WO2022030831A1

WO2022030831A1 - Process abnormality detection device and method using analysis of control section temperature signal

Info

Publication number: WO2022030831A1
Application number: PCT/KR2021/009532
Authority: WO
Inventors: 김남기; 왕지남; 박준표; 한강희; 한승우; 김연동
Original assignee: 주식회사 유디엠텍
Priority date: 2020-08-03
Filing date: 2021-07-23
Publication date: 2022-02-10
Also published as: CN116249945A; KR102220138B9; KR102220138B1

Abstract

A process abnormality detection device using the analysis of a control section temperature signal, of the present invention, comprises: a preprocessing unit for performing data preprocessing on an analog temperature signal corresponding to a PLC control process; a feature information extraction unit for extracting pieces of feature information from the preprocessed analog temperature signal; and a process abnormality detection unit, which applies the extracted pieces of feature information to a machine learning model to detect the abnormality of the PLC control process corresponding to the analog temperature signal.

Description

Process abnormality detection device and method through analysis of temperature signal in control section

The present invention relates to a programmable logic controller (PLC) control, and more particularly, to a technology capable of accurately detecting defects and abnormalities through analysis of a control section.

PLC is an industrial computer programmed in a low level language and used to control an automation system. The PLC program, which is the internal logic of PLC, controls the automation system through Boolean operation. The PLC program designed in the general manufacturing process goes through the verification process, and when the accuracy of the program is guaranteed, it controls the actual automation system.

In the recent automated manufacturing industry, the control logic is vast and very complexly designed as the complexity of the manufacturing line increases. Accordingly, the PLC program is also complicatedly logicized. For this reason, it is also becoming more and more difficult to diagnose and monitor the PLC program, and accordingly, the time it takes to find and correct an error is gradually increasing. According to Operational Diagnostics, The holy grail of control automation, the delay due to the time it takes to identify these diagnostic methods and errors accounts for more than 80% of the total equipment downtime. In particular, in the case of an automobile body assembly line, the average cycle time is about 1 minute, and therefore, when the line is stopped due to equipment failure, a large loss of profit is caused for a short period of time.

A typical automation system consists of numerous robots and automated transport devices. The robot and the transfer device perform various tasks such as welding or transfer according to the logic of the PLC program. Recently, automation systems have high complexity due to the construction of large-scale automated production lines, and include various work failure factors such as errors of the equipment itself or errors caused by external factors such as interference with the range of motion of the robot. Delay due to failure during operation causes huge economic loss due to error detection and increase in device set-up time.

In order to diagnose these errors, the automation system is monitored by adding a code for diagnosis inside the PLC program that controls the flow of process and logistics. In general, when an error occurs in an automated manufacturing line in the automobile industry, which is a representative of the automation system, the PLC program is monitored through the predicted abnormality display module. That is, the conventional monitoring method cannot monitor all signals because a code according to the target of error diagnosis is separately written and added in anticipation of a high error occurrence area. Therefore, it can be said that the process to be monitored is very limited, and there is a limit to the monitoring method to detect errors that occur gradually. That is, there is a problem in that it is difficult to respond in advance to the operation failure phenomenon that occurs due to the gradual wear of devices or accessories.

In particular, in the case of a system abnormality that does not significantly affect the operation of the PLC control system, there is a problem in that it is impossible to detect all types of errors related to analog input/output signals because it is not possible to determine whether there is an abnormality through the control process model. As a prior document related to the present invention, there is Republic of Korea Patent Registration No. 10-0414437 (registration date: December 24, 2003).

The present invention extracts characteristic information from a PLC analog signal, and applies the extracted characteristic information to a machine learning model to detect a process abnormality with respect to a PLC analog signal. A process abnormality detection method according to the analysis of a control section signal and systems.

In order to solve the above problems, an apparatus for detecting process anomalies through analysis of a temperature signal in a control section of the present invention includes: a pre-processing unit for performing data pre-processing on an analog temperature signal corresponding to a PLC control process; a feature information extraction unit for extracting feature information from the pre-processed analog temperature signal; and a process abnormality detection unit for detecting whether there is an abnormality in a PLC control process corresponding to the analog temperature signal by applying the extracted characteristic information to a machine learning model.

The pre-processing unit may include: a signal dividing module dividing the analog temperature signal into segments; and a loss recovery module for restoring a loss section for the analog temperature signal divided into segments.

The signal division module is characterized in that the product production cycle unit is divided as the segment unit in the analog temperature signal.

The signal division module divides each detailed process unit in the product production cycle from the analog temperature signal as the segment unit.

The loss recovery module includes: endpoint temperature data among sampling temperature data of a first segment signal belonging to the segment signals in which the analog temperature signal is divided into segments, and sampling temperature data of a second segment signal adjacent to the first segment signal It is characterized in that the sampling temperature data of the loss section for the first segment signal or the second segment signal divided by the segment unit is restored by using a linear relationship between the starting point temperature data.

The loss recovery module divides each of the segment signals divided into the segment unit into a predetermined sampling period, and restores the sampling temperature data divided by the predetermined sampling period using a linear relationship between the original sampling temperature data. characterized in that

The feature information extracting unit is, in the analog temperature signal, a total integral area, a bow-shaped integral area, a Y-axis parallel movement integral area, an integral area of a temperature rise section, an integral area of a temperature fall section, a start end slope, a start in one product production cycle unit It is characterized in that at least one of the highest displacement difference, the highest final displacement difference, the mean, the standard deviation, the root mean square, and the shape coefficient is extracted as the feature information.

The feature information extraction unit is characterized in that by using a random forest algorithm among the machine learning method, it is characterized in that the selection of some of the feature information of the extracted feature information.

In order to solve the above problems, there is provided a process abnormality detection method through analysis of a temperature signal in a control section of the present invention, comprising: performing data pre-processing on an analog temperature signal corresponding to a PLC control process; extracting characteristic information from the pre-processed analog temperature signal; and detecting an abnormality in a PLC control process corresponding to the analog temperature signal by applying the extracted feature information to a machine learning model.

The pre-processing of the data may include: dividing the analog temperature signal into segments; and restoring a loss section for the analog temperature signal divided into segments.

The dividing into segments may include dividing a product production cycle unit in the analog temperature signal as the segment unit.

The dividing into segments may include dividing each detailed process unit in the product production cycle in the analog temperature signal as the segment unit.

The restoring of the loss section includes: endpoint temperature data among sampling temperature data of a first segment signal belonging to segment signals in which the analog temperature signal is divided into segments, and a second segment signal adjacent to the first segment signal. It is characterized in that the sampling temperature data of the loss section with respect to the first segment signal or the second segment signal divided by the segment unit is restored by using a linear relationship between the starting point temperature data among the sampling temperature data.

The step of restoring the loss section includes dividing each of the segment signals divided into the segment unit into a certain sampling section, and using a linear relationship between the original sampling temperature data, the sampling temperature divided by the predetermined sampling section It is characterized in that the data is restored.

In the step of extracting the feature information, the total integral area, the bow-shaped integral area, the Y-axis parallel movement integral area, the integral area of the temperature rise section, the integral area of the temperature drop section, the start and end from the analog temperature signal in one product production cycle unit It is characterized in that at least one or more of a slope, a starting highest displacement difference, a highest ending displacement difference, a mean, a standard deviation, a root mean square root, and a shape coefficient is extracted as the feature information.

The extracting of the feature information may include selecting some of the extracted feature information using a random forest algorithm among machine learning methods.

According to the present invention, data pre-processing is performed using an analog temperature signal corresponding to the PLC control process, and feature information is extracted from the pre-processed analog temperature signal to detect an abnormality in the PLC control process corresponding to the analog temperature signal. By doing so, it is possible to easily detect whether there is an abnormality in the process for the detailed control section in the PLC control process.

1 is a block diagram of an embodiment for explaining an apparatus for detecting process anomalies through analysis of a temperature signal in a control section of the present invention.

FIG. 2 is a block diagram of an embodiment for explaining the preprocessing unit shown in FIG. 1 .

3 is a reference diagram for an example illustrating that an analog temperature signal is divided into segments by a signal dividing module.

4 is a reference diagram of another example illustrating that an analog temperature signal is divided into segments by a signal dividing module.

5 is a reference diagram illustrating an example that a loss section for a segment signal is applied by a loss recovery module.

6 is a reference diagram illustrating an example that a loss section for a segment signal is applied by a loss recovery module.

7 is a reference diagram illustrating characteristic information that the characteristic information extraction unit can extract from an analog temperature signal.

8 is a flowchart of an embodiment for explaining a process abnormality detection method through analysis of a temperature signal in a control section of the present invention.

FIG. 9 is a flowchart of an embodiment for explaining a step of performing data preprocessing shown in FIG. 8 .

Hereinafter, with reference to the drawings, embodiments of the present invention will be described in detail so that those of ordinary skill in the art to which the present invention pertains can easily implement them. The present invention may be embodied in many different forms and is not limited to the embodiments described herein. In order to clearly explain the present invention, parts irrelevant to the description are omitted, and the same reference numerals are assigned to the same or similar components throughout the specification.

In general, it is difficult to detect and isolate faults and anomalies in a PLC (Programmable Logic Controller) control system. This invention proposes an automation tool for manufacturing process failure diagnosis called MPFDT (Manufacturing Process Failure Diagnosis Tool) that can accurately detect and isolate defects and abnormalities of PLC control systems.

Hereinafter, preferred embodiments of the present invention will be described in detail with reference to the accompanying drawings.

1 is a block diagram of an embodiment for explaining an apparatus 100 for detecting process anomalies through analysis of a temperature signal in a control section of the present invention.

Referring to FIG. 1 , the process anomaly detection apparatus 100 includes a pre-processing performing unit 110 , a feature information extracting unit 120 , and a process anomaly detecting unit 130 .

The preprocessing unit 110 performs data preprocessing on the collected analog temperature signal corresponding to the PLC control process.

FIG. 2 is a block diagram of an embodiment for explaining the preprocessing unit 110 shown in FIG. 1 .

Referring to FIG. 2 , the preprocessing unit 110 includes a signal division module 110-1 and a loss recovery module 110-2.

The signal division module 110-1 divides the analog temperature signal into segments.

The signal division module 110-1 may divide a product production cycle unit as the segment unit in the analog temperature signal.

3 is a reference diagram for an example illustrating that an analog temperature signal is divided into segments by the signal dividing module 110-1.

Referring to (a) of FIG. 3 , in the related art, in general, the analog temperature signal is divided by dividing the entire analog temperature signal at a predetermined time interval (eg, 1 minute interval) without considering the control section for the analog temperature signal, and the divided signal analysis method is used. According to this, since the pattern is different for each analysis point, it cannot accurately reflect the data pattern change of the analog temperature signal. In addition, when a process abnormality occurs in the analog temperature signal, it is difficult to have reliability in the data analysis result because it is impossible to detect the exact time when the process abnormality occurs.

Meanwhile, referring to FIG. 3B , the signal dividing module 110-1 divides the analog temperature signal into product production cycle units (eg, 74 second units) among segment units in consideration of the control section. According to this, since the analog temperature signal is divided into product production cycle units by the signal dividing module 110-1, each analysis point has a constant pattern, it is possible to easily distinguish a normal pattern from an abnormal pattern. In addition, data parts that do not require analysis, that is, parts in which process abnormalities occur, can be easily purified.

In addition, the signal dividing module 110-1 may divide each detailed process unit in the product production cycle in the analog temperature signal as the segment unit.

4 is a reference diagram of another example illustrating that the analog temperature signal is divided into segments by the signal dividing module 110-1.

Referring to FIG. 4 , the signal dividing module 110-1 may divide the analog temperature signal into each detailed process unit within a product production cycle based on PLC I/O log information. By dividing each detailed process in the product production cycle into segments, it is possible to specify which process section in the product production cycle has an error. For example, the operation of process A in the product production cycle takes about 5 seconds on average for every cycle, but when it takes 7 seconds due to an abnormality, it can be determined that the analog temperature signal of the process A is abnormal.

The loss recovery module 110-2 restores a loss section for the analog temperature signal divided into segments.

The loss recovery module 110-2 includes an endpoint temperature data among sampling temperature data of a first segment signal belonging to the segment signals in which the analog temperature signal is divided into segments and a second segment signal adjacent to the first segment signal. The sampling temperature data of the loss section with respect to the first segment signal or the second segment signal divided by the segment unit is restored by using a linear relationship between the starting point temperature data among the sampling temperature data.

5 is a reference diagram illustrating an example that a loss section for a segment signal is applied by the loss recovery module 110-2.

Referring to FIG. 5 , when the analog temperature signal is divided into segments, the start time and end time of the segment signal and the start time and end time of the sampling temperature data of the analog temperature signal do not exactly match. Therefore, since the degree of loss is different for each segment, it is impossible to pre-process data under the same conditions, and the values of the start time and end time of the sampling temperature data are unreliable. The smaller the size of the segment unit, the greater the effect of the loss.

In order to solve this, as shown in FIG. 5, the loss restoration module 110-2 selects the end point data or start point data based on the start time or end time of the segment signals so that the straight line passing through the end point data and the start point data is New sampling temperature data can be created at the point where it meets the reference point divided into segments.

For example, using a linear relational expression for a straight line passing through the end point temperature data EP1 among the sampling temperature data of the first segment signal and the start point temperature data SP1 among the sampling temperature data of the second segment signal adjacent to the first segment signal Thus, it is possible to calculate the sampling temperature data RP1 corresponding to the division point of the divided first segment signal or the second segment signal divided in units of segments. Accordingly, the loss recovery module 110-2 calculates the sampling temperature data RP1 corresponding to the dividing point, thereby generating an analog temperature signal for the section (eg, 0.5 second section) lost in the second segment signal. can do.

In addition, linearity with respect to a straight line passing through the end point temperature data EP2 among the sampling temperature data of the second segment signal and the starting point temperature data SP2 among the sampling temperature data of the third segment signal (not shown) adjacent to the second segment signal The sampling temperature data RP2 corresponding to the division point of the second segment signal or the third segment signal divided in units of segments may be calculated using the relational expression. Accordingly, the loss restoration module 110-2 calculates the sampling temperature data RP2 corresponding to the dividing point, thereby converting the analog temperature signal for the section lost in the second segment signal (for example, the section for 1.5 seconds). can do.

In addition, the loss recovery module 110-2 divides each of the segment signals divided into segments into a predetermined sampling period, and uses a linear relationship between the original sampling temperature data to separate each of the segment signals divided by the predetermined sampling period. Restore sampling temperature data.

6 is a reference diagram of another example illustrating that a loss section for a segment signal is applied by the loss recovery module 110-2.

Referring to FIG. 6 , when an analog temperature signal is divided based on a segment signal of the same size, the number of points of the sampling temperature data may be different for each segment, and even if the number of points of the sampling temperature data is constant, the points The collection time may not be constant.

In order to solve this problem, the loss recovery module 110-2 collects data points at regular time intervals so that sampling temperature data can be extracted from all segment signals under the same conditions.

For example, the loss recovery module 110 - 2 divides the sampling period into predetermined time units for each of the segment signals and calculates a data point value corresponding to each time unit. The loss recovery module 110-2 calculates the sampling temperature data RP10 divided by the predetermined sampling section by using a linear relational expression of a straight line passing through each of the original sampling temperature data EP10 and EP11. Accordingly, the loss recovery module 110-2 can be referred to as an analog temperature signal in which the number of data points and the collection interval are constant in all segment signals.

The feature information extraction unit 120 extracts feature information from the pre-processed analog temperature signal. The feature information extraction unit 120 includes the total integral area for one product production cycle in the analog temperature signal, the arc integral area, the Y-axis parallel movement integral area, the integral area for the temperature rise section, the integral area for the temperature drop section, the starting and ending slope, At least one of a starting maximum displacement difference, a maximum ending displacement difference, a mean, a standard deviation, a root mean square root, and a shape coefficient may be extracted as feature information.

7 is a reference diagram illustrating characteristic information that the characteristic information extraction unit 120 can extract from an analog temperature signal. As shown in Figure 7, the total integral area for the product production cycle, the integral area of the bow, the integral area of the Y-axis parallel movement, the integral area of the temperature rise section, the integral area of the temperature drop section, the start and end slope, the start maximum displacement difference, the highest The end displacement difference, the mean, the standard deviation, the root mean square, and the shape coefficient may be calculated through the following equations, respectively.

The total integral area can be calculated using Equation 1 below.

The arc integral area can be calculated using Equation 2 below.

The Y-axis translation integral area can be calculated using Equation 3 below.

The integral area of the temperature rise section can be calculated using Equation 4 below.

The integral area of the temperature drop section can be calculated using Equation 5 below.

The starting and ending slope can be calculated using Equation 6 below.

The starting maximum displacement difference can be calculated using the following Equation (7).

The highest end displacement difference can be calculated using the following Equation (8).

The average can be calculated using Equation 9 below.

The standard deviation can be calculated using Equation 10 below.

The root mean square can be calculated using Equation 11 below.

The shape coefficient can be calculated using Equation 12 below.

The feature information extraction unit 120 may select some feature information from among the feature information calculated in this way by using a random forest algorithm of a machine learning method.

The random forest algorithm refers to a method of generating a single model by randomly selecting elements used in making a decision tree to create multiple decision trees and combining the multiple decision trees. By selecting some of the feature information from among all the feature information using a random forest algorithm, more accurate classification of feature information can be achieved compared to regression analysis. The feature information extracting unit 120 uses a forest (random forest) algorithm to list the ranks of the feature information according to the importance of the feature, and from among these ranks, the feature information for accurately classifying whether there is an abnormality in the process is selected in order. Here, feature importance is a measure to evaluate which variable is most important in the process of creating a tree.

The process abnormality detection unit 130 detects whether there is an abnormality in the PLC control process corresponding to the analog temperature signal by applying the selected characteristic information to the machine learning model. A machine learning model includes an input layer, a hidden layer, and an output layer, and is a model trained in advance using training data, and the output layer can function as a classifier to determine whether there is an abnormality in the PLC control process. A detailed description of the machine learning model used in the process anomaly detection unit 130 is omitted in that it is a general model, however, the process anomaly detection unit 130 reshapes the selected characteristic information and applies it to the input layer. It has a structure that can output the result value through the output layer that determines whether there is an abnormality through the hidden layer learned through training.

The following Table 1 is a table for explaining that the selected characteristic information is applied to the process abnormality detection unit 130 to determine whether there is a process abnormality.

Referring to Table 1, 3 pieces of feature information selected using a random forest algorithm from among 12 kinds of feature information, that is, (b) circular integral area, (f) start and end slope, (k) root mean square process It is applied to the abnormality detection unit 130 to exemplify the detection of the presence or absence of an abnormality (no abnormality: OK, abnormal occurrence: NG) for the corresponding process for each product production cycle.

Data preprocessing is performed on the analog temperature signal corresponding to the PLC control process (step S1000).

First, the analog temperature signal is divided into segments (step S1010).

The dividing into segments may include dividing a product production cycle unit in the analog temperature signal as the segment unit. In addition, the dividing into segments may include dividing each detailed process unit in the product production cycle in the analog temperature signal as the segment unit.

After step S1010, a loss section for the analog temperature signal divided into segments is restored (step S1012).

The restoring of the loss section includes: endpoint temperature data among sampling temperature data of a first segment signal belonging to segment signals in which the analog temperature signal is divided into segments, and a second segment signal adjacent to the first segment signal. The sampling temperature data of the loss section with respect to the first segment signal or the second segment signal divided by the segment unit may be restored by using a linear relationship between the starting point temperature data among the sampling temperature data.

In addition, in the step of restoring the loss section, each of the segment signals divided into the segment unit is divided into a predetermined sampling interval, and divided by the predetermined sampling interval using a linear relationship between the original sampling temperature data. Sampling temperature data can be restored.

Meanwhile, after step S1000, feature information is extracted from the preprocessed data (step S1002).

In the step of extracting the feature information, the total integral area, the bow-shaped integral area, the Y-axis parallel movement integral area, the integral area of the temperature rise section, the integral area of the temperature drop section, the start and end from the analog temperature signal in one product production cycle unit At least one of a slope, a starting maximum displacement, a maximum ending displacement, a mean, a standard deviation, a root mean square root, and a shape coefficient may be extracted as the feature information. The extracting of the feature information may include selecting some of the extracted feature information by using a random forest algorithm among machine learning methods.

After step S1002, the extracted feature information is applied to the machine learning model to detect whether there is an abnormality in the PLC control process corresponding to the analog temperature signal (step S1004). The selected feature information may be reshaped and applied to the input layer, and the result value regarding the presence or absence of abnormality may be output through the output layer through the hidden layer learned through training.

The present invention may be implemented as a software program and may be implemented by a computer-readable recording medium. For example, the recording medium may be a hard disk, flash memory, RAM, ROM, etc. built-in to each playback device, or an optical disk such as a CD-R or CD-RW, compact flash card, smart media, memory stick, or multimedia card as an external type. have.

Although the embodiments of the present invention have been described as described above, the embodiments disclosed in the specification of the present invention do not limit the present invention. The scope of the present invention should be construed by the following claims, and all technologies within the scope equivalent thereto should be construed as being included in the scope of the present invention.

Claims

a pre-processing unit for performing data pre-processing on an analog temperature signal corresponding to a PLC control process;

a feature information extraction unit for extracting feature information from the pre-processed analog temperature signal; and

Process abnormality detection device through analysis of temperature signal in the control section, characterized in that it includes a process abnormality detection unit for detecting whether there is an abnormality in the PLC control process corresponding to the analog temperature signal by applying the extracted characteristic information to a machine learning model .
The method according to claim 1,

The pre-processing unit,

a signal dividing module that divides the analog temperature signal into segments; and

Process abnormality detection device through the analysis of control section temperature signal, characterized in that it comprises a loss recovery module for restoring the loss section for the analog temperature signal divided into the segment unit.
3. The method according to claim 2,

The signal dividing module,

Process abnormality detection device through analysis of temperature signal in the control section, characterized in that dividing the product production cycle unit as the segment unit in the analog temperature signal.
3. The method according to claim 2,

The signal dividing module,

In the analog temperature signal, each detailed process unit in the product production cycle is divided as the segment unit.
3. The method according to claim 2,

The loss recovery module,

Between the end point temperature data of the sampling temperature data of the first segment signal belonging to the segment signals in which the analog temperature signal is divided by the segment unit and the start point temperature data of the sampling temperature data of the second segment signal adjacent to the first segment signal Process anomaly detection apparatus through analysis of a temperature signal in a control section, characterized in that by using a linear relationship, the sampling temperature data of the loss section for the divided first segment signal or the second segment signal divided in the segment unit is restored.
6. The method of claim 5,

The loss recovery module,

A control characterized in that each of the segment signals divided in the segment unit is divided into a predetermined sampling period, and the sampling temperature data divided by the predetermined sampling period is restored using a linear relationship between the original sampling temperature data. Process abnormality detection device through analysis of section temperature signal.
The method according to claim 1,

The feature information extraction unit,

In the analog temperature signal, the total integral area, bow-shaped integral area, Y-axis parallel movement integral area, temperature rise section integral area, temperature fall section integral area, start and end slope, start maximum displacement difference, highest end in one product production cycle unit in the analog temperature signal Process anomaly detection device through analysis of temperature signal in control section, characterized in that at least one of displacement difference, mean, standard deviation, root mean square, and shape coefficient is extracted as the characteristic information.
8. The method of claim 7,

The feature information extraction unit,

Process anomaly detection apparatus through analysis of temperature signal in control section, characterized in that by using a random forest algorithm among machine learning methods, some of the extracted feature information is selected.
performing data pre-processing on an analog temperature signal corresponding to a PLC control process;

extracting characteristic information from the pre-processed analog temperature signal; and

and detecting whether there is an abnormality in the PLC control process corresponding to the analog temperature signal by applying the extracted characteristic information to a machine learning model.
10. The method of claim 9,

The data pre-processing step includes:

dividing the analog temperature signal into segments; and

and restoring a loss section for the analog temperature signal divided into segments.
11. The method of claim 10,

The step of dividing into segments is

A process abnormality detection method through analysis of a temperature signal in a control section, characterized in that dividing a product production cycle unit as the segment unit in the analog temperature signal.
11. The method of claim 10,

The step of dividing into segments is

Process abnormality detection method through the analysis of the temperature signal in the control section, characterized in that in the analog temperature signal, each detailed process unit in the product production cycle is divided as the segment unit.
11. The method of claim 10,

Restoring the lost section comprises:

Between the end point temperature data of the sampling temperature data of the first segment signal belonging to the segment signals in which the analog temperature signal is divided by the segment unit and the start point temperature data of the sampling temperature data of the second segment signal adjacent to the first segment signal Process anomaly detection method through analysis of temperature signal in control section, characterized in that the sampling temperature data of the loss section with respect to the divided first segment signal or the second segment signal divided in the segment unit is restored using a linear relationship.
14. The method of claim 13,

Restoring the lost section comprises:

A control characterized in that each of the segment signals divided in the segment unit is divided into a predetermined sampling period, and the sampling temperature data divided by the predetermined sampling period is restored using a linear relationship between the original sampling temperature data Process abnormality detection method through analysis of section temperature signal.
10. The method of claim 9,

The step of extracting the feature information,

In the analog temperature signal, the total integral area, bow-shaped integral area, Y-axis parallel movement integral area, temperature rise section integral area, temperature fall section integral area, start end slope, start maximum displacement difference, highest end in one product production cycle unit in the analog temperature signal A method for detecting process anomalies through analysis of a temperature signal in a control section, characterized in that at least one of displacement difference, mean, standard deviation, root mean square, and shape coefficient is extracted as the characteristic information.
16. The method of claim 15,

The step of extracting the feature information,

A process abnormality detection method through analysis of a temperature signal in a control section, characterized in that by using a random forest algorithm among machine learning methods, some of the extracted characteristic information is selected.