TWI449907B - Method for discriminating gas leakage and system thereof - Google Patents

Description

Gas leakage discrimination method and system

The present invention relates to a gas leak discrimination method and system, and more particularly to a gas leak discrimination method and system for analyzing a sequential gas concentration signal.

As process line widths evolve to nanoscale sizes, chemical micro-contamination in the environment is often a key factor in the decline in advanced process yields. In the future, optoelectronic semiconductor manufacturing will become one of the important indicators for the new development of the optoelectronic semiconductor industry. It is necessary to determine the pollution caused by the abnormal leakage of process gas in the clean room through appropriate pollution monitoring means.

The traditional monitoring method is to use the sampler to take samples in the air for a long time, and then send it to the laboratory for analysis of the micro-polluting components in the air. Although this method can obtain reliable environmental pollution data, it takes several days to obtain a data point, and the analysis results obtained must be interpreted by experienced experts before being correctly interpreted. Under traditional monitoring methods, When the pollution occurs, the incident is solved. The time spent is less than a few days, and more than one month. The cleanliness of the production line and the yield of the product have been seriously affected.

Instant continuous monitoring technology continuously and instantaneously provides instantaneous concentration changes of various pollutants in the plant area, thereby shortening the time when the pollution event is detected. However, the judgment logic for abnormal diagnosis in the prior art can usually only be defined by setting the monitoring upper limit concentration value for individual pollutants. Due to the numerous sources of gas pollution in the plant air, including the leakage of the process machine, the gas escape during maintenance and the inhalation of external air pollutants, the absolute value of the concentration alone will cause a large number of false alarms. Generated, seriously affecting the reliability of the system. Therefore, there is a need for a gas leak detection system with a more accurate leak event discrimination mechanism to improve plant safety and product yield.

The invention provides a method for detecting a gas leak, comprising: obtaining a gas monitoring signal, the gas monitoring signal recording a change in gas concentration versus time; and using a processor to load the code to perform a detection process, the detecting The program includes: differentially processing the gas monitoring signal; smoothing the differentiated gas monitoring signal to generate a smoothing signal; obtaining a standard deviation of the differential gas monitoring signal to generate a standard deviation signal; The standard deviation signal and the smoothing signal generate an event square wave signal; each square wave period of the event square wave signal corresponds to the smoothing signal traversing the standard deviation signal until the smoothing signal passes the standard downward The elapsed time of the difference signal; the gas monitoring signal is divided into a series of event blocks according to the event square wave signal, and the occurrence period of each of the event blocks corresponds to each of the square wave periods of the event square wave signal According to the correlation between the concentration changes of the event blocks and the length of time, a classification is performed, and the event blocks are returned. At least one event group; and at least one of the number of events in an event group blocks added to obtain a leakage coefficient, wherein when the leakage coefficient is greater than a threshold value, issuing a warning gas leakage.

The invention further provides a gas leakage detecting system, comprising a sensor for sensing a gas concentration and generating a gas monitoring signal, wherein the gas monitoring signal records a change in gas concentration with time; and a memory unit stores the gas monitoring signal And a processor executing a detection process, the detecting process comprising: reading the gas monitoring signal from the memory unit; performing differential processing on the gas monitoring signal; and smoothing the differentiated gas monitoring signal Generating a smoothed signal; obtaining a standard deviation of the differential gas monitoring signal to generate a standard deviation signal; comparing the standard deviation signal and the smoothing signal to generate an event square wave signal; each of the event square wave signals The one-wave period corresponds to an elapsed time when the smoothed signal crosses the standard deviation signal until the smoothed signal crosses the standard deviation signal; the gas monitoring signal is divided into a series of event blocks according to the event square wave signal. The occurrence period of each of the event blocks corresponds to each of the square wave periods of the event square wave signal Performing a classification according to the correlation between the concentration changes of the event blocks and the length of time, classifying the event blocks into at least one event group, and adding the number of event blocks in the at least one event group, A leakage coefficient is obtained, wherein when the leakage coefficient is greater than a critical value, a gas leak warning is issued.

Since the processes of modern optoelectronics, semiconductors and related industries are mostly batch production, the gas leakage generated by the process machine will form a regular leakage gas concentration signal change phenomenon in the environment, which can be separated from other disturbing gases. The sources are scattered. The present invention proposes a gas leakage discrimination method and system according to the situation to effectively solve related problems.

1 is a block diagram of a gas leakage discrimination system 100 in an embodiment of the present invention. Referring to FIG. 1, the gas leakage discrimination system 100 includes a sensor 200, a memory unit 300, and a processor 400. The sensor 200 senses the concentration of a specific chemical gas GAS in the atmosphere and generates a gas monitoring signal 10 as an electrical signal that quantifies the concentration of the chemical gas GAS. In some embodiments, the sensor 200 can be a MOS gas sensor, a liquid electrolyte gas sensor, a solid-state electrolyte sensor (solid-state electrolyte) Gas sensor), catalytic combustion gas sensor, Fourier transform infrared spectroscopy, Ion-mobility spectrometry, etc., but is not limited thereto. The memory unit 300 in turn stores the gas monitoring signal 10 from the sensor 200. In some embodiments, the gas monitoring signal 10 is stored in the memory unit 300 in a time series, that is, the concentration of the chemical gas GAS versus time. The processor 400 further executes a discriminating program written in a code or instruction, reads the gas monitoring signal 10 stored in the memory unit 300, and performs a series of signal processing analysis to determine whether a gas leak occurs. When the discriminating program determines that the gas monitoring signal 10 indicates a gas leak, for example, due to the rupture of the machine exhaust line using the chemical gas GAS, the processor 400 further issues a warning signal ALM to notify the relevant personnel or the machine. , for subsequent security checks or processing.

Figure 2 is a block diagram showing the steps of the discriminating program executed by the processor 400 in the embodiment of Figure 1. The 3A to 3E diagrams are schematic diagrams of the waveform signal processing of the gas monitoring signal 10 in the partial steps in FIG. Referring to FIG. 2, in step S1, first, a gas monitoring signal is obtained from the memory unit 300. No. 10. FIG. 3A is an example of a gas monitoring signal 10 read in an embodiment. Then, in step S2, the gas monitoring signal 10 is differentiated to obtain a differential signal 20 as shown in FIG. 3B, and the differential signal 20 represents a rate at which the concentration of the chemical gas GAS increases or decreases with time; then, step S3 is performed. The smoothing signal 20 is smoothed to obtain a smoothing signal 30, as shown in FIG. 3C, for the purpose of reducing the misjudgment of the discriminating program. In some embodiments, the discriminating program smoothes the post-differentiation signal 20 by a moving average, that is, smoothes the differential signal 20 by averaging the gas monitoring signals 10 in the previous time interval. The time interval used for the calculation of the moving average may be, for example, 5 minutes, but is not limited thereto, and may be adjusted depending on factors such as the supply amount and risk of the chemical gas GAS. Simultaneously, in step S4, the standard deviation in the past time interval of the differential signal 20 is calculated, and a standard deviation signal 40 is obtained, which represents the degree of dispersion of the differential signal 20 in the time interval, as shown in FIG. 3D. In an embodiment, the time interval for calculating the standard deviation may be, for example, 3 hours, but is not limited thereto, and may be adjusted according to factors such as the actual operating frequency of the production line machine and the supply amount of the chemical gas GAS. Next, in step S5, the smoothed signal 30 and the standard deviation signal 40 are compared to generate a continuous event square wave signal 50, the waveform of which is shown in FIG. 3E. Here, an event is defined as the process of monitoring the gas concentration after rising from a low concentration, maintaining a high concentration for a certain period of time, and then returning to the original concentration. The discriminating program separates the band in which the concentration event occurs in the gas monitoring signal 10 by the operations of the above steps S1 to S5, and defines it as an event square wave. The event square wave signal 50 has a plurality of event square waves 55, i.e., a series of events respectively determined to be determined by the beginning of the discriminating program. The events of the square wave 55 The starting point corresponds to the data point of the smoothing signal 30 that crosses the standard deviation signal 40 from the bottom up, and the ending point corresponds to the data point of the smoothing signal 30 that crosses the standard deviation signal 40 from top to bottom (ie, When the smoothing signal 30 is greater than the standard deviation signal 40, a square wave is generated correspondingly; the event square wave signal 50 is more time-corresponding to the gas monitoring signal 10, and the gas monitoring signal is divided into a series of events according to the event square wave signal 50. The block, in which the occurrence period of the event square wave 55 defines the corresponding event block time period. The event blocks represent trends in the concentration of the chemical gas GAS in the corresponding concentration events. 4 is a schematic diagram of the comparison of the event square wave E1 and the event block C1 in an embodiment. Taking FIG. 4 as an example, the event square wave E1 timing of the event square wave signal 50 corresponds to the gas monitoring signal 10 . In the event block C1, the starting point P1 of the event square wave E1 corresponds to the data point of the smoothing signal 30 crossing the standard deviation signal 40 from the bottom up; the ending point F1 corresponds to the smoothing signal 30 from the bottom up standard. One of the data points of the difference signal 40, and the occurrence period of the event block C1 also corresponds to the period T1 of the event square wave E1 of the event square wave signal.

Then, in step S6, a classification is performed according to the correlation between the concentration change of each event block and the length of time, and each event block is classified into at least one event group. Since the gas leakage signal change of the process machine usually has a special regularity compared with other pollution sources, the present invention further discriminates the leakage events belonging to the same source to the same event according to the correlation between the analysis event blocks. Group. In turn, it can accurately diagnose the occurrence of machine leakage.

FIG. 5 is a block diagram showing the steps in which the processor 400 classifies each event block. First, in step M1, all event blocks are compared two or two, if two When the time lengths of the event blocks differ by less than a predetermined ratio and the correlation coefficient is greater than a predetermined value, the two event blocks are classified into the same event group. Figure 6 is a schematic diagram of the comparison of the event block C1 and the event block C2 in an embodiment. As shown in FIG. 6, when the comparison between the event block C1 and the event block C2 is performed, the ratio of the lengths of the occurrence periods T1 and T2 is first calculated, and if the ratio of the lengths of the lengths T1 and T2 is smaller than the predetermined schedule set by the discriminating program. For the ratio, the time length correlation between the two events is determined; and the correlation coefficient between the event block C1 and the event block C2 is calculated to be earlier by using the starting points P1 and P2 of the two events as reference points. The point in time of the end event (as in the end point F2 of the event block C2 in Fig. 6; the change in the event block C1 from the end point F2 of the event block C2 to its own end point F1 is ignored). If the correlation coefficient between the event block C1 and the event block C2 is greater than the preset value set by the discriminating program, it means that the change of the concentration between the two event blocks also has a certain degree of correlation, so the event block C1 and the event area are determined. Block C2 belongs to the same event group. If an event block does not belong to any event group, an event group is newly established and includes the event block. If an event block can be classified into more than one different event group, the event block is classified into a group to which the event block having the smallest difference in length of the event block and the correlation coefficient is the largest.

After the event group is established, in order to further clarify the possibility of leakage of the machine, it is necessary to exclude the event block that is not leaked by the machine and the corresponding event group. In the atmosphere, an abnormally high concentration of chemical gas GAS may be dissipated to the periphery of the sensor 200, or the gas leakage discrimination system 100 may be disturbed by noise from various sources to cause a surge in the gas monitoring signal 10, but The source is not a machine leak, so there is no regular trend of recurrence. For row In addition to the event group of this type, the classification further proceeds to step M2, and discriminates that when any event block occurs, more than a predetermined number of event blocks of different event groups are continuously generated, then the event block corresponding to the event block is deleted. Event group. For example, the discriminating program further determines that when an event block occurs, the 10 sets of events that occur in succession are different event groups, and the event group to which the event block belongs is deleted.

In order to exclude the event block that is not leaked by the machine and the corresponding event group, the classification further proceeds to step M3, and discriminates that after any event block occurs, no event class is classified as belonging to the same event group for more than a predetermined time. When the event block is deleted, the event group corresponding to the one event block is deleted. For example, the detection program further determines that after an event block occurs, if no leakage event occurs in the same event group as the event block within 12 hours, the event group to which the event block belongs is deleted.

To exclude the event block that is not leaked by the machine and the corresponding event group, the classification further proceeds to step M4, and deletes the event group whose number of event blocks is less than a predetermined amount.

After the classification process is completed, as shown in Fig. 2, step S7 is followed to add the number of event blocks in the event group to obtain a leakage coefficient. The leakage coefficient represents an index of whether or not a gas leak is generated after the gas leak determination system 100 analyzes the gas monitoring signal 10. When the leakage coefficient exceeds a predetermined threshold, the gas concentration event representing the same type continues to occur repeatedly. At this time, the possibility of the concentration change due to the leakage of the machine is high. Then, in step S8, the processor 400 issues a warning signal. ALM to notify the relevant personnel or machine for subsequent security checks or processing. Probability of repeated and highly correlated leakage events in the absence of instrument leakage Very low, so the leakage coefficient will not continue to increase, reducing the chance of system misjudgment.

Compared with the conventional gas leakage discriminating mode, the gas leakage discriminating system proposed by the present invention can continuously monitor the chemical gas concentration in real time, and utilizes the repeatability of the gas concentration change when the machine leaks, and analyzes the chemical gas concentration in a sequential manner and eliminates the non-existence. The misjudgment of machine leakage increases the efficiency and accuracy of gas leakage discrimination.

10‧‧‧ gas monitoring signal

20‧‧‧Differential signal

30‧‧‧Smoothing signal

40‧‧‧Standard signal

50‧‧‧Event square wave signal

55‧‧‧ event square wave

100‧‧‧Gas Leakage Identification System

200‧‧‧ sensor

300‧‧‧ memory unit

400‧‧‧ processor

ALM‧‧‧ Warning Signal

C1, C2‧‧‧ event block

F1, F2‧‧‧ end point

M1-M4‧‧‧ steps

E1, E2‧‧‧ event square wave

GAS‧‧‧ chemical gas

P1, P2‧‧‧ starting point

S1-S8‧‧‧ steps

During the period of T1, T2‧‧

1 is a block diagram of a gas leak determination system 100 in an embodiment of the present invention; and FIG. 2 is a block diagram of a discriminating program step executed by the processor 400 in the embodiment of FIG. 1; FIGS. 3A-3E The figure is a schematic diagram of the waveform signal processing of the gas monitoring signal 10 in some steps in FIG. 2; FIG. 4 is a schematic diagram of the comparison of the event square wave E1 and the event block C1 in an embodiment; 400 block diagram of the classification of each event block.

Figure 6 is a schematic diagram of the comparison of the event block C1 and the event block C2 in an embodiment.

S1-S8. . . step

Claims (12)

A method for discriminating gas leakage includes: obtaining a gas monitoring signal, wherein the gas monitoring signal records a change in gas concentration versus time; and differentially processing the gas monitoring signal; and smoothing the differentiated gas monitoring signal to generate a smoothing signal; obtaining a standard deviation of the differential gas signal after the differentiation, generating a standard deviation signal; comparing the standard deviation signal and the smoothing signal to generate an event square wave signal; each square wave of the event square wave signal The period corresponds to an elapsed time that the smoothing signal crosses the standard deviation signal until the smoothed signal crosses the standard deviation signal; the gas monitoring signal is divided into a series of event blocks according to the event square wave signal, The occurrence period of the event blocks corresponds to each of the square wave periods of the event square wave signal; and the event blocks are classified according to the correlation between the concentration changes of the event blocks and the length of time. Classified into at least one event group; and adding the number of event blocks in the at least one event group to obtain one Leak coefficient, wherein when the leakage coefficient is greater than a threshold value, issuing a warning gas leakage. The method for discriminating gas leakage according to claim 1, wherein the method further comprises smoothing the differentiated gas monitoring signal by a moving average method. The method for discriminating gas leakage according to claim 1, wherein the classification comprises: when the time lengths of any two event blocks differ by less than a predetermined ratio and the correlation coefficient is greater than a predetermined value, then the two event blocks Is classified in the same event group. The method for discriminating gas leakage as described in claim 3, wherein the classification further comprises: when any event block occurs, more than a predetermined number of event blocks of different event groups are continuously generated, Delete the event group corresponding to the one event block. The method for discriminating gas leakage as described in claim 3, wherein the classification further comprises: when any event block occurs, when an event block classified into the same event group does not occur for more than a predetermined time , delete the event group corresponding to the event block. The method for discriminating a gas leak as described in claim 3, wherein the classifying comprises: deleting an event group in which the number of event blocks in the at least one event group is less than a predetermined amount. A gas leakage discriminating system includes: a sensor that senses a gas concentration and generates a gas monitoring signal, the gas monitoring signal records a change in gas concentration versus time; a memory unit stores the gas monitoring signal; and a process The discriminating program includes: reading the gas monitoring signal from the memory unit; performing differential processing on the gas monitoring signal; smoothing the differentiated gas monitoring signal to generate a smoothing signal Obtaining a standard deviation of the gas monitoring signal after the differentiation, generating a standard deviation signal; comparing the standard deviation signal and the smoothing signal to generate an event square wave signal; each square wave period of the event square wave signal corresponds to The smoothing signal traverses the standard deviation signal until the smoothing signal traverses the elapsed time of the standard deviation signal; the gas monitoring signal is divided into a series of event blocks according to the event square wave signal, each of the events The occurrence period of the block corresponds to each of the square wave periods of the event square wave signal; Performing a classification on the relationship between the change of the concentration between the blocks and the length of time, classifying the event blocks into at least one event group; and adding the number of event blocks in the at least one event group to obtain one A leakage coefficient, wherein when the leakage coefficient is greater than a critical value, a gas leak warning is issued. The gas leakage discriminating system of claim 7, further comprising smoothing the differentiated gas monitoring signal by a moving average method. The gas leakage discriminating system according to claim 7, wherein the classification comprises: when the time lengths of the two event blocks differ by less than a predetermined ratio and the correlation coefficient is greater than a predetermined value, then the two event blocks are Classified into the same event group. The gas leakage discriminating system according to claim 9, wherein the classification further comprises: deleting more than a predetermined number of event groups of different event groups after any event block occurs, deleting The event group corresponding to the event block. The gas leakage discriminating system according to claim 9, wherein the classification further comprises: when any event block occurs, when an event block classified into the same event group does not occur for more than a predetermined time, Then delete the event group corresponding to the one event block. The gas leakage discriminating system of claim 9, wherein the classifying comprises: deleting an event group in which the number of event blocks in the at least one event group is less than a predetermined amount.