Method, system, and non-transitory computer-readable medium of parameter sensing and computer modeling for gas delivery health monitoring

Abstract

A method includes: receiving measurement data from a plurality of sensors positioned along a delivery line that delivers liquid as a gas to either a gas panel or a processing chamber; simulating one or more processing parameters associated with the delivery line and a plurality of heater sheaths positioned around the delivery line using a computer-generated model; comparing the measurement data with values ​​of the one or more processing parameters; and determining, based on at least a threshold deviation between the measurement data and the values ​​of the one or more processing parameters, the existence of a fault associated with maintaining the temperature within the delivery line consistent with the gaseous state of the liquid.

Description

Technical Field

[0001] This manual generally pertains to the delivery of gas to the processing chamber. More specifically, this manual pertains to parameter sensing and computer modeling for health monitoring of gas delivery. Prior Technology

[0002] The fabrication of modern materials often involves various deposition techniques, such as chemical vapor deposition (CVD) or physical vapor deposition (PVD), in which one or more types of atoms or molecules of a chosen type are deposited on a wafer (substrate) held in a low or high vacuum environment provided by a vacuum processing chamber (e.g., deposition, etching, etc.). For example, CVD deposition processes are used in a wide range of applications. These applications range from patterned films to insulating materials in transistor structures and between conductive metal layers forming circuits. Applications include shallow trench isolation, pre-metal dielectrics, intermetallic dielectrics, and passivation. They are also used in strain engineering, which uses compressive or tensile stress films to enhance transistor performance through improved conductivity. Depending on the type of film to be deposited on the substrate, a precursor (gas or liquid) is delivered to a processing chamber where thermal oxidation or reaction results in the deposition of the desired film.

[0003] Precursors (and other liquids) induce chemical reactions during semiconductor manufacturing by introducing various types of reactive gases into the processing (or reaction) chamber. In CVD processes, there is an increasing preference for using liquid precursors instead of gases. The popularity of liquid precursors is partly based on their physical properties, which make them less hazardous, flammable, corrosive, and toxic than gaseous precursors. For example, one of the more common liquids used in the manufacture of semiconductor devices is tetraethyl orthosilicate (TEOS), which frequently replaces silane. Using TEOS, conformal silica (SiO₂) films without detectable defects can be deposited, exhibiting better step coverage and significantly less hazard compared to silane. In metal-organic CVD (MOCVD) processes, liquid precursors for metals such as copper are frequently used because gaseous precursors are unavailable.

[0004] Because the liquid precursor (and other processing liquids) are the first liquids, they are converted into a gaseous state to be transported as gas through the delivery pipeline. Failures in the manufacturing operation or delivery pipeline environment (especially failures affecting temperature and pressure) can cause some of the gas in the delivery pipeline to condense. This condensation can absorb particles and carry them into the processing chamber and deposit particles on the substrate, thereby causing particle defects on the substrate and any resulting manufacturing apparatus. Summary of the Invention

[0005] In one implementation, a method is disclosed that includes receiving measurement data from multiple sensors positioned along a delivery line that delivers liquid as a gas to either a gas panel or a processing chamber. The method further includes using a computer-generated model executed on a processing device to simulate one or more processing parameters associated with the delivery line and a plurality of heater sheaths positioned around the delivery line. The method further includes comparing the measurement data with values ​​of the one or more processing parameters by the processing device. The method further includes determining, by the processing device, the presence of a fault associated with maintaining temperature consistency with the gaseous state of the liquid within the delivery line based on at least a threshold deviation between the measurement data and the values ​​of the one or more processing parameters.

[0006] In another implementation, a system is disclosed, including a memory and a processing device operatively coupled to the memory for receiving measurement data from a plurality of sensors positioned along a processing subsystem that delivers liquid as a gas into a processing chamber. The processing device further executes a computer-generated model to simulate one or more processing parameters associated with the processing subsystem. The processing device further compares the measurement data with values ​​of the one or more processing parameters. The processing device further determines the presence of a fault associated with maintaining temperature and pressure within the processing subsystem consistent with the gaseous state of the liquid, based on at least a threshold deviation between the measurement data and the values ​​of the one or more processing parameters.

[0007] In another implementation, a non-transitory computer-readable storage medium is disclosed that stores instructions, which, when executed by a processing device, cause the processing device to perform a plurality of operations, including receiving measurement data from a plurality of sensors positioned along a delivery line that delivers liquid as a gas into a processing chamber. The operations further include simulating one or more processing parameters associated with the delivery line and a plurality of heater sheaths positioned around the delivery line by executing a computer-generated model. The operations further include comparing the measurement data with the values ​​of the one or more processing parameters. The operations further include determining the presence of a fault associated with maintaining temperature and pressure within the delivery line consistent with the gaseous state of the liquid based on at least a threshold deviation between the measurement data and the values ​​of the one or more processing parameters. Simple Explanation of the Diagram

[0008] [No.] [1A] [Figure] illustrates an exemplary implementation of health monitoring within a liquid conversion and gas delivery system to a processing chamber, according to some implementation methods.

[0009] [No.] [1B] [Figure] is an exemplary cross-sectional view of a portion of a delivery line including multiple heater sheaths positioned around the delivery line, according to some implementations.

[0010] [No.] [1C] [Figure] shows delivery lines positioned inside and outside the gas panel according to one implementation.

[0011] [No.] [2A] [Figure] is an exemplary cross-section of a dual delivery pipeline carrying different gases according to some implementations.

[0012] [No.] [2B] [Figure] is a graph illustrating an example of how steam pressure changes with the temperature of tetraethyl orthosilicate (TEOS) in a delivery pipeline, according to one implementation method.

[0013] [No.] [3] [Figure] is a flowchart of a method for monitoring the health of a gaseous liquid being delivered to a processing chamber, based on some implementation methods.

[0014] [No.] [4] [Figure] is a flowchart of a method for generating and calibrating computer-generated models according to some implementation methods.

[0015] [No.] [5] [Figure] is a flowchart of a method for training a machine learning model to predict faults in liquid conversion and gas delivery systems, based on some implementations.

[0016] [No.] [6] [Figure] is a collection of graphs showing sensor measurement data compared to a computer-generated model with one or more processing parameters, and from which patterns of change can be generated, according to one implementation method.

[0017] [No. 7] [Figure] A block diagram depicting an example computing device that operates in one or more states according to this disclosure and is capable of parameter sensing and computer modeling for gas transport health monitoring according to various implementations. Implementation

[0018] The implementation disclosed in this paper provides parameter sensing and computer modeling for health monitoring of precursors and other gas deliveries. For example, precursors (such as TEOS) are liquids with a boiling point of 168°C at room temperature. Therefore, TEOS is typically delivered as a gaseous precursor to the processing chamber after passing through a gas panel. Methods such as boiling, foaming, or injection are used to convert the precursor into a gaseous state. To prevent condensation of the gaseous precursor, the delivery line is heated by using multiple heater jackets wrapped around the line. Improper installation of any heater jacket, heater jacket failure (physical damage or control issues), or inconsistencies in heater jacket installation after preventative maintenance can all cause the gas pressure to drop below its saturation pressure and condensation in the line. Condensation subsequently leads to particle defects, potential wafer waste, and lengthy and costly unplanned maintenance procedures.

[0019] Although this application relates to precursors by way of example, this disclosure is applicable to any gas that will be maintained within a temperature range to prevent condensation back into a liquid state. For example, the gas may be located in a fore-gas line (where gas is entering the processing chamber) or an exhaust gas line (where processed gas is leaving the processing chamber), and the gas deposited as a liquid will be avoided within the gas line and elsewhere in the gas delivery system. Preventing this type of conversion to liquid and the corresponding deposition within the delivery line can prevent blockages or other adverse pumping performance.

[0020] Therefore, to address these shortcomings in current technologies employing gaseous liquids, this implementation tracks parameters related to maintaining the liquid in this gaseous state. For example, these parameters may be one or more of the following: temperature, pressure, gas concentration, gas flow rate, and / or the power output of the heater sheath. More specifically, this implementation uses a computing device to calibrate or machine learn a computer-generated model that is based on physics and characterizes the physical system of the delivery pipeline and heater sheath over time by predicting one or more of these parameters.

[0021] In various implementations, the computing device may then compare the predicted parameter value with measurement data received from multiple sensors positioned (e.g., attached) along the delivery line. If there is a change in the predicted value and the measurement data value for a parameter that at least meets a threshold deviation, the computing device detects a fault. In some implementations, predicted and measurement values ​​for more than one parameter may be combined. The computing device may further determine (or identify) a deviation pattern of this threshold deviation between the measurement data and the predicted value of the parameter. The computing device may compare the deviation pattern with predetermined deviation patterns previously associated with various faults to determine the specific fault. The computing device may further warn the operator of the specific fault and, if appropriate, suspend processing when the specific fault is resolved and the operator clears the process to continue.

[0022] This approach can be further applied to larger assemblies of physical processing subsystems, such as those including delivery lines, heater sheaths positioned around the delivery lines, gas panels, and processing chambers. In this way, the computer-generated model can be extended for preventative maintenance, fault detection, or automated diagnostics of the gaseous state of precursors or other gases, both outside and inside the processing chamber.

[0023] In some implementations, the computer-generated model integrates statistical analysis of measurement data received from sensors. Furthermore, the computing device can further utilize measurement data from sensors located along the delivery pipeline (or processing subsystem) to perform reinforcement learning on the computer-generated (e.g., machine learning) model. Reinforcement learning and / or statistical analysis enable the updating of the computer-generated model and the tracking of trends in parameters across sections, bends, and areas of the delivery pipeline (in addition to other components across the processing subsystem, such as gas panels and processing chambers).

[0024] The disclosed implementations involve various manufacturing techniques using processing chambers (which may include deposition chambers, etching chambers, and the like), such as chemical vapor deposition (CVD), physical vapor deposition (PVD), plasma-enhanced CVD, plasma-enhanced PVD, sputtering deposition, atomic layer CVD, combustion CVD, catalytic CVD, evaporation deposition, molecular beam epitaxy, and so on. The disclosed implementations can be employed in techniques using vacuum deposition chambers (e.g., ultra-high vacuum CVD or PVD, low-pressure CVD, etc.) as well as in atmospheric pressure deposition chambers.

[0025] [1A] [Figure] illustrates an exemplary implementation of health monitoring within a liquid conversion and gas delivery system 100 leading to a processing chamber, according to some implementations. The liquid conversion portion of system 100 may include a carrier gas line 102 and a drive gas line 104. The carrier gas line 102 may include a flow control valve and a mass flow controller (MFC) to allow a carrier gas (typically an inert gas) to enter an evaporator 110. The drive gas line 104 may include a flow control valve, a liquid storage container 114, and a liquid flow meter (LFM) to also deliver liquid to the evaporator 110. The evaporator 110 may subsequently convert the liquid (such as a precursor) into a gas by boiling, foaming, or injecting the carrier gas into the liquid precursor while simultaneously heating the gas with an ambient heater.

[0026] In various implementations, the liquid conversion and gas delivery system 100 includes a gas delivery system 118, which in turn includes an assembly of delivery lines 120 preceding and following the gas panel 124, and a plurality of heater sheaths 128 positioned around the assembly of delivery lines 120. For example, the assembly of delivery lines 120 may include an assembly of facility gas lines 120A from a gas source within the facility (e.g., the evaporator 110 in the disclosed implementation) coupled to the gas panel 124, and an assembly of chamber gas lines 120B coupled between the gas panel 124 and the processing chamber 115 (or multiple processing chambers).

[0027] The heater sheath 128 is shown in a simplified manner and separated from the assembly of the delivery lines 120, so as not to obscure the illustration of these components. However, [No.] [1B] [Figure] is an exemplary cross-sectional view of a portion of a delivery line 120, including a plurality of heater sheaths 128A, 128B, 128C, and 128D positioned around the delivery line 120, according to some implementations. In other words, the plurality of heater sheaths completely wrap around the outer surface of the delivery line 120. The length of the delivery line 120 can therefore be covered by these heater sheaths 128 to maintain sufficient heating of the delivery line 120 to keep the gas in a gaseous state.

[0028] In these implementations, the heater housings 128A, 128B, 128C, and 128D will each typically include one or more heating elements 132 connected to a power source and have a power output that can be tracked as measurement data. Each heater housing 128 will also typically include at least one thermocouple 136 (e.g., a temperature sensor) to track the temperature of the heater housing. In some implementations, the heater housing includes a function coupled between the power source and the heater housing, similar to closed-loop control of a thermostat. When the heater housing 128 reaches a target temperature, the thermocouple 136 flattens the power supplied to the heater housing. The thermocouple 136 can then periodically measure the temperature of the heater housing, and when the temperature drops below a threshold temperature value, the power supplied to the heater housing can be increased again until the target temperature is reached.

[0029] [C1] [Figure] illustrates an assembly of delivery lines 120 positioned inside and outside a gas panel 124 according to one implementation. As shown, the assembly of delivery lines 120 entering the gas panel 124 may be a plurality of facility gas lines 120A or feeders running from facility gas lines as discussed. The assembly of delivery lines 120 exiting the gas panel 124 may be an assembly of chamber gas lines 120B coupled between the gas panel 124 and a processing chamber (such as processing chamber 115). The ends of the chamber gas lines 120B may be attached to chamber gas feeders 138, which are attached to a plurality of processing chambers.

[0030] In some embodiments, the assembly of delivery lines 120 is identified by segments or regions and includes individual bends, curves, and other changes of direction. These non-linear characteristics of the assembly of delivery lines 120 make it more challenging to completely and reliably cover the entire outer surface of the assembly of delivery lines 120 with heater sheaths 128. Additionally, heater sheaths 128 are removed from the assembly of delivery lines 120 during preventative (or required ad hoc) maintenance. If these heater sheaths 128 are not properly reinstalled around the assembly of delivery lines 120 in the sometimes congested space of the gas delivery system 118, the delivery lines 120 may still remain at least partially exposed to ambient air. This exposure can lead to incorrect temperature variations in those areas and therefore a risk of gas condensing back into a liquid state within at least a portion of the delivery lines.

[0031] As a specific example, [No.] [2A] [Figure] is an exemplary cross-section of a dual delivery line 220 carrying different gases according to some implementations. The dual delivery line 220 includes a top delivery line carrying TEOS and argon (Ar) gases and a bottom delivery line carrying nitrous oxide (N₂O). Note that the various light-colored arrows indicate heat flow from a heater sheath 228 positioned around the dual delivery line 220. For example, heat flows from the heater sheath 228 to the stainless steel (or other metal) conduit of the dual delivery line 220. Heat also flows across the two sections of each delivery line until heat also flows in the gap between the dual delivery lines 220. Within the dual delivery line 220, heat also flows from the heated gas to the stainless steel conduit of the dual delivery line 220. The physics-based model, which will be discussed in more detail, can take into account these surface factors (including contact resistance) and the gases themselves in the fluid dynamics and heat transfer equations.

[0032] [No.] [2B] [Figure] is a graph illustrating an example of how the vapor pressure changes with the temperature of TEOS in the top delivery line, according to one implementation method. Therefore, the temperature itself affects the pressure within the delivery line, which will remain sufficiently constant and above the saturation pressure to prevent condensation.

[0033] Additional References [No.] [1A] [Figure], multiple sensors 140 are positioned along the gas delivery system 118, including an assembly along the delivery line 120 and, where appropriate, also inside the gas panel 124 and the processing chamber 115 (together referred to herein as the processing subsystem). [No.] [1C] [Figure] illustrates some potential locations of these sensors 140 as a star shape overlapping at various locations along the assembly of the delivery line 120. In one implementation, the sensors 140 are a collection of temperature sensors synchronized and coupled to a computing device 101, which includes a processing device operatively coupled to a processing unit (see Figure 101). [No.] [7] [Figure]) memory. Sensor 140 may be coupled to and / or wirelessly coupled to computing device 101 via a wired connection. In an alternative embodiment, computing device 101 may include an edge device located at the end of sensor 140.

[0034] In various implementations, these sensors 140 may include one or more of the following: thermocouples, pressure sensors, concentration sensors, optical sensors, gas flow sensors, or heater jacket power output sensors. Thus, some sensors 140 are attached to the delivery line 120, while others are located or attached inside the delivery line, within the gas panel 124 and / or within the processing chamber 115. Temperature sensors can detect temperature, such as a reference... [No.] [2B] [Figure] The discussion relates to pressure. Alternatively, pressure can be measured directly by a pressure sensor. The pressure in a delivery line can be associated with a specific gas (e.g., precursor) mixture and determine whether it has dropped below its saturation pressure and thus condensed.

[0035] In related or different implementations, a concentration sensor can sense the gas concentration in the delivery line, which can be another way to detect condensation. For example, the concentration sensor can be a piezoelectric cone. Alternatively, an optical sensor, such as a tomographic sensor or a nondispersive infrared (NDIR) sensor, can be used, where the gas concentration is proportional to the absorption at a specific wavelength and can identify specific compounds in the scanned gas. The optical sensor can thus detect a phase change in the liquid that indicates condensation. A heater jacket power output sensor can be coupled to the heater jacket and detect anomalies or other deviations from the desired power level, which can indicate a defective heater jacket.

[0036] In various embodiments, [No.] [1A] The computing device 101 in [Figure] can execute instructions stored in memory to instantiate various modules or components of the computing device 101, including a sensor control module (SCM) 150, a sensor statistics module (SSM) 152, a machine learning module (MLM) 154, and a fault detection module (FDM) 170. In some embodiments, computer-generated models including machine learning models can be stored in memory and referenced during parameter sensing and computer modeling for gas delivery health monitoring.

[0037] The SCM 150 can activate sensors, deactivate sensors, put sensors into idle state, change sensor settings, detect sensor hardware or software problems, and so on. In some implementations, the SCM 150 can keep track of the processing operations performed by the system 100 and determine which sensors 140 will be sampled for specific processing (or diagnostic, maintenance, etc.) operations of the system 100.

[0038] SSM 152 can process the raw data obtained from sensor 140 via SCM 150 and determine the statistics representing the raw data. For example, for each or some of the raw sensor data distributions, SCM 150 can determine one or more parameters of the distribution, such as the mean, median, mode, upper limit, lower limit, variance (or standard deviation), skewness (third moment), kurtosis (fourth moment), or any other moment or cumulative quantity of the data distribution. In some implementations, SCM 150 can model the raw data using various model distributions (normal distribution, log-normal distribution, binomial distribution, Passon distribution, gamma distribution, or any other distribution) (e.g., fitted via regression analysis). In such implementations, one or more parameters may include the identification of the fitted distribution used with the fitted parameters determined by SCM 150.

[0039] In some implementations, the SCM 150 can use multiple distributions to fit the raw data from a single sensor, such as the principal and tail distributions for outlier data points. The parameters of the distributions obtained by the SCM 150 can be sensor-specific. For example, for some sensors, a small number of parameters (mean, median, variance) can be determined, while for others, more (e.g., 10 or 20) torques can be determined.

[0040] In various implementations, the computer-generated model 160 includes at least one of the following: fluid dynamics equations, heat transfer equations, and / or thermal contact resistance equations associated with the delivery pipeline or processing subsystem. The heat transfer equations may be of at least three types, including but not limited to heat conduction, heat convection, and heat radiation. The thermal contact resistance equations can model less-than-perfect heat transfer (e.g., conduction) between two surfaces in contact, such as in the case of various sensors. Other equations based on thermodynamics or physics are envisioned. The computer-generated model may be an aggregation or mixture of at least some of these equations.

[0041] In some implementations, computing device 101 can calibrate computer-generated model 160, such as by referencing... [No.] [Figure 4] More detailed discussion. In other implementations, the MLM 154 of computing device 101 can use a training dataset as input to train a machine learning model. This training dataset is measured by multiple sensors 140 and associated with one or more processing parameters to provide a computer-generated model 160 to replace rigorous algorithmic fusion. The implementation of the machine learning model will be referred to... [No.] [5 Figures] More detailed discussion. In some implementations, the MLM 154 is located on a remote server or can be distributed across multiple processing devices and operates on the training dataset to train the machine learning model, and sends the machine learning model back to the computing device 101 for storage and use.

[0042] In various implementations, FDM 170 can compare predicted parameter values ​​with measurement data received from multiple sensors 140 positioned (e.g., attached) along delivery line 120. If there is a variation between the predicted value and the measured value for a parameter that at least meets a threshold deviation, FDM 170 detects a fault. In some implementations, FDM 170 can combine predicted and measured values ​​for more than one parameter, for example, statistically or otherwise, to determine a fault based on more than one parameter, such as temperature, pressure, gas concentration, gas flow rate, and / or heater sheath power output. FDM 170 can further determine (or identify) a deviation pattern of this threshold deviation between the measured and predicted values ​​for at least the parameter. FDM 170 can compare the deviation pattern with predetermined deviation patterns previously associated with various faults to determine the specific fault. Computing device 101 can further warn the operator of the specific fault and, where appropriate, suspend processing when the specific fault is resolved and the operator clears the process to continue.

[0043] In some implementations, additional or more subtle patterns of change can be associated with specific faults and stored in tables or other data structures, such as the memory of computing device 101, during calibration runs and while tracking parameter values ​​that tend to change over time and are associated with identified faults. FDM 170 can access this table or data structure during automated diagnostics or post-continuous preventative maintenance analysis to determine whether a fault has been detected. Specific faults that FDM 170 can determine include, but are not limited to: ambient air leaking through heater sheath 128 into direct contact with delivery line 120; lack of sensitivity to changes in gas flow rate; a gas partial pressure predicted by computer-generated model 160 being lower than the liquid saturation pressure; input power to one or more of the plurality of heater sheaths being outside a predetermined range for maintaining a target temperature within the delivery line; a gas temperature in the chamber manifold being lower than a historical temperature value; or unexpected fluctuations in sensor signals from one of the plurality of sensors 140.

[0044] The above approach can be further applied to larger collections of physical processing subsystems, such as those including delivery lines 120 (or multiple delivery lines), heater sheaths 128 positioned (e.g., wound) around delivery lines 120, gas panels 124, and processing chambers 115. In this way, the computer-generated model 160 can be extended for preventative maintenance, fault detection, or automated diagnostics of the gaseous state of liquids both outside and inside the processing chamber 115.

[0045] In some implementations, the computer-generated model 160 integrates statistical analysis of measurement data received from the sensor 140. Furthermore, the computing device 101 can further utilize measurement data from sensors located along the delivery line (or processing subsystem) to perform reinforcement learning on the computer-generated model 160 (or machine learning model). Reinforcement learning and / or statistical analysis enable updating the computer-generated model and tracking trends in parameters across segments, bends, and regions of the delivery line (in addition to other components of the processing subsystem, such as gas panel 124 and processing chamber 115). Additionally, comparing measurement data with values ​​of one or more processing parameters can include comparing measurement data from one segment of multiple segments of the delivery line with specific values ​​of one or more processing parameters corresponding to that segment. Therefore, the diagnostic analysis discussed herein can be performed at the granularity of different segments, regions, or bends of the delivery line 120, or as a collection of such components.

[0046] Furthermore, by means of reinforcement learning and other updates performed on the computer-generated model 160, the computing device 101 can take one or more of the multiple sensors 140 offline. Alternatively, for example, during preventative maintenance, when an operator accesses the sensors 140, one or more of the multiple sensors 140 can be removed. This is because the computer-generated model 160 will be adequately calibrated or trained to maintain accuracy with only sparse and / or infrequent measurement data. The need for fewer sensors 140 saves resources and simplifies the system 100 in the long run.

[0047] [3rd] [Figure] is a flowchart of a method 300 for monitoring the health of a gaseous liquid being delivered to a processing chamber according to some implementation methods. Method 300 can be executed by processing logic, which may include hardware (e.g., processing device, circuit system, dedicated logic, programmable logic, microcode, device hardware, integrated circuitry, etc.), software (e.g., instructions that run or execute on the processing device), or a combination thereof. In some embodiments, method 300 is executed by... [No.] [1A] [Image] to [No.] [2A] The system 100 and components shown in the figure, or any combination thereof, are executed. Method 300 can be executed using a single processing device or multiple processing devices. Some operations of method 300 may be optional, as indicated by the dashed lines. In an implementation, some operations of method 300 are executed by the processing device (processor, central processing unit (CPU)) of computing device 101, for example, in response to instructions issued by fault detection module (FDM) 170.

[0048] In various implementations of method 300, at operation 310, the processing logic executes a sequence of plans for processing the substrate in processing chamber 115 and other processing chambers, the sequence including planned idle times. At operation 320, the processing logic proceeds to the next plan or item in the sequence of plans. In one implementation, the next plan to be processed is a preventative maintenance recovery check.

[0049] In operation 325, the processing logic receives measurement data from a plurality of sensors 140 positioned along a delivery line that delivers liquid as gas to either a gas panel or a processing chamber. In another implementation, operation 325 is extended to include receiving measurement data from a plurality of sensors 140 positioned along a processing subsystem that delivers liquid as gas to a processing chamber.

[0050] In operation 330, the processing logic simulates one or more processing parameters associated with the delivery line 120 and the plurality of heater sheaths 128 positioned (e.g., wound) around the delivery line by executing a computer-generated model. In another implementation, operation 330 is extended to include simulating one or more processing parameters across components of the processing subsystem (e.g., delivery line 120, gas panel 124, plurality of heater sheaths 128, and processing chamber 115). In some implementations, operations 325 and 330 are executed in parallel. In these and alternative implementations, operations 325 and 330 are executed periodically or continuously during processing to perform real-time diagnostics on at least the delivery line 120 and, where appropriate, the processing subsystem.

[0051] In operation 340, the processing logic compares measurement data (from multiple sensors 140) with the values ​​of one or more processing parameters. This comparison may be, for example, a comparison of a first signal received from multiple sensors 140 with a second signal generated by a computing device 101 executing a computer-generated model 160.

[0052] In operation 350, the processing logic determines whether the difference between the measured data and the values ​​of one or more processing parameters meets at least a threshold deviation. If so, operation 350 proceeds, and subsequently in operation 352, this threshold deviation can trigger the processing logic to determine that a fault has been detected and trigger a corrective action. For example, in operation 355, the processing logic can suspend processing in response to the fault determination (e.g., within processing chamber 115). Other corrective actions may also be taken, such as compensating for temperature deviations in the heater housing by increasing the power supplied to the heater housing, and the like. If at least a threshold deviation is not detected, the processing logic can loop back to operation 310, where the next processing step in the sequence of actions is executed without interruption.

[0053] Additional References [No.] [3] [Figure], in operation 360, the processing logic determines a deviation pattern of at least a threshold deviation between the measurement data and the values ​​of one or more processing parameters. In operation 370, the processing logic determines whether the deviation pattern corresponds to any one or more predetermined faults, with reference to... [No.] [1A] [Figure] This can be accessed in tables or other data structures in the memory of computing device 101. If operation 370 produces a negative response, then in operation 380, the processing logic warns the operator of a known fault, such as a specific fault that has been determined. If operation 370 produces a positive response, then in operation 390, the processing logic warns the operator of an unknown fault. Such an unknown fault may be used to perform preventative maintenance or additional diagnostic steps to attempt to verify the sufficient cause of the unknown fault. The processing logic may send or issue such warnings to the operator via input / output (I / O) devices of the processing unit coupled to computing device 101.

[0054] In either case, at operation 395, the processing logic may wait for operator input (e.g., via an I / O device) indicating that processing should proceed before continuing. This may force the processing logic into a safety cycle (similar to the emergency stop performed at operation 355) until the operator manually inserts a command to continue after verifying that the fault has been resolved.

[0055] [4th] [Figure] is a flowchart of a method 400 for generating and calibrating a computer-generated model according to some implementations. Method 400 can be executed by processing logic, which may include hardware (e.g., processing device, circuit system, dedicated logic, programmable logic, microcode, device hardware, integrated circuitry, etc.), software (e.g., instructions that run or execute on the processing device), or a combination thereof. In some embodiments, method 400 is executed by... [No.] [1A] [Image] to [No.] [2A] The system 100 and components shown in the figure, or any combination thereof, are executed. Method 400 can be executed using a single processing device or multiple processing devices. Some operations of method 400 are optional, as indicated by the dashed lines. In some implementations, some operations of method 400 are executed by the processing device (processor, central processing unit (CPU)) of computing device 101.

[0056] In operation 410, the processing logic incorporates algorithms for fluid dynamics, heat transfer, and / or, depending on, thermal contact resistance associated with the gas delivery line (or processing subsystem) to generate a computer-generated model. In operation 420, the processing logic receives a set of baseline measurement data from multiple sensors 140. The baseline measurement data is available when system 100 is first brought online or after preventative maintenance and reset of system 100.

[0057] In operation 430, the processing logic inputs a dataset into a computer-generated model, for example, for calibration purposes. In operation 440, the processing logic executes the computer-generated model to predict one or more faults based on the input dataset. In this manner, the processing logic calibrates the computer-generated model 160 using baseline measurement data, which is compared to (or inferred from) an updated dataset of measurement data. In operation 450, the processing logic, for calibration purposes, compares the faults predicted by the computer-generated model with known faults that have occurred.

[0058] In operation 460, the processing logic determines whether the prediction determined in operation 450 is correct based on a comparison. For example, it may determine that the prediction ensures the computer-generated model 160 predicts a fault within an accuracy threshold, and therefore is a correct prediction. If the prediction determined in operation 450 is correct, then in operation 470, the processing logic outputs a calibrated model as the computer-generated model 160 for fault diagnosis of the delivery pipeline 120 and / or the processing subsystem. Although temporarily completed in operation 470, method 400 can use historical measurement data received in the interim to repeat at a later date, thereby producing an updated calibrated model based on measurement data received during a period of time. If the prediction determined in operation 450 is incorrect, then in operation 480, the processing logic updates the computer-generated model based on the prediction failure and loops back to operation 410 to continue calibrating the computer-generated model.

[0059] [5th] [Figure] is a flowchart of a method 500 for training a machine learning model to predict faults in a liquid conversion and gas delivery system, according to some implementations. Method 500 can be executed by processing logic, which may include hardware (e.g., processing device, circuit system, dedicated logic, programmable logic, microcode, device hardware, integrated circuits, etc.), software (e.g., instructions that run or execute on the processing device), or a combination thereof. In some embodiments, method 500 is executed by... [No.] [1A] [Image] to [No.] [2A] The system 100 and components shown in the figure, or any combination thereof, are executed. Method 500 can be executed using a single processing device or multiple processing devices. Some operations of method 500 may be optional, as indicated by the dashed lines. In an implementation, some operations of method 500 are executed by the processing device (processor, central processing unit (CPU)) of computing device 101, for example, in response to instructions issued by machine learning module (MLM) 154.

[0060] In operation 510, the processing logic initializes (or updates, if in a later iteration of method 500) the untrained machine learning model. In operation 520, the processing logic receives a training dataset of measurement data from multiple sensors 140. Depending on the focus of the machine learning model, this measurement data can come from different types of sensors or multiple types of sensors. In one implementation, the focus of the machine learning model is to predict a single parameter, such as temperature, and this can be performed multiple times for different parameters, such as the pressure of the heater sheath 128, gas concentration, or power output. In another implementation, the focus of the machine learning model is to predict a combination of parameters, which can be related to, for example, temperature and pressure, or the concentration and composition of molecules detected from optical sensors.

[0061] In operation 530, the processing logic inputs the training dataset into an untrained machine learning module. In operation 540, the processing logic trains the untrained machine learning module to produce a machine learning model that is at least partially trained. In operation 550, the processing logic compares (at least partially) the faults predicted by the trained machine learning model with the generated known faults for the purpose of supervised learning.

[0062] Continue to refer to [No.] [5] [Figure], in operation 560, the processing logic determines whether the prediction determined in operation 550 matches a known fault that the processing logic compares with. If the prediction is incorrect, the processing logic loops back to operation 510 to continue (at least partially) training the still untrained machine learning model. However, if the prediction is correct, in operation 570, the processing logic outputs the trained machine learning model as a computer-generated model 160 for fault diagnosis of gas delivery pipeline 120 and / or the previously discussed processing subsystem.

[0063] [No.] [6] [Figure] illustrates a set of graphs showing sensor measurement data compared to a computer-generated model of one or more processing parameters, and from which patterns of change can be derived, according to one implementation. For example, a set of sensor measurement data 605 (temperature for illustrative purposes) is received from at least one sensor, identified as Ch36(C) at the fourth bend of delivery line 120. This sensor measurement data 605 can be plotted against gas flow (e.g., the flow rate of a gaseous precursor liquid). The sensor measurement data 605 can then be plotted on a single graph 607, shown as a lower line. The computer-generated model 160 (or machine learning model) can output its own predictions to the single graph 607, as an upper line in this case. Note that although there is a correspondence between the upper and lower lines, there are some deviations that can be considered significant enough to be considered faults, particularly as seen over time at the beginning and inflection points. The computing device 101 can generate a deviation pattern by subtracting the lower line from the upper line. This deviation pattern can then be used to determine whether the deviation pattern matches one of a number of possible faults when compared with a pre-stored deviation pattern.

[0064] [No.] [7] [Figure] A block diagram depicting an example computing device 700 for parameter sensing and computer modeling for gas transport health monitoring, operating according to one or more state-specific configurations of this disclosure, and more particularly capable of performing various implementations. In one implementation, the computing device 700 may be... [No.] [1] [Figure] The computing device 101 or the microcontroller of the computing device 101.

[0065] The example computing device 700 can be connected to other processing devices in a LAN, intranet, extranet, and / or internet. The computing device 700 can be a personal computer (PC), set-top box (STB), server, network router, switch or bridge, or any machine capable of executing a set of instructions (continuously or otherwise) that specifies the actions to be taken by that device. Furthermore, although only a single example processing device is shown, the term "processing device" should also be considered to include any set of processing devices (e.g., computers) that independently or jointly execute a set of instructions (or multiple sets of instructions) to perform any one or more methods discussed herein.

[0066] Example computing device 700 may include processing device 702 (e.g., CPU), main memory 704 (e.g., read-only memory (ROM), flash memory, dynamic random access memory (DRAM) such as synchronous DRAM (SDRAM), etc.), static memory 706 (e.g., flash memory, static random access memory (SRAM), etc.), and auxiliary memory (e.g., data storage device 718), which communicate with each other via bus 730.

[0067] Processing device 702 represents one or more general-purpose processing devices, such as microprocessors, central processing units, or the like. More specifically, processing device 702 may be a complex instruction set computing (CISC) microprocessor, a reduced instruction set computing (RISC) microprocessor, a very long instruction word (VLIW) microprocessor, a processor implementing other instruction sets, or a processor implementing a combination of instruction sets. Processing device 702 may also be one or more special-purpose processing devices, such as application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), digital signal processors (DSPs), network processors, or the like. According to one or more embodiments of this disclosure, processing device 702 may be configured to execute instructions for implementing methods 300 and / or 400 and 500 for monitoring the transport health of gaseous liquids.

[0068] The example computing device 700 may further include a network interface device 708 communicatively coupled to a network 720. The example computing device 700 may further include a video display 710 (e.g., a liquid crystal display (LCD), a touchscreen, or a cathode ray tube (CRT)), an alphanumeric input device 712 (e.g., a keyboard), an input control device 714 (e.g., a cursor control device, a touchscreen control device, a mouse), and a signal generation device 716 (e.g., an acoustic speaker).

[0069] The computing device may include a data storage device 718, including a computer-readable storage medium (or more specifically, a non-transitory computer-readable storage medium) 728 on which one or more sets of executable instructions 722 are stored. According to one or more embodiments of this disclosure, the executable instructions 722 may include executable instructions for implementing methods 300 and / or 400 and 500 for monitoring the transport health of gaseous liquids.

[0070] The executable instructions 722 may also be, during their execution, wholly or at least partially housed within the main memory 704 and / or the processing device 702, which also constitute computer-readable storage media. The executable instructions 722 may further be sent or received over a network via a network interface device 708.

[0071] Despite [No.] [7] [Figure] illustrates computer-readable storage medium 728 as a single medium. The term "computer-readable storage medium" (or "non-transitory computer-readable medium storing instructions") should be considered to include a single medium or multiple media (e.g., centralized or distributed data sets, and / or associated cache and servers) that store one or more sets of operating instructions. The term "computer-readable storage medium" (or "non-transitory computer-readable medium storing instructions") should also be considered to include any medium capable of storing or encoding a set of instructions for execution by a machine, which causes the machine to perform any one or more of the methods described herein. The term "computer-readable storage medium" (or "non-transitory computer-readable medium") should therefore be considered to include, but is not limited to, solid-state memory, and optical and magnetic media.

[0072] It should be understood that the above description is intended to be illustrative rather than restrictive. Many other implementation examples will become apparent to those skilled in the art after reading and understanding the above description. Although this disclosure describes specific examples, it will be appreciated that the systems and methods disclosed herein are not limited to the examples described herein, but can be practiced with modifications within the scope of the appended claims. Therefore, the specification and drawings are considered illustrative rather than restrictive. Consequently, the scope of this disclosure should be determined by referring to the entire scope of the appended claims together with their equivalents.

[0073] The methods, hardware, software, firmware, or code described above can be implemented via instructions or code stored on a machine-accessible, machine-readable, computer-accessible, or computer-readable medium executable by a processing element. "Memory" includes any mechanism that provides (i.e., stores and / or transmits) information in a form readable by a machine (such as a computer or electronic system). For example, "memory" includes random-access memory (RAM), such as static RAM (SRAM) or dynamic RAM (DRAM); ROM; magnetic or optical storage media; flash memory devices; electrical storage devices; optical storage devices; acoustic storage devices; and any type of tangible machine-readable medium suitable for storing or transmitting electronic instructions or information in a form readable by a machine (e.g., a computer).

[0074] Throughout this specification, the reference to "an implementation" or "a particular implementation" means that a specific feature, structure, or characteristic described in connection with that implementation is included in at least one implementation disclosed herein. Therefore, the phrases "in an implementation" or "in a particular implementation" appearing in various places throughout this specification do not necessarily refer to the same implementation. Furthermore, a particular feature, structure, or characteristic may be combined in one or more implementations in any suitable manner.

[0075] In the foregoing specification, detailed descriptions have been given with reference to specific exemplary implementations. However, it will be demonstrated that various modifications and changes can be made thereto without departing from the broader spirit and scope of the disclosure as set forth in the appended claims. The specification and drawings are therefore to be regarded as illustrative rather than restrictive. Furthermore, the foregoing use of the terms "implementation," "implementation," and / or other exemplary language does not necessarily refer to the same implementation or the same instance, but may refer to different and dissimilar implementations, as well as potentially similar implementations.

[0076] The terms “example” or “exemplary” are used herein to mean as an instance, example, or illustration. Any form or design described herein as an “example” or “exemplary” is not necessarily to be construed as preferred or advantageous over other forms or designs. Rather, the use of the terms “example” or “exemplary” is intended to provide a concept in a specific manner. As used in this application, the term “or” is intended to mean an inclusive “or” rather than an exclusive “or.” That is, unless otherwise stated or clear from the context, “X includes A or B” is intended to mean any naturally included substitution. That is, “X includes A or B” is satisfied in any of the foregoing cases if X includes A; X includes B; or X includes both A and B. Furthermore, the article “a / an” as used in this application and the appended claims should generally be construed as meaning “one or more” unless otherwise stated or clear from the context that it pertains to the singular form. Furthermore, the use of the terms "an implementation" or "a method of implementation" or "an implementation" throughout the text is not intended to imply the same implementation or method of implementation, unless otherwise described. Additionally, as used herein, the terms "first," "second," "third," "fourth," etc., are markers for distinguishing different elements and do not necessarily have ordinal meaning based on their numerical designation.

[0077] 100: Liquid Conversion and Gas Transport Systems 101: Computing device 102: Gas delivery pipeline 104: Propel the gas pipeline 110: Evaporator 114: Liquid storage containers 115: Processing Chamber 118: Gas delivery system 120: Delivery pipeline 120A: Facility gas pipeline 120B: Chamber gas line 124: Gas Panel 128: Heater sheath 128A: Heater jacket 128B: Heater jacket 128C: Heater jacket 128D: Heater jacket 132: Heating element 136: Thermocouple 138: Chamber gas feedthrough 140: Sensor 150: Sensor Control Module (SCM) 152: Sensor Statistics Module (SSM) 154: Machine Learning Module (MLM) 160: Computer-generated model 170: Fault Detection Module (FDM) 220: Dual delivery pipeline 228: Heater sheath 300: Method 310: Operation 320: Operation 325: Operation 330: Operation 340: Operation 350: Operation 352: Operation 355: Operation 360: Operation 370: Operation 380: Operation 390: Operation 395: Operation 400: Method 410: Operation 420: Operation 430: Operation 440: Operation 450: Operation 460: Operation 470: Operation 480: Operation 500: Methods 510: Operation 520: Operation 530: Operation 540: Operation 550: Operation 560: Operation 570: Operation 605: Sensor Measurement Data 607: Chart 660: Computer-generated (or trained machine learning) model 700: Computing device 702: Processing device 704: Main Memory 706: Static Memory 708: Network Interface Device 710: Video Monitor 712: Alphanumeric Input Device 714: Input control device 716: Signal generating device 718: Data storage device 720: Internet 722: Executable instructions 726: Processing Logic 728: Computer-readable storage media 730: Busbar LFM: Liquid Flow Meter MFC: Mass Flow Controller

[0078] Domestic storage information (please note in order of storage institution, date, and number) none Overseas storage information (please note in the order of storage country, institution, date, and number) none

Claims

1. A method for parameter sensing and computer modeling for gas delivery health monitoring, comprising the steps of: receiving measurement data from a plurality of sensors positioned along a delivery pipeline that delivers a liquid as a gas to either a gas panel or a processing chamber; using a computer-generated model executed on a processing device to simulate one or more processing parameters associated with the delivery pipeline and a plurality of heater sheaths positioned around the delivery pipeline; comparing the measurement data with values ​​of the one or more processing parameters by the processing device; and determining, by the processing device, the presence of a fault associated with maintaining the temperature within the delivery pipeline consistent with a gaseous state of the liquid based on at least one threshold deviation between the measurement data and the equivalent values ​​of the one or more processing parameters.

2. The method as described in Request 1 further includes the following steps: in response to the determination that the fault exists, triggering a correction action.

3. The method as described in claim 1, in response to the step of determining the fault, further comprises the following steps: suspending processing within the processing chamber; alerting an operator to the fault via an input / output device coupled to the processing device; and waiting for input from the operator via the input / output device, the input indicating that processing should proceed before resuming.

4. The method as claimed in claim 1, wherein the one or more processing parameters are at least one of the following: the temperature, pressure, concentration of the gas, or the power output of the plurality of heater sheaths, wherein the method further comprises the step of: training a machine learning model to provide the computer-generated model using a training dataset measured by the plurality of sensors and associated with the one or more processing parameters as input.

5. The method as described in claim 1, wherein the computer-generated model includes at least one of the following: fluid dynamics equations, heat transfer equations, and / or thermal contact resistance equations associated with the delivery pipeline.

6. The method as described in claim 1, wherein the delivery pipeline comprises a plurality of sections, including bends and areas, and wherein the comparison comprises the step of comparing the measurement data of one section of the plurality of sections with a specific value of one or more processing parameters corresponding to that section.

7. The method as described in claim 1, wherein the method further comprises the steps of: performing reinforcement learning based on the equivalent values ​​of the one or more processing parameters to update the computer-generated model, while processing within the processing chamber; and removing or taking offline one or more of the plurality of sensors.

8. The method as described in claim 1 further comprises the steps of: determining a deviation pattern of the at least one threshold deviation between the measurement data and the equivalent values ​​of the one or more processing parameters; and determining a specific fault corresponding to the deviation pattern of the one or more processing parameters.

9. The method as described in claim 8, wherein the specific fault includes at least one of the following: leakage of ambient air through the plurality of heater sheaths into direct contact with the delivery line; lack of sensitivity to changes in the flow rate of the gas; the partial pressure of the gas predicted by the computer-generated model being lower than a saturation pressure of the liquid; an input power of one or more of the plurality of heater sheaths being outside a predetermined range for maintaining a target temperature in the delivery line; or an unexpected fluctuation in a sensor signal from one of the plurality of sensors.

10. A system for parameter sensing and computer modeling for gas delivery health monitoring, comprising: a memory; and a processing device operatively coupled to the memory for: receiving measurement data from a plurality of sensors positioned along a processing subsystem that delivers a liquid as a gas into a processing chamber; executing a computer-generated model simulating one or more processing parameters associated with the processing subsystem; comparing the measurement data with values ​​of the one or more processing parameters; and determining, based on at least one threshold deviation between the measurement data and the equivalent values ​​of the one or more processing parameters, the existence of a fault associated with maintaining temperature and pressure within the processing subsystem consistent with a gaseous state of the liquid.

11. The system as claimed in claim 10, wherein the processing device further uses baseline measurement data from the plurality of sensors to calibrate the computer-generated model to ensure that the computer-generated model predicts the fault within a threshold level of accuracy.

12. The system as claimed in claim 10, wherein the computer-generated model includes at least one of the following: fluid dynamics equations, heat transfer equations, and / or thermal contact resistance equations associated with the processing subsystem and a statistical analysis of measurement data received from the plurality of sensors.

13. The system of claim 10, further comprising the processing subsystem comprising: a collection of delivery lines for carrying the liquid as the gas; a gas panel coupled to the collection of delivery lines for feeding the gas through the delivery lines into the processing chamber; a plurality of heater sheaths positioned around the delivery lines to maintain the temperature and pressure of the delivery lines, such as maintaining the liquid in the gaseous state; and the processing chamber.

14. The system of claim 13, wherein the plurality of sensors comprises one or more of the following: a thermocouple, a pressure sensor, an optical sensor, a concentration sensor, a gas flow sensor, or a heater sheath power output sensor, and the processing device further uses a training dataset measured by the plurality of sensors and associated with the one or more processing parameters as input to train a machine learning model to provide the computer-generated model.

15. The system as claimed in claim 13, wherein the delivery pipeline comprises a plurality of sections, including bends and zones, and wherein comparing the measurement data with the values ​​of the one or more processing parameters comprises comparing the measurement data of one section of the plurality of sections with a specific value of the one or more processing parameters corresponding to that section.

16. The system as claimed in claim 13, wherein the processing means is further configured to: determine a deviation pattern of the at least one threshold deviation between the measurement data and the equivalent values ​​of the one or more processing parameters; and determine a specific fault corresponding to the deviation pattern of the one or more processing parameters.

17. The system as claimed in claim 16, wherein the specific fault comprises at least one of the following: leakage of ambient air through the plurality of heater sheaths into direct contact with the delivery line; lack of sensitivity to changes in the flow rate of the gas; the partial pressure of the gas predicted by the computer-generated model being lower than a saturation pressure of the liquid; an input power to one or more of the plurality of heater sheaths being outside a predetermined range for maintaining a target temperature in the delivery line; the temperature of the gas in a chamber manifold being lower than a historical temperature value; and unexpected fluctuations in a sensor signal from one of the plurality of sensors.

18. A non-transitory computer-readable medium for parameter sensing and computer modeling for gas delivery health monitoring, the non-transitory computer-readable medium storing a plurality of instructions, which, when executed by a processing device, perform a plurality of operations, including the steps of: receiving measurement data from a plurality of sensors positioned along a delivery line that delivers a liquid as a gas to a processing chamber; simulating one or more processing parameters associated with the delivery line and a plurality of heater sheaths positioned around the delivery line by executing a computer-generated model; comparing the measurement data with the values ​​of the one or more processing parameters; and determining, based on at least one threshold deviation between the measurement data and the values ​​of the one or more processing parameters, the existence of a fault associated with maintaining the temperature and pressure within the delivery line consistent with a gaseous state of the liquid.

19. The non-transitory computer-readable medium as described in claim 18, wherein such operation further comprises the step of: calibrating the computer-generated model using baseline measurement data from the plurality of sensors to ensure that the computer-generated model predicts the fault within a threshold level of accuracy.

20. The non-transitory computer-readable medium as described in claim 18, wherein the computer-generated model includes at least one of the following: fluid dynamics equations, heat transfer equations, and / or thermal contact resistance equations associated with a statistical analysis of the delivery line and measurement data received from the plurality of sensors.

21. The non-transitory computer-readable medium as described in claim 18, wherein the one or more processing parameters are at least one of the following: the temperature, pressure, concentration of the gas, or the power output of the plurality of heater sheaths, wherein the plurality of operations further include the step of: training a machine learning model to provide the computer-generated model using a training dataset measured by the plurality of sensors and associated with the one or more processing parameters as input.

Images