has selected the superior processes, sets the upper critical value 6a and the lower critical value 6b based on the clusters corresponding to the superior processes. At this time, when the upper critical value 6a and the lower critical value 6b calculated by the calculation unit 30 are viewed in terms of the amount of change in the relative physical magnitude value over a certain period of time, they can be formed to include a small fluctuation upper and lower limit region 7a where the relative physical magnitude value does not change much over a certain period of time, and a large fluctuation upper and lower limit region 7b where the relative physical magnitude value changes rapidly instantaneously, like pulses and noise, and fluctuates greatly over a certain period of time.

[0068] In this case, the calculation unit 30 can set the upper limit critical value 6a and the lower limit critical value 6b by applying the upper and lower limit rates and upper and lower limit deviations to the average value of one process data associated with the excellent process at the time of setting, or to multiple process data extracted from the excellent process, or by setting them with the average value of the maximum value and the average value of the minimum value of multiple process data extracted from the excellent process.

[0069] For example, the calculation unit 30, which has selected the best processes, can select process data associated with being closest to the center of the cluster data 5 as the best processes, and then apply upper and lower limit percentage rates or upper and lower limit deviations to the selected process data to set the upper limit critical value 6a and the lower limit critical value 6b.

[0070] To give another example, the calculation unit 30 calculates the average value for each of the multiple process data associated with the excellent process, selects the process data associated with the excellent process, and, as shown in Figure 9, superimposes the multiple process data selected for the excellent process in terms of time-relative physical magnitude value, sets the process data showing the most maximum value as the upper limit critical value 6a, and the process data showing the most minimum value as the lower limit critical value 6b. Alternatively, as shown in Figure 9, the calculation unit 30 can find the average of the multiple process data associated with the excellent process, determine the upper and lower limit standard deviations with respect to the average, and set the upper limit critical value 6a and the lower limit critical value 6b.

[0071] The upper critical value 6a and lower critical value 6b calculated by the calculation unit 30 are transmitted to the critical value monitoring unit 40, which monitors the physical size value in conjunction with the sensor unit 10. The critical value monitoring unit 40 generates a warning alarm to notify the process processing device that an issue has occurred when the process data for the time-relative physical size value measured by the sensor unit 10 falls below the upper critical value 6a and lower critical value 6b.

[0072] The following section will describe a process processing method using the aforementioned process data processing device.

[0073] Figure 11 is a flowchart of a process processing method using a process data processing device according to one embodiment of the present invention.

[0074] Referring further to Figure 11, the process processing method according to one embodiment of the present invention includes a process data recording step (S10), a calculation step (S20), and a critical value monitoring step (S30).

[0075] First, in the process data recording stage (S10), the process data recording unit 20 works in conjunction with the sensor unit 10 to receive input of physical magnitude values ​​relative to time each time the process schedule progresses. The input physical magnitude values ​​relative to time are recorded as process data, and the process data is accumulated and recorded for each process schedule.

[0076] Next, the calculation stage (S20) proceeds, including the interval calculation stage (S21), the feature information extraction stage (S22), the superior process calculation stage (S23), and the critical value extraction stage (S24).

[0077] First, in the interval calculation stage (S21), as described above, the calculation unit 30 divides the process data into intervals 3a and performs calculations. In this case, the intervals 3a can be divided as described above, either by a change in the physical size value of the process data on the process schedule, or by a pre-set point in time when the process changes, or, as shown in Figure 3, by whether the differential data generated by differentiating the process data exceeds or does not exceed a pre-set reference value.

[0078] Next, in the feature information extraction stage (S22), the calculation unit 30 performs the superior process selection calculation, processing the process data in time series for each interval 3a to extract feature information 4. In this case, as mentioned above with reference to Figure 4, feature information 4 can be extracted from a total of 18 data points.

[0079] Next, in the superior process calculation stage (S23), the calculation unit 30 clusters the characteristic information 4 from the process data to generate cluster cluster data 5 as shown in Figure 5, and the cluster cluster data 5 that are within a certain range at the cluster center of the generated cluster cluster data 5 are extracted in the superior process.

[0080] In this case, the superior process calculation stage (S23) involves the calculation unit 30 generating multiple cluster cluster data 5 in order to extract clusters within the cluster cluster data 5 for superior processes, selecting a number of clusters from among the multiple cluster cluster data 5 that can be used for superior processes, and executing the selected clusters to extract superior processes from clusters that are within a certain range at the center of the cluster cluster data 5.

[0081] At this time, in the superior process calculation stage (S23), as described above, when the calculation unit 30 selects the number of clusters from multiple cluster group data 5, it can select using one of the silhouette method, the elbow method, and the prior information classification method, and it can extract superior processes using one of the K-Means model, the GMM (Gaussian Mixture Model) model, and the Agglomerative Hierarchical Clustering model.

[0082] Next, in the critical value extraction stage (S24), as shown in Figure 10, the upper limit critical value 6a and the lower limit critical value 6b of the process data are set. For this purpose, in the critical value extraction stage (S24), process data associated with at least one excellent process extracted in the excellent process calculation stage (S23) can be used, or multiple process data associated with clusters within a certain range at the cluster center can be used. As mentioned above, the values ​​can be set by applying upper and lower limit rates and upper and lower limit deviations to the average value of one process data associated with an excellent process or multiple process data extracted from an excellent process, or by setting the values ​​with respect to the average value of the maximum value and the average value of the minimum values ​​of multiple process data extracted from an excellent process.

[0083] Next, in the critical value monitoring stage (S30), the critical value monitoring unit 40 receives input of a physical magnitude value relative to time in conjunction with the sensor unit 10, and the critical value monitoring unit 40 generates a warning alarm when the physical magnitude value relative to time detected by the sensor unit 10 falls below the upper critical value 6a and the lower critical value 6b. Here, the critical value monitoring unit 40 can be composed of an input / output arithmetic device such as a PC, ROM, PLC, or programmable logic.

[0084] In this critical value monitoring stage (S30), the critical value monitoring unit 40 becomes linked to the process processing device that is advancing the process, and receives input from the calculation unit 30 of the upper limit critical value 6a and the lower limit critical value 6b of the process data associated with the excellent process, and monitors them.

[0085] More specifically, referring to Figures 12 and 13, Figures 12 and 13 are graphs that monitor whether the process data, as the process progresses according to the process schedule, has moved beyond the upper critical value 6a and lower critical value 6b stored in the critical value monitoring unit 40.

[0086] As shown in Figure 12, during the critical value monitoring stage (S30), the critical value monitoring unit 40 continuously records the process data input through the sensor unit 10 and monitors whether it exceeds the upper critical value 6a and the lower critical value 6b. In this way, during the critical value monitoring stage (S30), the critical value monitoring unit 40 monitors the process data and does not generate a warning alarm if it does not exceed the upper critical value 6a and the lower critical value 6b.

[0087] In contrast, as shown in Figure 13, during the critical value monitoring stage (S30), the critical value monitoring unit 40 monitors the process data and generates a warning alarm 6c if the process exceeds the upper critical value 6a and the lower critical value 6b. This allows the operator to check the warning alarm to see if it was triggered by an issue in the process.

[0088] In this way, the process data processing device and process processing method according to one embodiment of the present invention enable the setting of the upper and lower critical values ​​6a and 6b uniformly and stably, because the upper and lower critical values ​​6a and 6b are set based on process data generated during the process, without the operator having to directly set the upper and lower critical values ​​6a and 6b. Therefore, the process data processing device and process processing method according to one embodiment of the present invention reduces operator fatigue and improves the reliability of warning alarms by generating only warning alarms related to actual issues.

[0089] Furthermore, in the process data processing apparatus and process processing method according to one embodiment of the present invention, when the upper limit critical value 6a and lower limit critical value 6b of the process data are separated and viewed based on the amount of change in the relative physical magnitude value over a certain period of time when set, it becomes possible to uniformly and stably determine the process data even up to the upper and lower limit regions 7b of large fluctuations, where the relative physical magnitude value changes rapidly instantaneously, like pulses and noise, and fluctuates greatly over a certain period of time.

[0090] As described above, the present invention has been explained with specific details such as concrete components, limited embodiments, and drawings. However, these are provided only to aid in a more general understanding of the present invention, and the present invention is not limited to the above embodiments. A person with ordinary skill in the art to which the present invention belongs can make various modifications and variations from this description.

[0091] Therefore, the concept of the present invention should not be limited to the embodiments described, and all things that are equivalent to or have equivalent variations to the claims described later would also fall within the scope of the concept of the present invention.

Claims

1. A sensor unit that detects the physical magnitude value in relation to the time it takes for the process to progress; A process data recording unit that, in conjunction with the aforementioned sensor unit, receives input of a physical magnitude value relative to time each time the process schedule progresses, records the received physical magnitude value relative to time as process data, and accumulates and records the process data for each process schedule; and A process data processing device comprising: a calculation unit that selects at least one excellent process data from the aforementioned process data by a pre-set excellent process selection calculation, and sets pre-set range values ​​for the selected excellent process data using the upper and lower critical values ​​of the process data;

2. The aforementioned upper critical value and lower critical value are, The process data processing apparatus according to claim 1, wherein when the amount of change in the relative physical magnitude value over a certain period of time is viewed in classification based on this classification, the classification includes upper and lower limits of large fluctuations where the relative physical magnitude value changes rapidly instantaneously, like pulses and noise, and fluctuates significantly over a certain period of time.

3. When the calculation unit proceeds with the excellent process selection calculation, The process data processing apparatus according to claim 1, which divides the aforementioned process data into sections and performs calculations accordingly.

4. When the calculation unit proceeds with the excellent process selection calculation, The process data processing apparatus according to claim 3, wherein the intervals are divided based on whether the differential data generated by differentiating the process data exceeds a previously set reference value or does not.

5. When the calculation unit proceeds with the excellent process selection calculation, The process data processing apparatus according to claim 3, which processes the process data in time series for each of the intervals to extract feature information.

6. The aforementioned characteristic information is The process data processing apparatus according to claim 5, which extracts data for each of the aforementioned intervals, including a start time, end time, duration, start value, end value, minimum value, and maximum value.

7. The aforementioned characteristic information is The process data processing apparatus according to claim 6, wherein the extraction of data for each interval further includes at least one of the following: Range, Mean, Autocorrelation, Standard deviation, Slope, R², Kurtosis, Concavity, Convexity, Skewness, and Crossing ratio.

8. When the calculation unit proceeds with the excellent process selection calculation, The process data processing apparatus according to claim 5, comprising: clustering the characteristic information of the process data to generate cluster group data; and extracting at least one cluster within a certain range at the cluster center of the generated cluster group data as a superior process.

9. When the calculation unit generates the cluster data, A process data processing apparatus according to claim 8, which generates multiple cluster group data and selects the number of clusters among the multiple cluster group data.

10. When the calculation unit selects the number of clusters from the multiple cluster data sets, The process data processing apparatus according to claim 9, wherein the process data processing apparatus is selected from one of the following: the silhouette method, the elbow method, and the pre-information classification method.

11. The process data processing apparatus according to claim 8, wherein the calculation unit extracts the superior process using one of the K-Means model, the GMM (Gaussian Mixture Model) model, and the Agglomerative Hierarchical Clustering model.

12. The upper and lower critical values ​​of the aforementioned process data are: The process data processing apparatus according to claim 1, wherein the value is set by applying upper and lower limit rates or upper and lower limit deviations to the average value of one process data associated with the excellent process or multiple process data extracted in the excellent process, or by setting the value with respect to the average value of the maximum value and the average value of the minimum values ​​of multiple process data extracted in the excellent process.

13. The process data recording unit, in conjunction with the sensor unit, receives input of physical magnitude values ​​relative to time each time the process schedule progresses, records the received physical magnitude values ​​relative to time as process data, and accumulates and records the said process data for each process schedule in the process data recording stage; A calculation step in which the calculation unit selects at least one excellent process data from the process data using a pre-set excellent process selection calculation, and sets pre-set range values ​​for the selected excellent process data using the upper and lower critical values ​​of the process data; and, A process method comprising a critical value monitoring step in which a critical value monitoring unit generates a warning alarm when the physical magnitude value for time detected by the sensor unit falls below the upper and lower critical values.

14. In the calculation stage, the upper limit critical value and the lower limit critical value are The process processing method according to claim 12, wherein when the amount of change in the relative physical magnitude value over a certain period of time is divided and viewed based on this standard, the process is formed to include upper and lower limits of large fluctuations where the relative physical magnitude value changes rapidly instantaneously, like pulses and noise, and fluctuates greatly over a certain period of time.

15. During the calculation stage, when the calculation unit proceeds with the excellent process selection calculation, The process processing method according to claim 12, wherein the aforementioned process data are divided into sections and calculated accordingly.

16. During the calculation stage, when the calculation unit proceeds with the excellent process selection calculation, The process processing method according to claim 13, wherein the intervals are divided based on whether the differential data generated by differentiating the process data exceeds a previously set reference value or does not.

17. During the calculation stage, when the calculation unit proceeds with the excellent process selection calculation, The process processing method according to claim 15, wherein the process data is processed time-series by each of the intervals to extract feature information.

18. The aforementioned characteristic information is The process processing method according to claim 17, wherein the data is extracted for each of the aforementioned intervals, including a start time, end time, duration, start value, end value, minimum value, and maximum value.

19. The aforementioned characteristic information is The process processing method according to claim 18, wherein the extraction for each interval further includes at least one of the following: Range, Mean, Autocorrelation, Standard deviation, Slope, R², Kurtosis, Concavity, Convexity, Skewness, and Crossing ratio.

20. During the calculation stage, when the calculation unit proceeds with the excellent process selection calculation, The process processing method according to claim 17, comprising: clustering the characteristic information of the process data to generate cluster group data; and extracting at least one cluster within a certain range at the cluster center of the generated cluster group data in a superior process.

21. During the calculation stage, when the calculation unit generates the cluster data, The process method according to claim 20, comprising generating multiple cluster data sets and selecting the number of clusters among the multiple cluster data sets.

22. During the calculation stage, when the calculation unit selects the number of clusters, The process processing method according to claim 21, wherein one of the following methods is selected: the silhouette method, the elbow method, and the pre-information classification method.

23. In the aforementioned calculation stage The process processing method according to claim 21, wherein the superior process is extracted using one of the following: a K-Means model, a GMM (Gaussian Mixture Model) model, and an Agglomerative Hierarchical Clustering model.

24. In the calculation stage, when setting the upper and lower critical values ​​of the process data, The process processing method according to claim 13, wherein the value is set by applying upper and lower limit rates or upper and lower limit deviations to the average value of one process data associated with the excellent process or multiple process data extracted in the excellent process, or by setting the value with respect to the average value of the maximum value and the average value of the minimum values ​​of multiple process data extracted in the excellent process.

Images