Exhaust pump failure prediction method and pump failure prediction system
The pump failure prediction system, utilizing sensors and deep learning, addresses blockages in the PS Exhaust Line by predicting failures in the exhaust pump, ensuring continuous ALD process operation.
Patent Information
- Application Number
- KR1020210002503
- Authority / Receiving Office
- KR · KR
- Patent Type
- Patents
- Current Assignee / Owner
- Filing Date
- 2021-01-08
- Publication Date
- 2026-11-11
- Estimated Expiration
- 2041-01-08
AI Technical Summary
The accumulation of residues in the Pump-Scrubber Exhaust Line (PS Exhaust Line) can lead to blockages, potentially causing exhaust pump failure, which halts the entire ALD process.
A pump failure prediction system comprising a deposition chamber, exhaust gas trap device, exhaust pump, and exhaust gas scrubber, equipped with sensors to measure pressure and temperature, and a monitoring device that generates warning signals by comparing collected data with stored data using deep learning technology to predict potential failures.
The system effectively predicts exhaust pump failures by considering various conditions, allowing for proactive maintenance and preventing process halts by identifying precursor phenomena through data comparison and deep learning.
Smart Images

Figure 112021002543067-PAT00001_ABST
Abstract
Description
Technology Field
[0001] The present invention relates to an exhaust pump failure prediction method and a pump failure prediction system, and more specifically, to an exhaust pump failure prediction method and a pump failure prediction system including a prediction algorithm. Background Technology
[0002] Residues generated during the Atomic Layer Deposition (ALD) process pass through an exhaust gas trap and an exhaust pump, and are incinerated by an exhaust gas scrubber. During this process, deposited material continuously accumulates in the Pump-Scrubber Exhaust Line (PS Exhaust Line) located between the exhaust pump and the scrubber, which can eventually block the flow of residue processing. If the flow in the exhaust line becomes blocked, the exhaust pump, which provides the hydraulic pressure to move the residues, may fail, resulting in a problem where the entire process is halted. The problem to be solved
[0003] One objective of the present invention is to provide a pump failure prediction system including a monitoring device capable of accurately predicting the failure of an exhaust pump.
[0004] Another objective of the present invention is to provide a method for predicting exhaust pump failure using the monitoring device. means of solving the problem
[0005] A pump failure prediction system according to exemplary embodiments for achieving one objective of the present invention comprises: a deposition chamber; an exhaust gas trap device connected to the deposition chamber by a first exhaust line for capturing residue contained in gas discharged from the deposition chamber; an exhaust pump connected to the exhaust gas trap device by a second exhaust line for providing hydraulic pressure for moving the residue; an exhaust gas scrubber connected to the exhaust pump by a third exhaust line for processing the residue; a trap device sensor for measuring the pressure of the exhaust gas trap device; and a pump sensor for measuring the temperature and pressure of the exhaust pump; and a monitoring device comprising a control unit for generating a warning signal when it is determined that there is a possibility of a failure situation occurring by comparing collected data obtained by the sensor unit with stored data including data regarding the thickness and shape of a deposit accumulated inside the third exhaust line.
[0006] In a method for predicting failure of a pump failure prediction system according to exemplary embodiments for achieving one objective of the present invention, data regarding the thickness and shape of a deposit accumulated inside a third exhaust line connecting an exhaust pump and an exhaust gas scrubber is obtained in advance. Data regarding a state value including the pressure of an exhaust gas trap device connected to a deposition chamber by a first exhaust line, and a state value including the temperature and pressure of the exhaust pump, are collected. The collected data is compared with stored data containing the previously obtained data to determine whether there is a possibility of a failure situation occurring. If a failure situation is determined, a warning signal is generated. Effects of the invention
[0007] According to exemplary embodiments, a pump failure prediction system comprises a deposition chamber, an exhaust gas trap device connected to the deposition chamber via a first exhaust line for temporarily storing residue, an exhaust pump connected to the exhaust gas trap device via a second exhaust line for providing hydraulic pressure for moving the residue, an exhaust gas scrubber connected to the exhaust pump via a third exhaust line for processing the residue, a sensor unit including a trap device sensor for measuring the pressure of the exhaust gas trap device and a pump sensor for measuring the temperature and pressure of the exhaust pump, and a monitoring device having a control unit that generates a warning signal when a dangerous situation is determined by comparing collected data measured by the sensor unit with existing stored data.
[0008] Accordingly, the pump failure prediction system can predict failures of the pump failure prediction system by comprehensively considering various conditions, such as the temperature and pressure of the exhaust pump, as well as the condition of the exhaust gas trap device, the condition of the third exhaust line, and the condition of the exhaust gas scrubber. Furthermore, by using deep learning technology, it can efficiently predict precursor phenomena of failure by comparing stored data collected from the past with currently collected data.
[0009] However, the effects of the present invention are not limited to the effects mentioned above and may be extended in various ways without departing from the spirit and scope of the present invention. Brief explanation of the drawing
[0010] FIG. 1 is a conceptual diagram showing a pump failure prediction system according to exemplary embodiments. FIG. 2 is a configuration diagram showing a monitoring device according to exemplary embodiments. Figure 3 is a graph showing data stored in a data storage unit. Figure 4 is a graph showing data collected by the sensor unit. FIG. 5 is a flowchart illustrating a method for implementing a prediction algorithm according to exemplary embodiments. Figure 6 is a flowchart illustrating the data collection step of Figure 5. Figure 7 is a flowchart illustrating the data comparison step of Figure 5. Specific details for implementing the invention
[0011] Hereinafter, preferred embodiments of the present invention will be described in detail with reference to the attached drawings.
[0012] FIG. 1 is a conceptual diagram showing a pump failure prediction system according to exemplary embodiments. FIG. 2 is a configuration diagram showing a monitoring device according to exemplary embodiments.
[0013] Referring to FIGS. 1 and 2, the pump failure prediction system (10) may include a deposition chamber (500), an exhaust gas trap device (200), an exhaust pump (300), and an exhaust gas scrubber (400). Additionally, the pump failure prediction system (10) may additionally include a monitoring device (100).
[0014] The deposition chamber (500) may be an Atomic Layer Deposition (ALD) device for forming a deposition film on a semiconductor wafer. Alternatively, the deposition chamber (500) may be a Chemical Vapor Deposition (CVD) device. For example, the deposition chamber (500) may be a deposition facility for a Titanium Nitride (TiN) process, a Zirconium (ZrO) process, or an Aluminum Oxide (AlO) process. A deposition material for forming a deposition film may be present inside the deposition chamber (500). After the deposition process is completed, any remaining material and gas may be discharged from the deposition chamber (500) through an exhaust port. The discharged gas may be introduced into an exhaust gas trap device (200) through a first exhaust line (600).
[0015] The exhaust gas trap device (200) can collect residual material generated inside the deposition chamber (500) or precipitate generated by a chemical reaction inside the pump failure prediction system (10) through a trap plate (not shown) having a plurality of flow holes and temporarily store it in an internal space.
[0016] If residual material and powder flow into the exhaust pump (300) in excess of the amount that the exhaust pump (300) can accommodate, it may cause the exhaust pump (300) to fail; therefore, the exhaust gas trap device (200) can regulate the amount that the exhaust pump (300) can accommodate and deliver residual material, gas and powder to the exhaust pump (300) through the second exhaust line (700). Alternatively, the exhaust gas trap device (200) can collect residual material and powder using a trap plate (not shown) and discharge it to the outside of the pump failure prediction system (10) through a separate opening and closing device. For example, the material temporarily stored in the exhaust gas trap device (200) may include precipitate (TiCl4nNH3), ammonium chloride (NH4Cl), titanium tetrachloride (TiF4), titanium nitride (TiN), etc.
[0017] The exhaust pump (300) can provide hydraulic pressure to discharge residual material and gas generated in the deposition chamber (500) to the outside of the deposition chamber (500). Additionally, the exhaust pump (300) can provide hydraulic pressure to transfer residual material and gas discharged from the deposition chamber (500) to the exhaust gas scrubber (400). The exhaust pump (300) can form a vacuum using a motor rotating inside, and can generate the hydraulic pressure through this vacuum. The exhaust pump (300) can provide the hydraulic pressure by being directly connected to the second exhaust line (700) and the third exhaust line (800). However, if residual material, powder, and gas exceeding the capacity are supplied from the exhaust gas trap device (200), the motor may be strained. For example, the exhaust pump (300) may be a pump with a capacity of 20K, 30K, 50K, or 100K. The exhaust pump (300) may include a rotary pump or a dry pump.
[0018] The exhaust gas scrubber (400) can be used to safely process residual materials generated during the deposition process without prior reaction, and to process gases containing toxic gases (NF3, ClF4) generated during the cleaning process to remove contaminants remaining on the inner wall of the deposition chamber (500) after the process is performed. The exhaust gas scrubber (400) may include all types of scrubbers commonly used for processing exhaust gases in the field of semiconductor manufacturing equipment technology. The exhaust gas scrubber (400) may be directly connected to the third exhaust line (800) to receive residual materials and gases from the exhaust pump (300), and may receive oxidizing agents, reducing agents, gases, etc., for processing residual materials through a separate line.
[0019] The exhaust gas scrubber (400) can classify the above residual substances and gases according to the respective required treatment processes, and can purify them into a treatable state through processes such as catalytic oxidation, incineration, and adsorption, and then discharge them to the outside. For example, the exhaust gas scrubber (400) can apply heat to the precipitate powder (TiCl4nNH3) to convert it into a treatable state through thermal decomposition and discharge it, or it can be discharged by adsorbing it onto a chemical agent that absorbs harmful substances. Alternatively, it can use calcium hydroxide (Ca(OH)2) to chemically react with nitrogen trifluoride (NF3), chlorine trifluoride (ClF4), etc., convert them into a treatable state and treat them.
[0020] In exemplary embodiments, the monitoring device (100) may include a sensor unit (900). The sensor unit (900) may include a trap device sensor (210) and a pump sensor (310). Furthermore, the sensor unit (900) may additionally include an exhaust gas scrubber sensor (410) and a deposition chamber sensor (510).
[0021] The monitoring device (100) can collect information on the status of the exhaust gas trap device (200), exhaust pump (300), exhaust gas scrubber (400), and deposition chamber (500) through the trap device sensor (210), pump sensor (310), scrubber sensor (410), and deposition chamber sensor (510) constituting the sensor unit (900).
[0022] The trap device sensor (210) is located inside the exhaust gas trap device (200) and can measure the pressure of the exhaust gas trap device. Additionally, the trap device sensor (210) can measure the temperature inside the exhaust gas trap device (200), the amount of residual material present inside the exhaust gas trap device, and the amount of sediment collected on the trap plate. For example, if residual material and gas exceeding the capacity of the exhaust gas trap device (200) flow into the exhaust gas trap device (200), the trap device sensor (210) can detect this.
[0023] The pump sensor (310) is located inside the exhaust pump (300) and can measure the temperature and pressure of the exhaust pump. Additionally, the pump sensor (310) can measure the current and frequency inside the exhaust pump (300). For example, if residual material and gas exceeding the capacity of the exhaust pump (300) flow into the interior of the exhaust pump (300), the pump sensor (310) can detect this. The pump sensor (310) can detect sudden changes in the temperature, pressure, current, and frequency of the exhaust pump.
[0024] The exhaust gas scrubber sensor (410) is located inside the exhaust gas scrubber (400) and can measure the lifespan of the agent in the exhaust gas scrubber (400). For example, the exhaust gas scrubber agent (not shown) may be an adsorbent for treating exhaust gas. Since the exhaust gas scrubber agent cannot perform its adsorption function normally if it exceeds its adsorption capacity, it may need to be replaced periodically. Therefore, when the lifespan of the exhaust gas scrubber agent is exhausted because precipitates have been adsorbed in excess of the adsorption capacity, the exhaust gas scrubber sensor (410) can detect this.
[0025] The deposition chamber sensor (510) is located inside the deposition chamber (500) and can measure the pressure inside the deposition chamber (500). For example, if the pressure inside the deposition chamber (500) rises above an allowable range, the deposition chamber sensor (510) can detect this.
[0026] In exemplary embodiments, the monitoring device (100) may include a control unit (1000). The control unit (1000) may include a data input unit (150), a data processing unit (110), a data storage unit (120), a data receiving unit (130), and an output unit (140).
[0027] The control unit (1000) can collect data regarding the exhaust gas trap device (200), pump (300), exhaust gas scrubber (400), and deposition chamber (500) through the trap device sensor (210), pump sensor (310), exhaust gas scrubber sensor (410), and deposition chamber sensor (510) constituting the sensor unit (900). The control unit (1000) can receive data from the sensor unit (900) through the data receiving unit (130). Additionally, the control unit (1000) can input data through the data input unit (150). The collected data and input data can be compared with existing data stored in the data storage unit (120) through the data processing unit (110).
[0028] In exemplary embodiments, the data input unit (150) may additionally input the thickness and shape of the sediment accumulated inside the first exhaust line (600) and the second exhaust line (700). For example, the shape may refer to the shape of the area formed by the sediment settling inside the first exhaust line (600) and the second exhaust line (700), and may mean an area through which residual material and gas can move.
[0029] For example, the first exhaust line (600) may have a diameter of 90 phi to 110 phi. The second exhaust line (700) may have a diameter of 40 phi to 60 phi. Residual materials and gases generated in the semiconductor process may pass through the interior of the first exhaust line (600) and the second exhaust line (700). The residual materials and gases may include materials such as TiCl4, and these materials may react with ammonia (NH3) to be replaced by a precipitate (TiCl4nNH3). Alternatively, the precipitate may include ammonium chloride (NH4Cl), titanium tetrachloride (TiF4), titanium nitride (TiN), etc. These precipitates may settle and accumulate from the side walls of the first exhaust line (600) and the second exhaust line (700).
[0030] Due to the sedimentation of precipitates, the area of the gas flow passage through which residues can move within the first exhaust line (600) and the second exhaust line (700) may be reduced, and the pressure within the first exhaust line (600) and the second exhaust line (700) may increase. Consequently, problems may arise in transferring the residues and gases supplied from the deposition chamber (500) to the exhaust gas trap device (200), or in transferring the residues and gases supplied from the temporary storage unit (200) to the exhaust pump (300). As the pressure within the first exhaust line (600) and the second exhaust line (700) increases, the pressure of the exhaust pump (300) may also increase, which may cause the exhaust pump (300) to malfunction.
[0031] To prevent this, the process may be periodically stopped and the first exhaust line (600) and the second exhaust line (700) opened to remove the internal sediment. At this time, the thickness and shape of the sediment deposited inside the first exhaust line (600) and the second exhaust line (700) may be measured and converted into data. For example, the shape may refer to the shape of the area formed by the sediment deposited inside the first exhaust line (600) and the second exhaust line (700), and may represent an area through which residual material and gas can move. This data may be input through the data input unit (150) and stored in the data storage unit (120), and may be used for data comparison by the data processing unit (110), which will be described later, and for predicting the state of the first exhaust line (600) and the second exhaust line (700).
[0032] In exemplary embodiments, the data input unit (150) may input the thickness and shape of the sediment accumulated inside the third exhaust line (800). For example, the shape may refer to the shape of the area formed by the sediment settling inside the third exhaust line (800), and may mean an area through which residual material and gas can move.
[0033] For example, the third exhaust line (800) may have a diameter of 30 to 50 pi. Residual materials and gases generated in the semiconductor process may pass through the interior of the third exhaust line (800). The residual materials and gases may include substances such as TiCl4, and these substances may react with ammonia (NH3) to be replaced by precipitates (TiCl4nNH3). Alternatively, the precipitates may include ammonium chloride (NH4Cl), titanium tetrachloride (TiF4), titanium nitride (TiN), etc. These precipitates may settle and accumulate from the sidewalls of the third exhaust line (800).
[0034] Due to the sedimentation of sediment, the area of the gas flow passage through which residue can move within the third exhaust line (800) may be reduced, and the pressure within the third exhaust line (800) may increase. Consequently, problems may arise in transferring the residue and gas supplied from the exhaust pump (300) to the exhaust gas scrubber (400). As the pressure within the third exhaust line (800) increases, the pressure of the exhaust pump (300) may also increase, which may cause the exhaust pump (300) to malfunction.
[0035] To prevent this, the process may be periodically stopped and the third exhaust line (800) opened to remove the internal sediment. At this time, the thickness and shape of the sediment deposited inside the third exhaust line (800) may be measured and converted into data. For example, the shape may refer to the form of the area formed by the sediment deposited inside the third exhaust line (800), and may represent an area through which residual material and gas can move. This data may be input through the data input unit (150) and stored in the data storage unit (120), and may be used for data comparison and state prediction of the third exhaust line (800) by the data processing unit (110) to be described later.
[0036] When the third exhaust line (800) is opened, the thickness and shape of the sediment inside the third exhaust line (800) can be measured by classifying them according to the distance from the exhaust pump (300) connected to the third exhaust line (800). For example, the third exhaust line (800) located between the exhaust pump (300) and the exhaust gas scrubber (400) can be divided into six parts, and the thickness and shape of the sediment according to each position can be measured. The inside of the third exhaust line (800) may receive less pressure and accumulate more sediment as it moves further away from the exhaust pump (300). Data on the thickness and shape of the sediment according to this distance can be stored in the data storage unit (120) through the data input unit (150) and utilized for more accurate fault prediction.
[0037] Data regarding a failure that occurred during the use of the exhaust system of the pump failure prediction system (10), that is, the exhaust pump (300), can be stored in the data storage unit (120). The data regarding the failure stored in the data storage unit (120) can be defined as stored data. For example, the stored data may include the overall state values of the pump failure prediction system (10) corresponding to the time of the failure of the exhaust pump (300). Such system states may include the pressure of the exhaust gas trap device (200), the pressure and temperature of the exhaust pump (300), the lifespan of the chemical agent in the exhaust gas scrubber (400), and the thickness and shape of the sediment accumulated inside the third exhaust line (800). Furthermore, the system states may be stored in the data storage unit (120) in a labeled state through data labeling, which will be described later.
[0038] New data can be accumulated in the data storage unit (120) while continuously using the pump failure prediction system (10). Through this iterative process, it is possible to form Big Data regarding the failure status of the pump failure prediction system (10), and as the process progresses, reliability can be increased.
[0039] The data processing unit (110) can compare the data collected from the sensor unit (900) with the stored data stored in the data storage unit (120). If the state of each component of the pump failure prediction system (10) stored in the data storage unit (120) at the time of the previous failure is the same as the state of each component of the pump failure prediction system (10) collected from the sensor unit (900), the data processing unit (110) can generate a warning signal. For example, the thickness and shape of the sediment inside the third exhaust line (800) can be predicted through the collected data to generate a warning signal. The warning signal can be transmitted to the manager through the output unit (140). Through this warning signal, the occurrence of a failure can be predicted in advance, and the situation where the entire process is stopped can be prevented.
[0040] When a failure occurs, the data processing unit (110) can label the status of the exhaust gas trap device (200), exhaust pump (300), exhaust gas scrubber (400), deposition chamber (500), first exhaust line (600), second exhaust line (700), and third exhaust line (800) constituting the pump failure prediction system (10) and record them in the data storage unit (120). At this time, the stored data can be used as data to predict failures that may occur later.
[0041] Figure 3 is a graph showing data stored in a data storage unit. Figure 4 is a graph showing data collected by a sensor unit.
[0042] Referring to FIGS. 3 and 4, the data processing unit (110) may use deep learning technology when comparing collected data with stored data. Since the amount of data collected by the sensor unit (900) and the amount of data stored in the data storage unit (120) are vast, it may be difficult to compare them individually. Furthermore, when considering the mutual influence between each sensor constituting the sensor unit (900), the amount of data may become even more massive. Therefore, by using deep learning technology, key content can be summarized from a large amount of data, and effective data comparison may be possible.
[0043] As illustrated in FIG. 3, a vast amount of data stored in the data storage unit (120) can be simplified and represented as a single graph over time. For example, the X-axis may correspond to the time axis, and the Y-axis may correspond to a value obtained by uniformly converting data such as temperature and pressure collected from the sensor unit (900). A may correspond to data such as the temperature and pressure of each facility, and the thickness and shape of the sediment inside the third exhaust line (800), which are recorded and stored at the time of the failure. B may correspond to the time when the actual process equipment fails. Data A at the time of the failure may be labeled and stored in the data storage unit (120). The data labeling may include the task of attaching a value (label) to the source data to create AI training data.
[0044] As illustrated in FIG. 4, data collected by the sensor unit (900) can be simplified and represented as a single graph over time, such as the data stored in the data storage unit (120) of FIG. 3. For example, the X-axis may correspond to the time axis, and the Y-axis may correspond to a value obtained by uniformly converting data such as temperature and pressure collected by the sensor unit (900). The collected data (C) can be compared with data (A) labeled and stored in the data storage unit (120), and if it is determined that the data is identical to A, such as C, D can be predicted as the failure prediction point, and the temperature and pressure of each facility at that time, as well as the thickness and shape of the sediment inside the third exhaust line (800), can be predicted.
[0045] When the data processing unit (110) uses deep learning technology, it may use a convolutional neural network model. For example, data collected from the sensor unit (900) can be represented as a graph as shown in FIGS. 3 and FIGS. 4. The X-axis may correspond to the time axis, and the Y-axis may correspond to a value obtained by uniformly converting data such as temperature and pressure collected from the sensor unit (900) and data input through the data input unit (150).
[0046] These graphs can be visualized using a Convolutional Neural Network model, and a vast amount of data can be simplified into an image. Therefore, data collected from the sensor unit (900) and the existing vast amount of stored data stored in the data storage unit (120) can be effectively compared.
[0047] Below, a method for predicting the failure of an exhaust pump performed by the monitoring device of FIGS. 1 and FIG. 2 will be described.
[0048] FIG. 5 is a flowchart illustrating a method for predicting failure of an exhaust pump according to exemplary embodiments. FIG. 6 is a flowchart illustrating the data collection step of FIG. 5. FIG. 7 is a flowchart illustrating the data comparison step of FIG. 5.
[0049] Referring to FIGS. 5 through 7, in exemplary embodiments, the method for predicting failure of an exhaust pump may include a data collection step (S10), a data comparison step (S20), and a warning signal generation step (S30). The method for predicting failure of an exhaust pump may further include a data labeling step (S40).
[0050] The data collection step (S10) can collect data for the exhaust gas trap device (200), exhaust pump (300), exhaust gas scrubber (400), deposition chamber (500), etc., which constitute the pump failure prediction system (10). Specifically, data can be collected through the trap device sensor (210), pump sensor (310), exhaust gas scrubber sensor (410), and deposition chamber sensor (510) which constitute the monitoring device (100).
[0051] The data comparison step (S20) can compare existing data stored in the data storage unit (120) with collected data through the data processing unit (110). The data storage unit (120) may store data regarding failures that occurred during the process of using the exhaust system of the pump failure prediction system (10), that is, the exhaust pump (300). The data regarding the failure stored in the data storage unit (120) may be defined as stored data. For example, the stored data may include the overall state values of the pump failure prediction system (10) corresponding to the time of the failure of the exhaust pump (300). Such system states may include the pressure of the exhaust gas trap device (200), the pressure and temperature of the exhaust pump (300), the lifespan of the chemical agent in the exhaust gas scrubber (400), and the thickness and shape of the sediment accumulated inside the third exhaust line (800). Furthermore, this system state may be stored in the data storage unit (120) in a labeled state through the data labeling step (S40) to be described later.
[0052] If an abnormal sign is detected by determining that the data collected through the data comparison step (S20) is identical to the data that caused the problem among the existing stored data, the process may proceed to the warning signal generation step (S30). If there is no existing data of the same problem phenomenon, the process may return to the data collection step (S10) to continuously collect data through the sensors of the pump failure prediction system (10).
[0053] In the data comparison step (S20), when comparing the collected data and the stored data, deep learning technology may be used. Since the amount of data collected by the sensor unit (900) and the amount of data stored in the data storage unit (120) are vast, it may be difficult to compare them individually. In addition, if the mutual influence between each sensor constituting the sensor unit (900) is taken into account, the amount of data may become even more massive. Therefore, by using deep learning technology, key content can be summarized from a large amount of data, and effective data comparison may be possible.
[0054] In the data comparison step (S20), if deep learning technology is used, a Convolutional Neural Network model can be used. The Convolutional Neural Network model can simplify a vast amount of collected data by visualizing it as an image through a graph. Therefore, the data comparison step (S20) can compare the collected data and the stored data more efficiently.
[0055] In the warning signal generation step (S30), if it is determined through the data comparison step (S20) that there is a possibility of a problem occurring in the exhaust system of the pump failure prediction system (10) because the same phenomenon as when the conventional failure occurred is occurring, a warning signal may be generated. For example, the phenomenon at the time of the conventional failure may include an excessive rise in the pressure of the exhaust gas trap device (200), an excessive rise in the pressure and temperature of the exhaust pump (300), the expiration of the lifespan of the chemical agent in the exhaust gas scrubber (400), and an increase in the thickness of the sediment in the third exhaust line (800).
[0056] If a warning signal is generated, the process may be stopped, and subsequent processing may be performed on the device determined to have a problem based on the collected data. For example, the subsequent processing may include replacing the chemical in the exhaust gas scrubber (400), removing foreign substances inside the exhaust gas trap device (200) or exhaust pump (300), and adjusting the pressure of the deposition chamber (500).
[0057] When a warning signal occurs, the thickness and shape of the sediment deposited inside the first exhaust line (600) and the second exhaust line (700) can be predicted through the data stored in the data storage unit (120). At this time, the process can be stopped, the first exhaust line (600) and the second exhaust line (700) can be opened, and the sediment deposited inside can be removed to prevent problems from occurring in advance. Additionally, the thickness and shape of the internal sediment at the time the first exhaust line (600) and the second exhaust line (700) are opened can be measured and the data can be stored in the data storage unit (120) through the data input unit (150). This stored data can be reused later in the data comparison step (S20).
[0058] When a warning signal occurs, the thickness and shape of the sediment deposited inside the third exhaust line (800) can be predicted through the data stored in the data storage unit (120). At this time, the process can be stopped, the third exhaust line (800) can be opened, and the sediment deposited inside can be removed to prevent problems from occurring. Additionally, the thickness and shape of the sediment inside at the time the third exhaust line (800) is opened can be measured and the data can be stored in the data storage unit (120) through the data input unit (150). This stored data can be reused later in the data comparison step (S20).
[0059] In the data labeling step (S40), the process of comparing existing data with collected data through the data comparison step (S20) is repeated. However, if a failure occurs unexpectedly or if the failure is predicted successfully, data regarding the status of the exhaust gas trap device (200), exhaust pump (300), exhaust gas scrubber (400), deposition chamber (500), third exhaust line (800), first exhaust line (600), and second exhaust line (700) constituting the pump failure prediction system (10) can be labeled and stored. The data stored in this way is stored in the data storage unit (120) and can be used later as existing data to be compared. Through this iterative process, it is possible to form Big Data regarding the failure status of the pump failure prediction system (10), and reliability can be increased as the process progresses.
[0060] As illustrated in FIG. 6, in exemplary embodiments, the data collection step (S10) constituting the method for implementing the prediction algorithm of FIG. 5 may include an exhaust gas trap device state measurement step (S14) and a pump state measurement step (S16). Furthermore, the data collection step (S10) may additionally include a deposition chamber state measurement step (S12) and an exhaust gas scrubber state measurement step (S18).
[0061] In the exhaust gas trap device state measurement step (S14), data regarding the exhaust gas trap device (200) can be collected. In the exhaust gas trap device state measurement step (S14), the pressure of the exhaust gas trap device (200) can be measured using the trap device sensor (210). Additionally, the temperature inside the exhaust gas trap device (200) and the amount of residual material present inside the exhaust gas trap device can be measured. For example, in the exhaust gas trap device state measurement step (S14), if residual material and gas exceeding the capacity of the exhaust gas trap device (200) flow into the interior of the exhaust gas trap device (200), this can be detected through the trap device sensor (210).
[0062] In the pump state measurement step (S16), data regarding the exhaust pump (300) can be collected. The pump state measurement step (S16) can measure the temperature and pressure of the exhaust pump (300) using the pump sensor (310). Additionally, the pressure, current, and frequency inside the exhaust pump (300) can be measured. For example, in the pump state measurement step (S16), if residual material and gas exceeding the capacity of the exhaust pump (300) flow into the interior of the exhaust pump (300), this can be detected through the pump sensor (310).
[0063] The deposition chamber state measurement step (S12) can measure the pressure inside the deposition chamber (500). For example, if the pressure inside the deposition chamber (500) rises above an allowable range, the deposition chamber sensor (510) can detect this.
[0064] The exhaust gas scrubber state measurement step (S18) can measure the lifespan of the agent of the exhaust gas scrubber (400). For example, the exhaust gas scrubber agent may be an adsorbent for treating exhaust gas. Since the exhaust gas scrubber agent cannot perform its adsorption function normally if it exceeds its adsorption capacity, there may be a need to replace it periodically. If the lifespan of the exhaust gas scrubber agent is exhausted during the exhaust gas scrubber state measurement step (S18), this can be detected through the exhaust gas scrubber sensor (410).
[0065] The deposition chamber state measurement step (S12), exhaust gas trap device state measurement step (S14), pump state measurement step (S16), and exhaust gas scrubber state measurement step (S18) constituting the data collection step (S10) can measure the state of the deposition chamber (500), exhaust gas trap device (200), exhaust pump (300), and exhaust gas scrubber (400) regardless of the order in the prediction algorithm for collecting data.
[0066] As illustrated in FIG. 7, in exemplary embodiments, the data comparison step (S20) constituting the method for implementing the prediction algorithm of FIG. 5 may include a deposition chamber state comparison step (S21), an exhaust gas trap device state comparison step (S22), a pump state comparison step (S24), an exhaust gas scrubber state comparison step (S26), and a pump failure prediction system state prediction step (S28).
[0067] In the deposition chamber state comparison step (S21), exhaust gas trap device state comparison step (S22), pump state comparison step (S24), and exhaust gas scrubber state comparison step (S26), the data for each deposition chamber (500), exhaust gas trap device (200), exhaust pump (300), and exhaust gas scrubber (400) collected in the data collection step (S10) can be compared with the conventional stored data recorded in the data storage unit (120).
[0068] The deposition chamber state comparison step (S21), exhaust gas trap device state comparison step (S22), pump state comparison step (S24), and exhaust gas scrubber state comparison step (S26) constituting the data comparison step (S20) can compare the states of the deposition chamber (500), exhaust gas trap device (200), exhaust pump (300), and exhaust gas scrubber (400) regardless of the order in the prediction algorithm for comparing data.
[0069] In the pump failure prediction system state prediction step (S28), the situation inside the pump failure prediction system (10) can be predicted through the compared data. In particular, by predicting the thickness and shape of the sediment deposited inside the third exhaust line (800), it is possible to predict when the third exhaust line (800) should be opened to remove the sediment.
[0070] In the data comparison step (S20), the information collected in the data collection step (S10) can be compared with the stored data to predict a failure. Specifically, if data is collected that is identical to the state of the exhaust gas trap device (200), exhaust pump (300), exhaust gas scrubber (400), and deposition chamber (500) in a state where a failure has occurred, the process can proceed to the warning signal generation step (S30) to generate a warning signal.
[0071] As described above, the method for predicting failure of the exhaust pump can predict the state of the pump failure prediction system (10) by comprehensively considering various conditions, such as the temperature and pressure of the exhaust pump (300), the pressure of the exhaust gas trap device (200), the thickness and shape of the sediment accumulated inside the third exhaust line (800), and the condition of the exhaust gas scrubber (400) agent. Furthermore, by using deep learning technology, the premonitory phenomenon of failure can be efficiently predicted by comparing the stored data collected from the past with the currently collected data.
[0072] Although the present invention has been described above with reference to embodiments thereof, those skilled in the art will understand that various modifications and changes can be made to the invention without departing from the spirit and scope of the invention as described in the following claims. Explanation of the symbols
[0073] 10: Pump failure prediction system 100: Monitoring device 110: Data processing unit 120: Data storage unit 130: Data receiving unit 140: Output unit 150: Data input section 200: Exhaust gas trap device 210: Trap device sensor 300: Exhaust pump 310: Pump sensor 400: Exhaust gas scrubber 410: Exhaust gas scrubber sensor 500: Deposition chamber 510: Deposition chamber sensor 600: First exhaust line 700: 2nd exhaust line 800: 3rd exhaust line 900: Sensor unit 1000: Control unit
Claims
Claim 1 A pump failure prediction system comprising: a deposition chamber; an exhaust gas trap device connected to the deposition chamber by a first exhaust line and configured to capture residues contained in gas discharged from the deposition chamber; an exhaust pump connected to the exhaust gas trap device by a second exhaust line and configured to provide hydraulic pressure for moving the residues; an exhaust gas scrubber connected to the exhaust pump by a third exhaust line and configured to process the residues; a sensor unit comprising a trap device sensor for measuring the pressure of the exhaust gas trap device and a pump sensor for measuring the temperature and pressure of the exhaust pump; and a control unit configured to generate a warning signal when it is determined that there is a possibility of a failure occurring by comparing collected data obtained by the sensor unit with stored data including data regarding the thickness and shape of the sediment accumulated inside the third exhaust line, wherein the shape refers to the shape of the area formed by the sediment settling inside the third exhaust line. Claim 2 A pump failure prediction system according to claim 1, wherein the trap device sensor additionally measures the temperature of the exhaust gas trap device, and the sensor unit further includes an exhaust gas scrubber sensor for measuring the chemical life of the exhaust gas scrubber and a deposition chamber sensor for measuring the pressure of the deposition chamber. Claim 3 In claim 1, the control unit is a pump failure prediction system that compares the collected data and the stored data using deep learning technology. Claim 4 In claim 3, the deep learning technology is a pump failure prediction system utilizing a convolutional neural network model. Claim 5 A pump failure prediction system according to claim 1, wherein data regarding the thickness and shape of sediment accumulated inside the third exhaust line is classified and measured according to the distance from the exhaust pump. Claim 6 A method for predicting failure in a pump failure prediction system comprising a deposition chamber, an exhaust gas trap device connected to the deposition chamber by a first exhaust line, an exhaust pump connected to the exhaust gas trap device by a second exhaust line, and an exhaust gas scrubber connected to the exhaust pump by a third exhaust line, wherein the method comprises: acquiring data in advance regarding the thickness and shape of a deposit accumulated inside the third exhaust line; collecting data regarding a state value including the pressure of the exhaust gas trap device and a state value including the temperature and pressure of the exhaust pump; determining whether there is a possibility of a failure situation occurring by comparing the collected data with stored data including the data acquired in advance; and generating a warning signal when a failure situation is determined, wherein the shape refers to the shape of an area formed by the deposition of the deposit inside the third exhaust line. Claim 7 A method for predicting failure of a pump failure prediction system, wherein collecting the data further includes collecting data on the chemical life of the exhaust gas scrubber and the pressure within the deposition chamber, and the state value of the exhaust gas trap device further includes the temperature of the exhaust gas trap device. Claim 8 A method for predicting failure of a pump failure prediction system according to claim 6, wherein comparing the collected data and the stored data includes using deep learning technology. Claim 9 In claim 8, the deep learning technology is a failure prediction method for a pump failure prediction system using a convolutional neural network model. Claim 10 In claim 6, the failure prediction method of a pump failure prediction system in which data regarding the thickness and shape of sediment accumulated inside the third exhaust line is classified, input, and stored according to the distance from the exhaust pump.
Citation Information
Patent Citations
Diagnostic system for vacuum pump
JP2005009337A
Exhaust conditioning system for semiconductor reactors
JP2007522649A
Trap apparatus
KR1020020037280A
Ventilation Structure For Semiconductor ManufacturingEquipment
KR1020030021302A
Detector for identifying residual life time ofabsorbent, gas scrubber comprising the detector andmethod of identifying residual life time of absorbent
KR1020050020358A
