System and method for sample injection monitoring and diagnosis of gas chromatography

Abstract

The present disclosure relates to systems and methods for sample monitoring and diagnostics for gas chromatography. A system for gas chromatography includes an inlet configured to receive a sample through a sample introduction, a column having a stationary phase, a flow control system, and a sample introduction monitoring system. The flow control system is configured to regulate flow of a mobile phase through the inlet and the column based on flow control parameters. The sample introduction monitoring system is configured to: obtain flow control data representing a measurement of the flow control parameter over time during a time period including sample introduction of the sample into the inlet; determining that the sample introduction is unsuccessful based on the flow control data; and performing a mitigation operation to mitigate unsuccessful introduction of the sample based on determining that the introduction is unsuccessful.

Description

[0001] Background Information

[0002] Gas chromatography (GC) is an analytical technique for separating and analyzing (e.g., detecting, identifying and / or quantifying) the chemical components of a sample mixture. GC is performed by injecting a sample into the inlet (also referred to as an injector) of a gas chromatograph by inserting a syringe needle through the diaphragm of the inlet. The injected sample evaporates in the inlet, and a mobile phase (commonly referred to as a carrier gas) flows through the inlet and carries the evaporated sample through a column (elongated tube) with a stationary phase. The mobile phase can be an inert gas or a non-reactive gas, such as helium, argon, nitrogen, hydrogen or argon / methane. Based on the various chemical and physical properties of the components, the components of the sample are retained in the column differently by the stationary phase and eluted from the column at different times. The eluted components are carried to a detector by the mobile phase, which can detect the components and generate a signal representing the detected components.

[0003] When the sample is injected into the inlet, various problems may occur. For example, when the sample is drawn from the vial, bubbles may be drawn into the syringe, causing the bubbles to be injected into the inlet. In some cases, the sample level in the vial may be lower than the tip of the syringe needle, so that the sample is not drawn into the syringe, resulting in no sample being injected into the inlet. In other cases, if the syringe needle pierces the septum of the inlet, the syringe needle may be blocked, thereby preventing the sample from being injected into the inlet. In some cases, the syringe needle and / or the syringe plunger may be bent, thereby preventing the complete injection of the sample.

[0004] However, when these problems occur, they may not be obvious to the user, or may not even be detected or diagnosed by the user. For example, once the sample has been injected and passed through the stationary phase, any problems with the injection may no longer be detectable, such as insufficient sample volume drawn by the syringe, injection of bubbles, and / or injection of less than the entire amount of the sample. The volumes of some injections are very small, perhaps about a fraction of a milliliter (mL), which may make the detection of incorrect injections virtually impossible for the user. If no signal for injection is detected, the user may not be able to diagnose the cause or determine whether the problem occurs at the inlet, at the detector, or at some other location (e.g., within the column).

[0005] [TP387104CNSEC1]

[0006] Conventional methods for monitoring and detecting incorrect injections include doping an internal standard into the sample and comparing a signal representing the internal standard with the expected signal of the internal standard. However, the use of an internal standard increases the cost, complexity, and time of performing a GC experiment. Although an internal standard can help detect incorrect injections, the use of an internal standard is not helpful in diagnosing the cause of an incorrect injection. Summary of the invention

[0007] The following description presents a simplified overview of one or more aspects of the methods and systems described herein to provide a basic understanding of such aspects. This summary is not an extensive overview of all covered aspects, and is neither intended to identify key or critical elements of all aspects, nor to delineate the scope of any or all aspects. Its sole purpose is to present some concepts of one or more aspects of the methods and systems described herein in a simplified form as a prelude to the more detailed description presented below.

[0008] In some illustrative examples, a system for gas chromatography includes: an inlet configured to receive a sample via injection; a column including a stationary phase; a flow control system configured to regulate the flow of a mobile phase through the inlet and the column based on a flow control parameter; and an injection monitoring system configured to perform a process including: obtaining flow control data representing a measured value of the flow control parameter over time during a time period including injecting the sample into the inlet; determining that the injection was unsuccessful based on the flow control data; and performing a mitigation operation based on determining that the injection was unsuccessful.

[0009] In some illustrative examples, an injection monitoring system for a gas chromatography system includes: one or more processors; and a memory storing executable instructions that, when executed by the one or more processors, cause a computing device to perform a process including: obtaining flow control data from a flow control system, the flow control system being included in the gas chromatography system and being configured to regulate a flow of a fluid through an inlet of the gas chromatography system based on a flow control parameter, wherein the flow control data represents a measured value of the flow control parameter over time during a time period including injecting a sample into the inlet; determining that the injection is unsuccessful based on the flow control data; and directing the gas chromatography system to perform a mitigation operation to mitigate the unsuccessful injection of the sample based on the determination that the injection is unsuccessful.

[0010] In some illustrative examples, a non-transitory computer-readable medium stores instructions that, when executed, direct at least one processor of a computing device for a gas chromatography system to perform a process comprising: obtaining flow control data from a flow control system configured to regulate a flow of a fluid through an inlet of the gas chromatography system based on a flow control parameter, wherein the flow control data represents a measurement of the flow control parameter over time during a time period including injecting a sample into the inlet.

[0011] a quantity; determining that the injection was unsuccessful based on the flow control data; and performing a mitigation action based on determining that the injection was unsuccessful. BRIEF DESCRIPTION OF THE DRAWINGS

[0012] The accompanying drawings illustrate various embodiments and are part of the specification. The embodiments shown are examples only and do not limit the scope of the present disclosure. Throughout the drawings, the same or similar reference numerals represent the same or similar elements.

[0013] Figure 1 A functional diagram of an exemplary gas chromatography (GC) system capable of split mode injection and splitless mode injection is shown.

[0014] Figure 2 The process includes injecting the sample into Figure 1 An illustrative graph of a pulse width modulated (PWM) valve drive signal over time during a time period in the inlet of a GC system.

[0015] Figure 3 A functional diagram of an exemplary injection monitoring system is shown.

[0016] Figure 4 An exemplary method of detecting unsuccessful injections is shown.

[0017] Figure 5 An exemplary method for performing a GC experiment with injection monitoring and diagnostics is shown.

[0018] Figure 6 A block diagram showing an exemplary training phase for training an injection classification model and a vapor volume estimation model is shown.

[0019] Figure 7 An illustrative method is shown that may be performed to train a machine learning model to classify an injection of a sample or estimate the vapor volume of an injected sample.

[0020] Figure 8 An illustrative method for training a machine learning model using training examples is shown.

[0021] Fig. 9 An illustrative computing device is shown that may be specifically configured to perform one or more of the processes described herein. DETAILED DESCRIPTION

[0022] This paper describes a system and method for monitoring the injection in the inlet of a gas chromatography (GC) system and diagnosing an unsuccessful injection. For example, an injection monitoring system can obtain flow control data from a flow control system, and the flow control system is configured to regulate the flow of the inlet of a fluid through a GC system based on a flow control parameter. The flow control data can represent the measured values ​​of the flow control parameter over time during a time period including injecting a sample into the inlet. Based on the flow control data, the injection monitoring system can determine that the injection is unsuccessful. Based on determining that the injection is unsuccessful, the injection monitoring system can guide the GC system to perform a mitigation operation to mitigate the unsuccessful injection of the sample.

[0023] [TP387104CNSEC1]

[0024] In some examples, the flow control parameter is a pulse width modulation (PWM) valve drive signal for a valve of the flow control system, a pressure signal output by a pressure sensor of the flow control system, or a flow rate signal output by a flow rate sensor of the flow control system.

[0025] The systems and methods described herein improve GC systems and GC methods by detecting these unsuccessful injections when they occur, even when the user may not otherwise be able to discern any problems in the injection. The systems and methods described herein also improve GC systems and GC methods by alleviating unsuccessful injections, such as by discarding data obtained based on other unsuccessful injections, performing diagnostic processes to identify the cause of unsuccessful injections and / or providing alarms. In some examples, the systems and methods described herein quantify the sample volume injected by each injection (including partial injections), and this information can be used to appropriately scale (if necessary) the GC data obtained. The systems and methods described herein eliminate the need to use expensive, complex and time-consuming internal standards to monitor injection quality. The systems and methods described herein can be implemented under a wide range of instrument and experimental conditions, with little or no additional work for the user. The systems and methods described herein can also be implemented on traditional GC systems without the need to install new hardware.

[0026] Various embodiments will now be described in more detail with reference to the accompanying drawings.The systems and methods described herein may provide one or more of the benefits described above and / or various additional and / or alternative benefits will be apparent herein.

[0027] Exemplary systems and methods for monitoring injections, diagnosing unsuccessful injections, and mitigating unsuccessful injections will now be described with reference to an exemplary gas chromatography (GC) system. The GC system described is exemplary and non-limiting.

[0028] Figure 1 A functional diagram of an exemplary GC system 100 capable of performing split mode injection and splitless mode injection is shown. The GC system 100 includes an inlet 102, a column 104, an input path 106, a column path 108, a detector 109, a split path 110, a purge path 112, a flow control system 115, and a GC controller 117. The GC system 100 may include additional or alternative components useful for a particular implementation, such as a charcoal trap (not shown) for trapping contaminants, an oven, and / or an autosampler.

[0029] The inlet 102 (also referred to as the injector) includes a septum 114 that covers and seals the inlet 102. The septum 114 can be formed of a self-sealing material, such as silicone. Alternatively, the septum 114 can be a mechanical spring-assisted device that opens and closes when a needle is inserted. When a syringe needle (not shown) pierces or opens the septum 114, the inlet 102 receives the sample by injection, and the sample is injected from the syringe into the inlet 102. The sample may include one or more analytes of interest dissolved in a solvent. Exemplary solvents include, but are not limited to, methanol, acetone, pentane, hexane, isooctane, etc. The inlet 102 is connected via an input [TP387104CNSEC1]

[0030] Path 106 receives mobile phase (e.g., carrier gas). Sample is mixed with mobile phase in inlet 102, and a portion of the fluid mixture leaves inlet 102 and passes through column 104 via column path 108. Column 104 includes a stationary phase, which can be solid or liquid. Column 104 separates these components based on the interaction of the components in the sample injected with the stationary phase. Detector 109 is coupled to the output end of column 104 and detects the components of the sample when the components are eluted from column 104. In some examples, detector 109 is a gas chromatograph detector, such as a flame ionization detector or a thermal conductivity detector. In other examples, detector 109 is a mass spectrometer. The data generated by detector 109 can be output to GC controller 117.

[0031] A small portion of the fluid exits the inlet 102 via the purge path 112. The purge path 112 provides a flow path to expel a portion of the fluid out of the inlet 102, thereby clearing contaminants that may be introduced into the inlet 102 by the septum 114 when the septum 114 is pierced by the syringe needle. The purge path 112 expels the fluid before any contaminants from the septum 114 mix with the injected sample.

[0032] The inlet 102 can be a split / splitless (SSL) inlet or a programmed temperature vaporization (PTV) inlet. The PTV inlet can also be operated in a split mode and / or a splitless mode. In the split mode, a portion of the fluid leaves the inlet 102 via the split path 110. The split path 110 provides a flow path to discharge the fluid out of the inlet 102. The ratio of the flow rate of the fluid leaving the inlet 102 via the split path 110 to the flow rate of the fluid leaving the inlet 102 via the column 104 is referred to as the "split ratio". Any suitable split ratio can be used, such as but not limited to 10:1, 20:1, 50:1, 100:1. In the splitless mode, the fluid does not leave the inlet 102 via the split path 110. In the split mode and the splitless mode, the flow rate of the carrier gas flowing into the inlet 102 is equal to the sum of the flow rates of the fluid leaving the inlet 102 (the volume of the sample injected is considered to be negligible).

[0033] The GC controller 117 is communicatively coupled to the GC system 100 and is configured to control the operation of the GC system. The GC controller 117 may include any suitable hardware (e.g., processor, circuit, etc.) and / or software configured to control the operation of and / or connect to various components of the GC system 100 (e.g., detector 109, flow control system 115, oven, autosampler, etc.). The GC controller 117 receives data output by the detector 109 and may process the data (e.g., generate a chromatogram, generate a mass spectrum, analyze the data, transmit the data to another computing system, etc.) and / or store the data (e.g., in a memory). The GC controller 117 may also include and / or provide a user interface that is configured to enable interaction between a user and the GC controller 117. The user may interact with the GC controller 117 via the user interface through tactile, visual, auditory, and / or other sensory types of communication. For example, the user interface may include a display device (e.g., a liquid crystal display (LCD) display screen, a touch screen, etc.) for displaying information (e.g., a chromatogram, a mass spectrum, a notification, etc.) to the user. [TP387104CNSEC1]

[0034] The user interface may also include an input device (e.g., a keyboard, a mouse, a touch screen device, etc.) that allows a user to provide input to the GC controller 117. In other examples, the display device and / or the input device may be separate from the GC controller 117 but communicatively coupled to the controller. For example, the display device and the input device may be included in a computer (e.g., a desktop computer, a laptop computer, etc.) that is communicatively connected to the GC controller 117 via a wired connection (e.g., via one or more cables) and / or a wireless connection. Although Figure 1 The GC controller 117 is shown to be included in the GC system 100, but the GC controller 117 may alternatively be implemented in a manner that is completely or partially separate from the GC system 100, such as by a computing device that is communicatively coupled to the GC system 100 via a wired connection (e.g., a cable) and / or a network (e.g., a local area network, a wireless network (e.g., Wi-Fi), a wide area network, the Internet, a cellular data network, etc.).

[0035] The flow control system 115 is configured to regulate the flow of fluid into and out of the inlet 102. The flow control system 115 includes a set of valves, a set of pressure sensors, and a flow controller. The flow control system 115 may include any additional or alternative components that may be suitable for a specific implementation. The valve 116 on the input path 106 regulates the flow of the carrier gas into the inlet 102, the valve 118 on the shunt path 110 regulates the flow of the fluid leaving the inlet 102 via the shunt path 110, and the valve 120 on the purge path 112 regulates the flow of the fluid leaving the inlet 102 via the purge path 112. The valves 116, 118, and 120 may include any suitable valves, such as proportional valves. The pressure sensor 122 downstream of the valve 116 measures the pressure head of the inlet 102 / column 104, the pressure sensor 124 upstream of the valve 118 measures the pressure at the valve 118, and the pressure sensor 126 upstream of the valve 120 measures the pressure at the valve 120. The pressure sensors 122, 124, and 126 can sample the inlet pressure at any suitable sampling rate. In some examples, the sampling rate ranges from 0.5 Hz to 500 Hz or 1 kHz. In further examples, the sampling rate ranges from 1 Hz to 100 Hz.

[0036] The flow controller 128 is configured to control (e.g., open and close) the valves 116, 118, and / or 120 to regulate the flow of fluid into and out of the inlet 102. For example, in a flow control mode, the flow controller 128 may regulate the flow of fluid through any one or more of the input path 106, the shunt path 110, and / or the purge path 112 to maintain a target flow rate through the column path 108 and the column 104 and / or to maintain a target split ratio. In a pressure control mode, the flow controller 128 may regulate the flow of fluid through any one or more of the input path 106, the shunt path 110, and / or the purge path 112 to maintain a target pressure within the inlet 102.

[0037] In some examples, the flow controller 128 uses feedback control to regulate the flow of fluid into and out of the inlet 102. To this end, the flow controller 128 is coupled to the valves 116, 118, and 120 and the pressure sensor [TP387104CNSEC1]

[0038] 122, 124 and 126 are communicatively coupled. The flow controller 128 receives the pressure signal output by any one or more of the pressure sensors 122, 124 and 126, and generates a valve drive signal for one or more of the valves 116, 118 and 120 based on the pressure signal. In some examples, the valve drive signal is a pulse width modulation (PWM) valve drive signal (e.g., voltage, current or digital value) that specifies the on / off duty cycle of the valve. For example, a PWM valve drive signal with a 50% duty cycle is half the on time and half the off time, while a PWM valve drive signal with a 30% duty cycle is 30% of the on time and 70% of the off time. The flow controller 128 outputs the valve drive signal to the valves 116, 118 and / or 120 to regulate the flow of the fluid through the input path 106, the column path 108, the shunt path 110 and / or the purge path 112. The conductivity of the column 104 is known or can be determined based on instrument properties and characteristics, so the measured pressure signal can be easily related to the flow rate.

[0039] In some examples, GC system 100 operates using fore pressure regulation, where flow controller 128 generates PWM valve drive signals for any one or more of valves 116, 118, and 120 based on pressure signals output by any one or more of pressure sensors 122, 124, and 126. In some examples, the inlet pressure is fore pressure regulated, while split flow path 110 and / or purge path 112 are back pressure regulated or fore pressure regulated.

[0040] Flow controller 128 may include any suitable hardware (eg, processor, circuitry, etc.) and / or software configured to control and / or interface with valves 116, 118, and 120 and pressure sensors 122, 124, and 126. Figure 1 The flow controller 128 is shown separate from the GC controller 117 , but the flow controller 128 may alternatively be implemented in whole or in part by the GC controller 117 .

[0041] It should be understood that the GC system 100 is merely illustrative and may be modified to suit a particular implementation. For example, in addition to or in place of the pressure sensors 122, 124, and / or 126, the GC system 100 may further include a flow sensor that measures the flow rate of the input path 106, the column path 108, the shunt path 110, and / or the purge path 112. The flow sensor may be any suitable type of sensor configured to measure flow rate, such as a mass flow sensor or a combination of a pressure sensor and a restrictor. In some examples, a flow sensor is used instead of a pressure sensor.

[0042] In some examples, flow control system 115 does not adjust or change the operation of valves 116, 118, and 120, but maintains the duty cycle of the valve drive signal even when a pressure change is detected. For example, pressure may be regulated using back pressure regulation or forward pressure regulation.

[0043] When the sample is injected into the inlet 102, the sample evaporates at the high temperature of the inlet. The increased vapor volume sample momentarily increases the pressure within the inlet 102. The flow control system 115 detects the increased pressure [TP387104CNSEC1]

[0044] (or change in flow rate), and responds to the increased pressure by adjusting the PWM valve drive signals to valves 116 , 118 , and / or 120 to reduce the pressure within inlet 102 .

[0045] Figure 2 An exemplary graph 200 of a PWM valve drive signal (e.g., for valve 116) over time during a time period including the injection of a sample into inlet 102 is shown. The PWM valve drive signal may be extracted from raw time data obtained or generated by flow control system 115 (e.g., by valves 116, 118, and / or 120, by pressure sensors 122, 124, and / or 126, and / or by flow controller 128). Figure 2 In the example of , the PWM valve drive signal is a voltage signal supplied to a valve (e.g., valve 116). However, in other examples, the PWM valve drive signal may be a current signal or a digital signal representing a voltage or current supplied to the valve. Curve 202 shows the average voltage (V) of the PWM valve drive signal as a function of time during the time period. It can be seen that the PWM valve drive signal is in a steady state before time t0, indicating that the pressure within the inlet 102 is in a steady state. A volume of sample is injected into the inlet 102 at time t0, which increases the pressure within the inlet 102. Curve 202 includes a disturbance 204 indicating the response of the flow controller 128 to return the pressure within the inlet 102 to a steady state. As shown in the waveform of disturbance 204, the average PWM valve drive signal initially decreases (e.g., to close the valve and reduce the pressure within the inlet 102). Feedback control of the PWM valve drive signal continues until the system returns to a steady state at time t1 (approximately 85 seconds after injection at time t0). The damping of the flow control system 115 may be adjusted to change the time to return to a steady state.

[0046] Although not shown, measurements of inlet pressure over time during a period of time including injection (as measured by one or more pressure sensors (such as pressure sensors 122, 124, and / or 126)) may also have a similar waveform with perturbations, such as in FIG. Figure 2Similarly, the flow rate over time measurements (such as measured by one or more flow sensors) during the time period including the injection may also have a similar waveform with disturbances, such as in Figure 2 However, the waveform of the inlet pressure measurement or flow rate measurement will likely be perturbed in the opposite direction of the PWM valve drive signal. For example, when the PWM valve drive signal decreases, the pressure measurement or flow rate measurement will increase, and vice versa.

[0047] The PWM valve drive signal, inlet pressure, and flow rate are flow control parameters that can be extracted from the raw time data and used alone or in combination to detect and / or diagnose unsuccessful injections, as will be described in more detail below. The measured values ​​of the flow control parameters over time during the time period including the injection can be characterized by one or more characterizing metrics. Exemplary characterizing metrics may include, but are not limited to:

[0048] a. Baseline (e.g., steady-state) measurements of flow control parameters (e.g., Figure 2 About 7V in the middle);

[0049] [TP387104CNSEC1]

[0050] b. the maximum amplitude of the disturbance waveform (e.g., the maximum change in the duty cycle of the valve drive signal or the maximum change in the PWM valve drive signal, the maximum change in the inlet pressure, the change in the maximum flow rate through the input path 106, etc.), measured as the difference between the maximum value of the flow control parameter and the baseline;

[0051] c. the integrated change of the disturbance waveform, calculated as the sum of the differences between each value and the baseline (e.g., the integrated change of the PWM valve drive signal during the time period, the integrated inlet pressure during the time period, or the integrated flow rate during the time period);

[0052] d. the absolute value of the integrated change of the disturbance waveform, calculated as the sum of the absolute values ​​of the difference between each value and the baseline (e.g., the absolute value of the integrated change of the PWM valve drive signal during the time period, the absolute value of the integrated inlet pressure during the time period, or the absolute value of the integrated flow rate during the time period);

[0053] e. the root mean square (RMS) error of the disturbance waveform, calculated as the square root of the sum of the squared differences between each value and the baseline;

[0054] f. Duration of the perturbation waveform, calculated as the time difference from sample injection to reaching steady state again;

[0055] g. The oscillation period of the disturbance waveform; and

[0056] h. The sign (eg, positive or negative) of the first peak of the disturbance waveform (eg, the initial direction of the disturbance waveform).

[0057] It should be appreciated that other characterizing metrics may be extracted from the raw time data as may be appropriate for a particular implementation.The injection monitoring system may use any one or more of the characterizing metrics, alone or in combination, to detect an unsuccessful injection and, in some cases, diagnose the cause of an unsuccessful injection.

[0058] Figure 3 A functional diagram of an exemplary injection monitoring system 300 ("system 300") is shown. System 300 may be implemented in whole or in part by GC system 100 (e.g., by GC controller 117, by flow controller 128, and / or by some other computing system included in GC system 100). Alternatively, system 300 may be implemented in whole or in part separate from GC system 100 (e.g., a remote computing device, system, and / or server that is separate from but communicatively coupled to GC controller 117 or flow controller 128).

[0059] System 300 may include, but is not limited to, a memory 302 and a processor 304 selectively and communicatively coupled to each other. Memory 302 and processor 304 may each include hardware and / or software components (e.g., processor, memory, communication interface, instructions stored in the memory for execution by the processor, etc.) or be composed of these hardware and / or software components (e.g., processor, memory, communication interface, [TP387104CNSEC1]

[0060] In some examples, memory 302 and processor 304 may be distributed among multiple devices and / or multiple locations that may serve a particular implementation.

[0061] The memory 302 may maintain (e.g., store) executable data used by the processor 304 to perform any of the operations described herein. For example, the memory 302 may store instructions 306 that may be executed by the processor 304 to perform any of the operations described herein. The instructions 306 may be implemented by any suitable application, software, code, and / or other executable data instance.

[0062] The memory 302 may also maintain any data collected, received, generated, managed, used, and / or transmitted by the processor 304. For example, the memory 302 may maintain and / or store a cross-correlation algorithm, an injection classification model, and / or a vapor volume estimation model, as described below.

[0063] The processor 304 may be configured to perform (e.g., execute instructions 306 stored in the memory 302 to perform) various processing operations described herein. It will be appreciated that the operations and examples described herein are merely illustrative of many different types of operations that may be performed by the processor 304. In the following description, any reference to operations performed by the system 300 may be understood to be performed by the processor 304 of the system 300. Furthermore, in the description herein, any operations performed by the system 300 may be understood to include the system 300 directing or instructing another system or device to perform an operation.

[0064] Figure 4 An exemplary method 400 for detecting an unsuccessful injection is shown. Figure 4 Exemplary operations according to one embodiment are shown, but other embodiments may omit, add, reorder and / or modify Figure 4 Any of the operations shown.

[0065] At operation 402, the system 300 obtains flow control data from a flow control system (e.g., flow control system 115), which regulates the flow of a fluid through an inlet (e.g., inlet 102) of a GC system (e.g., GC system 100) based on a flow control parameter. The flow control data represents a measured value of a flow control parameter over time during a time period including the injection of a sample into the inlet. The flow control data includes raw time data and / or data representing one or more characterizing measurements extracted from the raw time data. In some examples, the system 300 obtains flow control data in response to the injection of a sample. For example, the system 300 may determine that an injection has been performed, and in response to determining that an injection has been performed, obtains flow control data. The system 300 may determine that an injection has been performed in any suitable manner, such as based on injection data transmitted by an automatic sampler and / or based on an injection schedule. In other examples, the system 300 obtains flow control data continuously or periodically, regardless of the execution of any injection.

[0066] In some examples, the flow control parameter is a PWM valve drive signal for a valve (e.g., valves 116, 118, and / or 120) of the flow control system. In other examples, the flow control parameter is a PWM valve drive signal for a valve (e.g., valves 116, 118, and / or 120) of the flow control system.

[0067] The pressure signal output by the pressure sensor of the flow control system (e.g., pressure sensors 122, 124, and / or 126) indicates the pressure within the inlet 102, at the diversion path 110, or at the purge path 112. In yet another example, the flow control parameter is a flow rate signal output by a flow sensor of the flow control system (e.g., a flow sensor on the input path 106).

[0068] The time period of flow control data may have any suitable duration. In some examples, the time period is set in advance (e.g., before injection or before performing a GC experiment), and is set to include the expected disturbance in the measured value of the flow control parameter, the time period before the expected disturbance (e.g., when the measured value of the flow control parameter is in a steady state), and the time period after the expected disturbance (e.g., when the measured value of the flow control parameter is in a steady state). The duration of the disturbance may depend on various factors, such as the damping of the flow control system (e.g., overdamping, underdamping, critical damping), the time constant of the feedback control of the flow control system, the volume of the injection, the volume of the inlet capacity, and the type of the inlet (e.g., SSL, PTV, etc.). In some examples, the range of the duration of the time period is from 10 seconds to 120 seconds. In other examples, the range of the duration of the time period is from 15 seconds to 60 seconds. In some examples, the time period includes a preset time period before the disturbance (e.g., 2 seconds to 5 seconds) and a preset time period after the disturbance (e.g., 2 seconds to 5 seconds). The flow control data may have any suitable sampling rate, such as 0.5 Hz to 1 kHz, 5 Hz to 500 Hz, 10 Hz to 100 Hz, or any other suitable sampling rate.

[0069] At operation 404, system 300 determines that injection is unsuccessful based on flow control data. Unsuccessful injection (also referred to as "error" injection in this article) is any injection in which a sample less than a threshold amount is injected into the inlet. The threshold amount can be the percentage (e.g., 95%) of the total volume or total volume of the sample expected or specified to be injected into the inlet. For example, the method parameters for GC experiments can specify a specific amount (e.g., 2.0 μL, 1.0 μL, 0.5 μL) of the sample to be injected into the inlet. Therefore, the automatic sampler can draw a specified amount of sample from a sample vial into a syringe, and then inject the contents of the syringe into the inlet. An injection less than a specified amount (e.g., due to less than a specified amount, injection bubbles, needle blockage, needle bending, etc. drawn from a sample vial) is an unsuccessful injection. Unsuccessful injection includes partial injection (wherein at least a portion of the sample is injected) and empty injection (wherein no sample is injected).

[0070] In some examples where the system 300 continuously or periodically obtains flow control data regardless of any injection, the system 300 may determine that the injection was unsuccessful in response to the injection of the sample. For example, the system 300 may determine that the injection has been performed, and in response to determining that the injection has been performed, determine whether the injection was unsuccessful. The system 300 may determine that the injection has been performed in any manner described herein. In another embodiment [TP387104CNSEC1]

[0071] In one example, system 300 can determine that an injection has been performed based on the flow control data (eg, based on detection of a disturbance in a measurement of the flow control data).

[0072] If the measured value (eg, perturbation) of the flow control parameter does not behave as expected in response to the injection, the system 300 determines that the injection was unsuccessful. The system 300 can determine that the flow control parameter is not behaving as expected in a variety of different ways.

[0073] In some examples, the system 300 determines that the flow control parameter does not perform as expected by comparing the flow control data with the reference flow control data and determining that the flow control data is different from the reference flow control data based on the comparison. The reference flow control data represents the expected measured value of the flow control parameter (e.g., PWM valve drive signal, inlet pressure or flow rate) over time during the time period including the reference injection. In some examples, the reference flow control data represents the measured value of the flow control parameter during the previous time period, and the previous time period includes the previous injection of the sample into the inlet during the current GC experiment. In these examples, the previous successful injection during the same GC experiment can be used as a reference injection for monitoring subsequent injections. In other examples, the reference flow control data represents the measured value of the flow control parameter during multiple previous time periods, and the multiple previous time periods include multiple previous injections of the sample into the inlet during the current GC experiment. In these examples, the flow control data associated with multiple previous successful injections are aggregated (e.g., averaged or otherwise statistically processed) to generate reference flow control data for reference injection.

[0074] In other examples, the reference flow control data is generated based on one or more previous injections prior to the current GC experiment. In some examples, the reference flow control data may be updated as subsequent injections (through the current GC experiment and / or any other GC experiment performed by the same or different GC system) are successfully performed.

[0075] System 300 can use any suitable technology to determine that flow control data is different from reference flow control data in any suitable manner. In some examples, system 300 will be cross-correlated with the flow control data associated with the current injection and the reference flow control data. Any suitable cross-correlation algorithm or technology can be used. Flow control data and reference flow control data are arranged in time relative to injection. Cross-correlation can compare any one or more characterization measures derived or extracted from flow control data and reference flow control data. In conventional cross-correlation techniques, time series data will be normalized to a reference value along the y-axis. However, in the example where the characterization measure includes the amplitude of the disturbance, system 300 will not normalize the time series data, because the amplitude of the disturbance indicates the vapor volume of the sample injected and is therefore used in the cross-correlation.

[0076] The system 300 may determine that the flow control data is different from the reference flow control data in any suitable manner, such as when the result of the cross-correlation is less than a threshold value. For example, where the correlation coefficient of the cross-correlation ranges from +1.0 (perfect positive correlation) to -1.0 (perfect negative correlation) (where 0 indicates no linear relationship), the threshold value may be +0.97, +0.95, +0.90, or any other suitable value.

[0077] In addition to or in lieu of using reference flow control data, the system 300 determines that the flow control parameters are not behaving as expected based on an injection classification model that is trained to classify the injection as successful or unsuccessful based on the flow control data. For example, the system 300 may extract data representing one or more characterizing measurements from the flow control data (e.g., from measurements of the flow control parameters over time during a time period associated with the injection), and apply the extracted characterizing measurement data to (e.g., input to) the injection classification model. In some examples, the characterizing measurement data applied to the injection classification model represents the maximum amplitude of the perturbation waveform, the integral change of the perturbation waveform, and / or the absolute value of the integral change of the perturbation waveform, as described above. In some examples, data representing one or more other experimental condition parameters may also be applied as input to the injection classification model, including but not limited to the split ratio (when operating in split mode), the type of inlet, the volume of the inlet, the temperature of the inlet, the pressure in the inlet, the type of solvent for the sample, and the type of carrier gas.

[0078] Based on the inputs to the injection classification model, the injection classification model classifies the injection as successful or unsuccessful. The injection classification model and an exemplary method for training the injection classification model are described in more detail below.

[0079] In yet another example, the system 300 determines that the flow control parameters are not behaving as expected by determining a theoretical vapor volume of an injection, estimating an actual vapor volume of the injection, and comparing the estimated vapor volume of the injection to the theoretical vapor volume of the injection.

[0080] System 300 can calculate the theoretical vapor volume of the injected sample using the ideal gas law according to equation (1):

[0081]

[0082] Where V t (g) is the theoretical vapor volume of the sample injected, n is the amount of sample injected (in moles), R is the universal gas constant, T is the temperature at the inlet, and P is the pressure in the inlet. The amount of sample injected n can be calculated according to formula (2):

[0083]

[0084] Wherein V(l) is the volume of the sample in the liquid phase (before injection), p is the density of the sample in the liquid phase, and M is the molar mass of the sample in grams (g) / mol. As described above, the sample includes an analyte of interest dissolved in a solvent. In the case where the amount of the analyte of interest is low (e.g., the sample solution is dilute), the contribution of the analyte of interest can be assumed to be negligible and therefore ignored. As described above, the method parameters for a GC experiment can specify the volume V(l) of the sample to be injected into the inlet, and other information can be accessed from the GC system 100, from a remote computing system (e.g., a server), and / or input by a user. In some examples, the theoretical vapor volume V of the injected sample t (g) Can be calculated or estimated using non-ideal gas equations, such as the van der Waals equation.

[0085] The system 300 can estimate the actual vapor volume of the injection based on a vapor volume estimation model, which is trained to estimate the actual vapor volume of the sample injected based on the flow control data. The method is based on the fact that the perturbation characteristics of the flow control parameters are at least partially based on the actual vapor volume of the sample injected. Therefore, the system 300 can extract data representing one or more characterizing measurements from the original time data (e.g., from the measured values ​​of the flow control parameters over time during the time period associated with the injection), and input the extracted characterizing measurement data into the vapor volume estimation model. One or more other experimental condition parameters can also be input into the injection classification model, including but not limited to the split ratio (when operating in split mode), the type of inlet, the volume of the inlet, the temperature of the inlet, the pressure in the inlet, the solvent type of the sample, and the type of carrier gas.

[0086] Based on the input to the vapor volume estimation model, the vapor volume estimation model estimates the actual vapor volume V of the sample injected into the inlet e (g). The vapor volume estimation model and an exemplary method for training the vapor volume estimation model are described in more detail below.

[0087] The system 300 can calculate the estimated actual vapor volume V of the injected sample. e (g) and the theoretical vapor volume V of the injected sample t (g) performing a comparison. Based on (e.g., in response to) determining the estimated actual vapor volume V e (g) and the theoretical vapor volume V t The change in (g) exceeds a threshold amount, such as exceeding a threshold volume amount (e.g., 0.05 μL, 0.1 μL, etc.) or exceeding a theoretical vapor volume V t(g) exceeds a threshold percentage (e.g., 5%, 10%, etc.), the system 300 may determine that the injection was unsuccessful. The estimated actual vapor volume V e The percentage change of (g) can be given, for example, by the following formula (3):

[0088]

[0089] [TP387104CNSEC1]

[0090] Reference again Figure 4 At operation 406, the system 300 directs the GC system (e.g., GC system 100) to perform a mitigation operation to mitigate the unsuccessful injection of the sample based on (e.g., in response to) determining that the injection was unsuccessful. In some examples, the mitigation operation includes providing a notification of the unsuccessful injection. The notification may be provided via a display associated with the system 300 or with the GC system (e.g., included therein or communicatively coupled thereto). In additional or alternative examples, the mitigation operation includes discarding data associated with the unsuccessful injection. For example, the GC system (e.g., detector 109 or GC controller 117) may discard any data obtained by GC analysis of the sample that was not successfully injected. In yet another example, the mitigation operation includes a diagnostic process for determining the cause of the unsuccessful injection. Reference will be made to Figure 5 An exemplary diagnostic procedure is described.

[0091] Figure 5 An exemplary method 500 for performing a GC experiment with injection monitoring and diagnostics is shown. Figure 5 Exemplary operations according to one embodiment are shown, but other embodiments may omit, add, reorder and / or modify Figure 5 Any of the operations shown.

[0092] A sample is injected into an inlet of a gas chromatograph at operation 502. The injection may be performed in any manner described herein.

[0093] At operation 504, the system 300 checks whether the injection was successful. The system 300 can perform operation 504 in any manner described herein. If the system 300 determines that the injection was successful, the process returns to operation 502 for the next injection. If the system 300 determines that the injection was unsuccessful, the system 300 performs a diagnostic process to determine the cause of the unsuccessful injection.

[0094] The diagnostic process begins at operations 506 and 508, which are performed to check whether the syringe needle is damaged (e.g., bent) so that the needle cannot pierce the septum. At operation 506, the system 300 guides the GC system (e.g., GC system 100) to perform a "blank injection" by inserting the needle through the septum into the inlet when the inlet pressure is relatively high but no sample is injected (or the injection of the sample is delayed long enough to allow any response of the flow control system to return to a steady state). In some cases, the GC system may need to increase the pressure in the inlet to perform a blank injection. For example, the pressure for the blank injection may be greater than or equal to a threshold pressure level of 20psig, 25psig, 30psig, 40psig, or even 45psig. Piercing or opening the septum when the inlet is under high pressure will usually cause a temporary reduction in pressure due to the escape of carrier gas through the hole in the septum (as opposed to a temporary increase in pressure when the sample is injected and evaporated). This temporary reduction in pressure in the inlet results in a characteristic disturbance in the measured value of the flow control parameter (eg, PWM valve drive signal, inlet pressure and / or flow rate).

[0095] At operation 508, the system 300 determines whether the inlet is leaking slightly as expected. Operation 508 may be performed using any of the methods described above with reference to method 400. For example, the system 300 [TP387104CNSEC1]

[0096] Flow control data may be obtained that represents the measured values ​​of the flow control parameter over time during a time period that includes the empty injection. The system 300 may then determine based on the flow control data whether the measured values ​​of the flow control parameter do not behave as expected in response to the empty injection (e.g., whether the measured values ​​of the flow control parameter include a perturbation characteristic of a slight leak through the septum).

[0097] If the system 300 determines that the measured value of the flow control parameter does not perform as expected, the system 300 determines that the inlet is not leaking slightly as expected, and therefore, the needle is damaged (e.g., bent). Then, processing of method 500 continues to operation 510. At operation 510, the system 300 determines that the needle is damaged and directs the GC system to perform a needle correction operation. The needle correction operation may include providing a notification (e.g., via a display associated with the GC controller 117 and / or a display associated with the system 300) that the needle may be damaged and / or should be inspected. Additionally or alternatively, the needle correction operation may include directing the GC system's autosampler to discard the syringe and use a replacement syringe. Then, processing returns to operation 502 to perform the next injection or repeat the injection.

[0098] However, if the system 300 determines that the measured values ​​of the flow control parameters behave as expected, then the system 300 determines that the inlet is slightly leaking as expected, and therefore, the needle is not damaged. Processing of the method 500 then continues to operation 512 for the next step of the diagnostic process (operations 512 and 514) to check for vial errors (e.g., whether the sample vial is empty or too low).

[0099] At operation 512, the system 300 directs the GC system to perform additional sample injections using additional samples drawn from different vials (eg, sample vials or wash vials). The additional injections may be performed in any manner described herein.

[0100] At operation 514, the system 300 obtains additional flow control data representing measured values ​​of flow control parameters during a time period including the additional injection, and checks whether the additional injection performed at operation 512 is successful based on the additional flow control data. The system 300 may perform operation 514 in any manner described herein. If the system 300 determines that the additional injection performed at operation 512 is successful, the system 300 determines that the vial used for injection at operation 502 may be empty, and proceeds to operation 516.

[0101] At operation 516, the system 300 directs the GC system to perform a vial correction operation. In some examples, the vial correction operation includes providing a notification (e.g., via a display associated with the GC controller 117 and / or a display associated with the system 300) that the vial used at operation 502 may be empty or too low. Additionally or alternatively, the vial correction operation may include discarding any data acquired at operation 502. Processing of the method 500 then returns to operation 502 for the next iteration. At [TP387104CNSEC1]

[0102] In some examples, operation 516 may be omitted such that processing of method 500 returns to operation 502 in response to determining that the injection performed at operation 512 was successful.

[0103] Referring again to operation 514, if the system 300 determines that the additional injection is unsuccessful, the system 300 proceeds to operation 518. At operation 518, the system 300 determines that there is a problem with the syringe (e.g., the needle may be blocked or the syringe plunger is bent), and directs the GC system to perform a syringe correction operation. The syringe correction operation may include providing a notification that there may be a problem with the syringe and / or that the syringe should be checked (e.g., via a display screen associated with the GC controller 117 and / or a display screen associated with the system 300). Additionally or alternatively, the syringe correction operation may include directing the GC system's automatic sampler to discard the syringe and use a replacement syringe. Additionally or alternatively, the syringe correction operation may include discarding any data acquired at operation 502. Then, processing returns to operation 502 to perform the next injection or repeat the injection.

[0104] An illustrative method for training an injection classification model and a vapor volume estimation model will now be described. Figure 6 A block diagram of an exemplary training phase 600 is shown, in which a training module 602 trains a machine learning model 604 to classify an injection or estimate the actual vapor volume of an injected sample using training data 606 and an evaluation unit 608. When trained as described herein, the machine learning model 604 can implement the injection classification model or vapor volume estimation model used in method 400 or method 500.

[0105] The training module 602 may perform any suitable heuristic steps, processes, and / or operations that may be configured to train the machine learning model 604. In some examples, the training module 602 is implemented by a hardware component and / or a software component (e.g., a processor, a memory, a communication interface, instructions stored in the memory for execution by the processor, etc.). In some examples, the training module 602 is implemented by the system 300 or any component or specific implementation thereof. For example, the training module 602 may be implemented by the GC controller 117. Alternatively, the training module 602 may be implemented by a computing system (e.g., a personal computer or a remote server) that is separate from the GC system 100 but communicatively coupled thereto.

[0106] In some examples, the machine learning model 604 is implemented using one or more supervised and / or unsupervised learning algorithms. In some examples where the trained machine learning model 604 implements a sample classification model, the machine learning model 604 is implemented by a classification algorithm such as, but not limited to, an AdaBoost classifier, a gradient boosting classifier, a random forest classifier, or a support vector classifier. In some examples where the trained machine learning model 604 implements a vapor volume estimation model, the machine learning model 604 is implemented by a neural network (e.g., a convolutional neural network (CNN)) having an input layer, one or more hidden layers, and an output layer. In some examples, the CNN includes a long short-term memory (LSTM) network. In other examples where the trained machine learning model 604 implements a vapor volume estimation model, [TP387104CNSEC1]

[0107] The machine learning model 604 is implemented by a multivariate regression model, such as but not limited to a LASSO regression model, a ridge regression model, a boosted decision tree regression model, a decision forest regression model, a fast forest quantile regression model, or an ordinal regression model.

[0108] The training data 606 includes a set of training examples 610 (e.g., training examples 610-1 to 610-N). The training data 606 can be generated in any suitable manner. In some examples, the training data 606 is generated based on a series of injections performed over time. Each training example 610 corresponds to a specific injection and includes input data 612 and target output data 614.

[0109] The input data 612 includes flow control data, including raw time data associated with the corresponding injection and / or data representing one or more characterization metrics derived from the raw time data. The characterization metrics may include any of the above-mentioned characterization metrics that characterize the measured values ​​of the flow control parameters over time during the time period including the injection. In some examples, the input data 612 also includes data representing experimental condition parameters, such as but not limited to the split ratio (when the injection is performed in the split mode), the type of inlet, the volume of the inlet, the temperature of the inlet, the pressure in the inlet, the type of solvent for the sample, and the type of carrier gas.

[0110] The target output data 614 is a known expected output from the machine learning model 604 and can be used to evaluate the output of the machine learning model 604. When the trained machine learning model 604 implements an injection classification model, the target output data 614 represents the classification of the injection, for example, "successful" or "unsuccessful" (or other similar or suitable classifications). When the trained machine learning model 604 implements a vapor volume estimation model, the target output data 614 represents the vapor volume of the injected sample. The vapor volume of the injected sample can be determined in any suitable manner, including empirically and / or theoretically as described above.

[0111] like Figure 6 As shown, the input data 612 of the training example 610-1 is provided as an input vector to the machine learning model 604, which is trained to provide processed output data 616 (e.g., injection classification or estimated vapor volume). The target output data 614 can be provided as an input to the evaluation unit 608, which is configured to determine (e.g., calculate) an evaluation value provided to the machine learning model 604 based on the processed output data 616 output from the machine learning model 604 and the input data 612. Based on the evaluation value, the training module 602 can adjust one or more model parameters of the machine learning model 604. Then, the machine learning model 604 can be trained for the next training example (e.g., training example 610-2), and the training can be performed in all training examples 610. The training for the training example 610 can be repeated.

[0112] [TP387104CNSEC1]

[0113] In some examples, training data 606 is split into two subsets of data such that a first subset of training data 606 is used to train machine learning model 604 and a second subset of training data is used to score machine learning model 604. For example, training data 606 may be split such that a first percentage (e.g., 75%) of training examples 610 is used as a training set for training machine learning model 604 and a second percentage (e.g., 25%) of training examples 610 is used as a scoring set to generate an accuracy score for machine learning model 604.

[0114] Figure 7 An exemplary method 700 is shown that may be performed to train a machine learning model 604 to classify an injection of a sample or estimate the vapor volume of an injected sample. Figure 7 Exemplary operations according to one embodiment are shown, but other embodiments may omit, add, reorder and / or modify Figure 7 One or more operations of method 700 are depicted. Figure 7Each of the operations of the depicted method 700 may be performed in any of the ways described herein.

[0115] At operation 702, the training module 602 obtains flow control data associated with a plurality of injections. The flow control data associated with each injection represents measured values ​​of a flow control parameter over time during a time period including the associated injection.

[0116] At operation 704, the training module 602 generates training data 606 including multiple training examples 610 based on the flow control data. Each training example 610 includes input data 612 and target output data 614. The input data 612 includes data representing one or more characterization metrics derived from the flow control data associated with the corresponding injection. The target output data 614 is a known expected output from the machine learning model 604. When the machine learning model 604 is trained to classify the injection, the target output data 614 represents the classification of the injection, for example, "successful" or "unsuccessful" (or other similar or suitable classification). When the machine learning model 604 is trained to estimate the vapor volume of the sample injected, the target output data 614 represents the vapor volume of the sample injected. The vapor volume of the sample injected can be determined in any suitable manner, including the empirical and / or theoretical manner as described above.

[0117] At operation 706, the training module 602 uses the training data 606 to train the machine learning model 604 to provide processed output data 616 that classifies the injection and / or estimates the vapor volume of the injection. Once trained, the machine learning model 604 can be used in the method 400 and / or the method 500 to determine whether the injection is successful.

[0118] Figure 8 Shown in Figure 6 illustrative method 800 for training a machine learning model 604 using training examples 610 during a training phase 600 of Figure 8 Exemplary operations according to one embodiment are shown, but other embodiments may omit, add, reorder and / or modify one of the methods 800 [TP387104CNSEC1]

[0119] or multiple operations. Figure 8 Each of the operations of the depicted method 800 may be performed in any of the ways described herein.

[0120] At operation 802, the training module 602 uses the machine learning model 604 to generate processed output data 616 (e.g., injection classification data or estimated sample vapor volume data) based on the input data 612 in the training example 610-1 (e.g., flow control data, such as characterization metric data derived from the raw time data).

[0121] At operation 804, the training module 602 determines an evaluation value based on the target output data 614 and the processed output data 616 in the training example 610-1. Figure 6 As shown, the training module 602 may provide the processed output data 616 generated by the machine learning model 604 and the target output data 614 in the training example 610-1 to the evaluation unit 608. The evaluation unit 608 may determine an evaluation value based on the target output data 614 and the processed output data 616. The evaluation value is any value representing a comparison of the target output data 614 with the processed output data 616, such as a mean square error. Other specific implementations for determining the evaluation value are also possible and contemplated.

[0122] At operation 806, the training module 602 adjusts one or more model parameters of the machine learning model 604 based on the determined evaluation value. Figure 6 As shown, the training module 602 can back-propagate the evaluation value determined by the evaluation unit 608 to the machine learning model 604, and adjust the model parameters of the machine learning model 604 (e.g., the weight values ​​assigned to various data elements in the training example 610) based on the evaluation value.

[0123] In some embodiments, the training module 602 may determine whether the model parameters of the machine learning model 604 have been sufficiently adjusted. For example, the training module 602 may determine that the machine learning model 604 has been subjected to a predetermined number of training cycles and has therefore been trained using a predetermined number of training examples. Additionally or alternatively, the training module 602 may determine that the evaluation value satisfies a predetermined evaluation value threshold for a threshold number of training cycles, and therefore determines that the model parameters of the machine learning model 604 have been sufficiently adjusted. Additionally or alternatively, the training module 602 may determine that the evaluation value remains substantially unchanged for a predetermined number of training cycles (e.g., the difference between the evaluation values ​​calculated in sequential training cycles satisfies the difference threshold), and therefore determines that the model parameters of the machine learning model 604 have been sufficiently adjusted.

[0124] In some embodiments, in response to determining that the model parameters of the machine learning model 604 have been sufficiently tuned, the training module 602 can determine that the training phase of the machine learning model 604 has been completed and select the current values ​​of the model parameters as the values ​​of the model parameters in the trained machine learning model 604. The trained machine learning model 604 can implement an injection classification model or a vapor volume estimation model, as the case may be.

[0125] [TP387104CNSEC1]

[0126] In some examples, the machine learning model 604 is trained based on training data 606 acquired during multiple different experiments performed under different sets of experimental conditions. Thus, the trained machine learning model 604 can be used under a wide range of experimental conditions. The experimental conditions include, but are not limited to, the split ratio (when operating in split mode), the type of inlet, the volume of the inlet, the temperature of the inlet, the pressure in the inlet, the type of solvent for the sample, and the type of carrier gas. The machine learning model 604 can be trained in any suitable manner for use under a wide range of experimental conditions.

[0127] In some examples, the training data 606 includes multiple subsets of the training data. Each subset of the training data is acquired based on a different set of experimental conditions. The training module 602 may train the machine learning model 604 for each separate subset of the training data in multiple training stages in a row. For example, the training module 602 may train the machine learning model 604 for a first subset of the training data 606 in a first training stage. When the training with the first subset is completed, the training module 602 may train the machine learning model 604 for a second subset of the training data 606 in a second training stage. When the training with the second subset is completed, the training module 602 may train the machine learning model 604 for a third subset of the training data 606 in a third training stage, and so on.

[0128] Alternatively, data from multiple different subsets of training data can be mixed so that the machine learning model 604 is trained on different subsets of training data in one training phase. For example, training examples from various different subsets of training data can be mixed (e.g., randomly) to form training data 606.

[0129] In some examples, the machine learning model 604 is trained based on training data 606 configured for specific experimental conditions. In such examples, the trained machine learning model 604 can then be used only in subsequent iterations of that specific experiment.

[0130] In some examples, the machine learning model 604 may be refined or further trained in real time during the analysis experiment. In some embodiments, the training module 602 may continue to collect training examples 610 over time during the experiment and train the machine learning model 604 using the collected training examples 610. For example, when the training module 602 collects one or more additional training examples from one or more data sources, the training module 602 may update the plurality of training examples to include both the existing training examples and the additional training examples, and train the machine learning model 604 using the updated plurality of training examples according to the training process described herein. Additionally or alternatively, the training module 602 may periodically collect additional training examples from one or more other data sources, update the plurality of training examples to include both the existing training examples and the additional training examples, and train the machine learning model 604 using the updated plurality of training examples at predetermined intervals.

[0131] [TP387104CNSEC1]

[0132] The trained machine learning model 604 may also be scored and / or updated (e.g., retrained) in real-time during the analysis experiment based on data acquired during the analysis experiment (e.g., based on an analysis acquisition or scan). At different times throughout the analysis experiment, the system 300 may perform evaluations using the analysis data that has been acquired up to that point to evaluate the performance of the trained machine learning model 604. The evaluations may be performed at any suitable time, such as periodically (e.g., every nth acquisition), randomly, or in response to a triggering event (e.g., detection of agglomeration or peak broadening exceeding a threshold amount). Each evaluation may evaluate the quality of the trained machine learning model 604. If the system 300 determines during an evaluation that an error condition is met, the system 300 may retrain and / or update the machine learning model 604 using the acquired experimental data.

[0133] Various modifications may be made to the methods, apparatus, and systems described herein. For example, although the methods described herein are described as being performed in real time during a GC experiment, the methods may be performed at any other time (e.g., after the GC experiment is complete) to determine whether the acquired data is reliable. If it is determined that the data has been acquired using an erroneous injection, the data may be discarded.

[0134] The various examples above are described as using flow control data (e.g., a measurement of the inlet pressure) based on the output of pressure sensor 122 to determine whether the injection is unsuccessful. In other examples, flow control data based on the output of pressure sensor 124 (on shunt path 110) or pressure sensor 126 (on purge path 112) can be used to determine whether the injection is unsuccessful. Pressure sensor 124 and pressure sensor 126 may not be as sensitive as pressure sensor 122 (which measures the inlet pressure), but may still exhibit disturbances in the flow control data. In other examples, flow control data based on the output of any two or three pressure sensors (e.g., pressure sensors 122 and 124; pressure sensors 122 and 126; pressure sensors 124 and 126; or pressure sensors 122, 124, and 126) can be used. In a similar manner, flow control data based on the output of any one or more flow sensors can be used to determine whether the injection is unsuccessful.

[0135] The various examples above are described as using a single flow control parameter (e.g., PWM valve drive signal, inlet pressure, or flow rate) to determine whether an injection is unsuccessful. In other examples, any two or more flow control parameters may be used. For example, the system 300 may apply both the PWM valve drive signal and the inlet pressure to an injection classification model or a vapor volume estimation model to determine whether an injection is unsuccessful. In a similar manner, an injection classification model and / or a vapor volume estimation model may be trained based on two or more flow control parameters. For example, in Figure 6 In the example of FIG. 6 , the input data 612 may include flow control data for each of two different flow control parameters (e.g., PWM valve drive signal and inlet pressure). In practice, the injection classification model and / or vapor volume estimation model may be specific to [TP387104CNSEC1]

[0136] Training is performed with any set of representative measures of any set of flow control parameters associated with any combination of flow control data sources (eg, any one or more valves, any one or more pressure sensors, and / or any one or more flow sensors).

[0137] In some examples, the system 300 can be configured to classify an injection as successful (or complete), partial, or empty, and these different classifications can be used to diagnose an erroneous injection. For example, the system 300 can determine that the injection is classified as partial, and therefore omit operations 506, 508, and 510 of the method 500 (based on the assumption that a partial injection implies that the needle is not damaged). In a similar manner, the system 300 can determine that the needle is damaged based on classifying the injection as an empty injection.

[0138] Additionally or alternatively, system 300 may determine the vapor volume of the partial injection based on flow control data associated with the partial injection, as described above, and then adjust the acquired GC data by scaling the GC data based on the estimated vapor volume and theoretical vapor volume for the injection.

[0139] In some examples, one or more of the systems, components, and / or processes described herein may be implemented and / or performed by one or more appropriately configured computing devices. To this end, one or more of the above-described systems and / or components may include or be implemented by any computer hardware and / or computer-implemented instructions (e.g., software) embodied on at least one non-transitory computer-readable medium configured to perform one or more of the processes described above. Specifically, the system components may be implemented on one physical computing device, or may be implemented on more than one physical computing device. Thus, the system components may include any number of computing devices and may employ any number of computer operating systems.

[0140] In some examples, one or more of the processes described herein may be implemented at least in part as instructions embodied in a non-transitory computer-readable medium and executable by one or more computing devices. Typically, a processor (e.g., a microprocessor) receives instructions from a non-transitory computer-readable medium (e.g., a memory, etc.) and executes those instructions, thereby performing one or more processes, including one or more of the processes described herein. Any of a variety of known computer-readable media may be used to store and / or transmit such instructions.

[0141] Computer-readable media (also called processor-readable media) include any non-transitory media that participate in providing data (e.g., instructions) that can be read by a computer (e.g., by a processor of a computer). Such media can take many forms, including but not limited to non-volatile media and / or volatile media. Non-volatile media can include, for example, optical or magnetic disks and other permanent memory. Volatile media can include, for example, dynamic random access memory ("DRAM"), which typically constitutes main memory. Common forms of computer-readable media include, for example, disks, hard drives, tapes, any other magnetic media, [TP387104CNSEC1]

[0142] Compact Disc Read Only Memory (“CD-ROM”), Digital Video Disc (“DVD”), any other optical medium, Random Access Memory (“RAM”), Programmable Read Only Memory (“PROM”), Electrically Erasable Programmable Read Only Memory (“EPROM”), FLASH-EEPROM, any other memory chip or cartridge, or any other tangible medium from which a computer can read.

[0143] Fig. 9An exemplary computing device 900 is shown, which may be specifically configured to perform one or more of the processes described herein. Fig. 9 As shown, computing device 900 may include a communication interface 902, a processor 904, a storage device 906, and an input / output ("I / O") module 908 communicatively coupled to one another via a communication infrastructure 910. Fig. 9 An exemplary computing device 900 is shown, but Fig. 9 The components shown are not intended to be limiting. In other embodiments, additional or alternative components may be used. Fig. 9 Components of computing device 900 are shown.

[0144] The communication interface 902 may be configured to communicate with one or more computing devices. Examples of the communication interface 902 include, but are not limited to, a wired network interface (such as a network interface card), a wireless network interface (such as a wireless network interface card), a modem, an audio / video connection, and any other suitable interface.

[0145] The processor 904 generally represents any type or form of processing unit capable of processing data and / or interpreting, executing and / or directing one or more of the instructions, processes and / or operations described herein. The processor 904 may perform operations by executing computer-executable instructions 912 (e.g., applications, software, code and / or other executable data instances) stored in the storage device 906.

[0146] The storage device 906 may include one or more data storage media, devices, or configurations, and may employ any type, form, and combination of data storage media and / or devices. For example, the storage device 906 may include, but is not limited to, any combination of non-volatile media and / or volatile media described herein. Electronic data, including data described herein, may be temporarily and / or permanently stored in the storage device 906. For example, data representing computer executable instructions 912 configured to direct the processor 904 to perform any of the operations described herein may be stored within the storage device 906. In some examples, the data may be arranged in one or more databases residing within the storage device 906.

[0147] I / O module 908 may include one or more I / O modules configured to receive user input and provide user output. One or more I / O modules may be used to receive input for a single virtual experience. I / O module 908 may include any hardware, firmware, software, or combination thereof that supports input capabilities and output capabilities. For example, I / O module 908 may include hardware and / or software for capturing user input, [TP387104CNSEC1]

[0148] This includes, but is not limited to, a keyboard or keypad, a touch screen component (e.g., a touch screen display), a receiver (e.g., an RF or infrared receiver), a motion sensor, and / or one or more input buttons.

[0149] The I / O module 908 may include one or more devices for presenting output to a user, including but not limited to a graphics engine, a display (e.g., a display screen), one or more output drivers (e.g., a display driver), one or more audio speakers, and one or more audio drivers. In certain embodiments, the I / O module 908 is configured to provide graphical data to the display for presentation to the user. The graphical data may represent one or more graphical user interfaces and / or any other graphical content such as may serve a particular implementation.

[0150] In some examples, any of the systems, computing devices, and / or other components described herein may be implemented by computing device 900. For example, memory 302 may be implemented by storage device 906, and processor 304 may be implemented by processor 904.

[0151] Those of ordinary skill in the art will recognize that, although in the foregoing description, various exemplary embodiments have been described with reference to the accompanying drawings. However, it will be apparent that various modifications and alterations may be made thereto, and additional embodiments may be implemented, without departing from the scope of the invention as set forth in the appended claims. For example, certain features of one embodiment described herein may be combined with or substituted for features of another embodiment described herein. Accordingly, the description and drawings should be regarded as illustrative rather than restrictive.

[0152] The advantages and features of the present disclosure are further described by the following examples:

[0153] Embodiment 1. A system for gas chromatography, the system comprising: an inlet, the inlet being configured to receive a sample by injection; a column, the column comprising a stationary phase; a flow control system, the flow control system being configured to regulate the flow of a mobile phase through the inlet and the column based on a flow control parameter; and an injection monitoring system, the injection monitoring system being configured to perform a process comprising: obtaining flow control data, the flow control data representing a measured value of the flow control parameter over time during a time period including injecting the sample into the inlet; determining that the injection is unsuccessful based on the flow control data; and performing a mitigation operation based on determining that the injection is unsuccessful.

[0154] Example 2. A system according to Example 1, wherein the flow control system includes: a pressure sensor, which measures an inlet pressure; a valve, which regulates the flow of the mobile phase into the inlet; and a flow controller, which receives a pressure signal output by the pressure sensor and outputs a pulse width modulation (PWM) valve drive signal to the valve based on the pressure signal.

[0155] [TP387104CNSEC1]

[0156] Embodiment 3. The system of embodiment 2, wherein the flow control parameter comprises the inlet pressure.

[0157] Embodiment 4. The system of Embodiment 2, wherein the flow control parameter comprises the PWM valve drive signal.

[0158] Embodiment 5. A system according to embodiment 1, wherein: the flow control system includes a flow sensor, which measures the flow rate of the mobile phase into the inlet; and the flow control parameter includes the flow rate of the mobile phase.

[0159] Embodiment 6. A system according to embodiment 1, wherein: the flow control data indicates a disturbance in the measured value of the flow control parameter; and determining that the injection is unsuccessful includes determining that the disturbance did not behave as expected.

[0160] Embodiment 7. The system of embodiment 6, wherein determining that the disturbance did not behave as expected is based on one or more characterization metrics that characterize the disturbance.

[0161] Embodiment 8. The system of Embodiment 7, wherein the one or more characterization measures include at least one of: a maximum magnitude of the disturbance, an integrated change of the disturbance, or an absolute value of an integrated change of the disturbance.

[0162] Embodiment 9. An injection monitoring system for a gas chromatography system, the injection monitoring system comprising: one or more processors; and a memory storing executable instructions, wherein the executable instructions, when executed by the one or more processors, cause a computing device to perform a process comprising: obtaining flow control data from a flow control system, the flow control system being included in the gas chromatography system and being configured to regulate the flow of a fluid through an inlet of the gas chromatography system based on a flow control parameter, wherein the flow control data represents a measured value of the flow control parameter over time during a time period including injecting a sample into the inlet; determining that the injection is unsuccessful based on the flow control data; and directing the gas chromatography system to perform a mitigation operation to mitigate the unsuccessful injection of the sample based on the determination that the injection is unsuccessful.

[0163] Embodiment 10. A sample injection monitoring system according to embodiment 9, wherein: the flow control system includes a valve; and the flow control parameter includes a pulse width modulation (PWM) valve drive signal for the valve.

[0164] Embodiment 11. An injection monitoring system according to embodiment 10, wherein determining that the injection is unsuccessful is based on at least one of the following: a maximum change of the PWM valve drive signal during the time period, an integrated change of the PWM valve drive signal during the time period, or an absolute value of the integrated change of the PWM valve drive signal during the time period.

[0165] [TP387104CNSEC1]

[0166] Embodiment 12. A sample injection monitoring system according to embodiment 9, wherein: the flow control system includes a pressure sensor for measuring inlet pressure; and the flow control parameter includes a pressure signal output by the pressure sensor.

[0167] Embodiment 13. An injection monitoring system according to embodiment 12, wherein determining that the injection is unsuccessful is based on at least one of the following: the maximum change in inlet pressure during the time period, the integrated inlet pressure during the time period, or the absolute value of the integrated inlet pressure during the time period.

[0168] Embodiment 14. A sample injection monitoring system according to embodiment 9, wherein: the flow control system includes a flow sensor in the inlet; and the flow control parameter includes a flow rate signal output by the flow sensor.

[0169] Embodiment 15. An injection monitoring system according to Embodiment 9, wherein determining that the injection is unsuccessful includes applying the flow control data to an injection classification model, the injection classification model being trained to classify the injection as successful or unsuccessful based on the flow control data.

[0170] Embodiment 16. A sample injection monitoring system according to embodiment 9, wherein determining that the injection is unsuccessful includes: determining a theoretical vapor volume of the injection; estimating an actual vapor volume of the injection; and comparing the estimated actual vapor volume of the injection with the theoretical vapor volume of the injection.

[0171] Embodiment 17. A sample injection monitoring system according to embodiment 16, wherein estimating the actual vapor volume of the sample injection includes applying the flow control data to a vapor volume estimation model, and the vapor volume estimation model is trained to estimate the actual vapor volume of the sample injection based on the flow control data.

[0172] Embodiment 18. A sampling monitoring system according to embodiment 9, wherein determining that the sampling is unsuccessful includes: obtaining reference flow control data, wherein the reference flow control data represents the expected measured value of the flow control parameter over time during a time period including the sampling of the sample; and determining that the flow control data is different from the reference flow control data.

[0173] Embodiment 19. The injection monitoring system of Embodiment 18, wherein determining that the flow control data is different from the reference flow control data comprises cross-correlating the flow control data with the reference flow control data.

[0174] Embodiment 20. The injection monitoring system of embodiment 9, wherein the mitigation action comprises providing a notification that the injection was unsuccessful.

[0175] [TP387104CNSEC1]

[0176] Embodiment 21. An injection monitoring system according to embodiment 9, wherein the mitigation operation includes a diagnostic process, the diagnostic process comprising: performing a blank injection into the inlet when the pressure within the inlet is at or above a threshold pressure level; and detecting based on the blank injection that the measured value of the flow control parameter does not behave as expected in response to the blank injection.

[0177] Embodiment 22. According to the injection monitoring system of embodiment 21, the injection monitoring system further comprises: guiding the automatic sampler to perform additional injection of the sample using a replacement syringe.

[0178] Example 23. An injection monitoring system according to Example 9, wherein: the injected sample is drawn from a first vial; and the mitigation operation includes a diagnostic process, the diagnostic process including: performing an additional injection using an additional sample drawn from a second vial different from the first vial; obtaining additional flow control data from the flow control system, wherein the additional flow control data represents the measured value of the flow control parameter over time during a time period including the additional injection; and determining whether the additional injection is successful or unsuccessful based on the additional flow control data.

[0179] Embodiment 24. A non-transitory computer-readable medium storing instructions which, when executed, direct at least one processor of a computing device for a gas chromatography system to perform a process comprising: obtaining flow control data from a flow control system, the flow control system being configured to regulate the flow of a fluid through an inlet of the gas chromatography system based on a flow control parameter, wherein the flow control data represents a measured value of the flow control parameter over time during a time period including injecting a sample into the inlet; determining that the injection was unsuccessful based on the flow control data; and performing a mitigation operation based on determining that the injection was unsuccessful.

[0180] Embodiment 25. A computer-readable medium according to embodiment 24, wherein determining that the injection is unsuccessful includes applying the flow control data to an injection classification model, the injection classification model being trained to classify the injection as successful or unsuccessful based on the flow control data.

[0181] Embodiment 26. A computer-readable medium according to embodiment 24, wherein determining that the injection is unsuccessful includes: determining a theoretical vapor volume for the injection; estimating an actual vapor volume for the injection; and comparing the estimated actual vapor volume for the injection with the theoretical vapor volume for the injection.

[0182] Embodiment 27. A computer-readable medium according to Embodiment 26, wherein estimating the actual vapor volume of the injection includes applying the flow control data to a vapor volume estimation model, and the vapor volume estimation model is trained to estimate the actual vapor volume of the injection based on the flow control data.

[0183] [TP387104CNSEC1]

[0184] Embodiment 28. A computer-readable medium according to embodiment 24, wherein determining that the injection is unsuccessful includes: obtaining reference flow control data, wherein the reference flow control data represents expected measured values ​​of the flow control parameters over time during a time period including the injection of the sample; and determining that the flow control data is different from the reference flow control data based on a cross-correlation between the flow control data and the reference flow control data.

[0185] Embodiment 29. A computer-readable medium according to embodiment 24, wherein the flow control parameter includes at least one of the following: a pulse width modulation (PWM) valve drive signal for a valve of the flow control system, an inlet pressure measured by a pressure sensor of the flow control system, or a flow rate of the mobile phase measured by a flow sensor of the flow control system.

[0186] Embodiment 30. A computer-readable medium according to embodiment 24, wherein: the flow control data indicates a disturbance in the measured value of the flow control parameter; and determining that the injection was unsuccessful includes determining that the disturbance did not behave as expected.

[0187] Embodiment 31. The computer-readable medium of Embodiment 30, wherein determining that the disturbance does not behave as expected is based on one or more characterization metrics that characterize the disturbance.

[0188] Embodiment 32. A computer-readable medium according to embodiment 31, wherein the one or more characterization measures include at least one of the following: a maximum amplitude of the disturbance, an integrated change of the disturbance, or an absolute value of the integrated change of the disturbance.

Claims

1. A system for gas chromatography, the system comprising: an inlet configured to receive a sample via injection; a column comprising a stationary phase; a flow control system configured to regulate flow of a mobile phase through the inlet and the column based on a flow control parameter; and An injection monitoring system, the injection monitoring system being configured to perform a process comprising: obtaining flow control data representing measurements of the flow control parameter over time during a time period including injecting the sample into the inlet; determining that the injection was unsuccessful based on the flow control data; as well as A mitigation action is performed based on determining that the injection was unsuccessful.

2. The system of claim 1, wherein the flow control system comprises: a pressure sensor, the pressure sensor measuring an inlet pressure; a valve regulating flow of the mobile phase into the inlet; and A flow controller receives the pressure signal output by the pressure sensor and outputs a pulse width modulation (PWM) valve drive signal to the valve based on the pressure signal. 3 . The system of claim 2 , wherein the flow control parameter comprises the inlet pressure or the PWM valve drive signal.

4. The system of claim 1, wherein: The flow control system includes a flow sensor that measures a flow rate of a mobile phase into the inlet; and The flow control parameters include the flow rate of the mobile phase.

5. The system of claim 1, wherein: The flow control data indicates a disturbance in the measured value of the flow control parameter; and [TP387104CNSEC1] Determining that the injection was unsuccessful includes determining that the perturbation did not behave as expected.

6. The system of claim 5, wherein: Determining that the disturbance does not behave as expected is based on one or more characterization metrics characterizing the disturbance; and The one or more characterization measures include at least one of: a maximum magnitude of the disturbance, an integrated change of the disturbance, or an absolute value of an integrated change of the disturbance.

7. A sampling monitoring system for a gas chromatography system, the sampling monitoring system comprising: one or more processors; and A memory storing executable instructions that, when executed by the one or more processors, cause a computing device to perform a process comprising: obtaining flow control data from a flow control system included in the gas chromatography system and configured to regulate flow of a fluid through an inlet of the gas chromatography system based on a flow control parameter, wherein the flow control data represents a measured value of the flow control parameter over time during a time period including injecting a sample into the inlet; determining that the injection was unsuccessful based on the flow control data; as well as The gas chromatography system is directed to perform a mitigation operation to mitigate the unsuccessful injection of the sample based on determining that the injection was unsuccessful.

8. The sample injection monitoring system according to claim 7, wherein: The flow control system includes a valve; and The flow control parameters include a pulse width modulated (PWM) valve drive signal for the valve.

9. The injection monitoring system of claim 8, wherein determining that the injection is unsuccessful is based on at least one of: a maximum change in the PWM valve drive signal during the time period, an integrated change in the PWM valve drive signal during the time period, or an absolute value of an integrated change in the PWM valve drive signal during the time period.

10. The sample injection monitoring system according to claim 7, wherein: The flow control system includes a pressure sensor that measures an inlet pressure; and The flow control parameter includes a pressure signal output by the pressure sensor.

11. The injection monitoring system of claim 10, wherein determining that the injection is unsuccessful is based on at least one of: a maximum change in inlet pressure during the time period, an integrated inlet pressure during the time period, or an absolute value of the integrated inlet pressure during the time period.

12. The sample injection monitoring system according to claim 7, wherein: The flow control system includes a flow sensor within the inlet; and The flow control parameter includes a flow rate signal output by the flow sensor.

13. The injection monitoring system of claim 7, wherein determining that the injection is unsuccessful comprises applying the flow control data to an injection classification model, the injection classification model being trained to classify the injection as successful or unsuccessful based on the flow control data.

14. The injection monitoring system of claim 7, wherein determining that the injection is unsuccessful comprises: determining a theoretical vapor volume of the injection; estimating the actual vapor volume of the injection; as well as The estimated actual vapor volume of the injection is compared to the theoretical vapor volume of the injection.

15. The injection monitoring system of claim 7, wherein determining that the injection is unsuccessful comprises: obtaining reference flow control data representing expected measured values ​​of the flow control parameter over time during a time period including the introduction of a sample; as well as It is determined that the flow control data is different from the reference flow control data.

16. The injection monitoring system of claim 7, wherein the mitigation operation comprises a diagnostic process, the diagnostic process comprising: performing a null injection into the inlet when the pressure within the inlet is at or above a threshold pressure level; as well as The measured value of the flow control parameter detected based on the empty injection does not behave as expected in response to the empty injection.

17. The sample injection monitoring system according to claim 7, wherein: The injected sample was drawn from the first vial; and The mitigation operation includes a diagnostic process, and the diagnostic process includes: [TP387104CNSEC1] performing an additional injection using an additional sample drawn from a second vial different from the first vial; obtaining additional flow control data from the flow control system, wherein the additional flow control data represents measurements of the flow control parameter over time during a time period including the additional injection; and A determination is made whether the additional injection is successful or unsuccessful based on the additional flow control data.

18. A non-transitory computer readable medium storing instructions that, when executed, direct at least one processor of a computing device for a gas chromatography system to perform a process comprising: obtaining flow control data from a flow control system configured to regulate flow of a fluid through an inlet of a gas chromatography system based on a flow control parameter, wherein the flow control data represents a measured value of the flow control parameter over time during a time period including injecting a sample into the inlet; determining that the injection was unsuccessful based on the flow control data; and A mitigation action is performed based on determining that the injection was unsuccessful.

19. The computer-readable medium of claim 18, wherein determining that the injection is unsuccessful comprises applying the flow control data to an injection classification model, the injection classification model being trained to classify the injection as successful or unsuccessful based on the flow control data.

20. The computer readable medium of claim 18, wherein determining that the injection is unsuccessful comprises: determining a theoretical vapor volume of the injection; estimating the actual vapor volume of the injection; as well as The estimated actual vapor volume of the injection is compared to the theoretical vapor volume of the injection.

21. The computer readable medium of claim 18, wherein determining that the injection is unsuccessful comprises: obtaining reference flow control data representing expected measured values ​​of the flow control parameter over time during a time period including a reference injection; and [TP387104CNSEC1] The flow control data is determined to be different from the reference flow control data based on a cross-correlation between the flow control data and the reference flow control data.

22. A computer-readable medium according to claim 18, wherein the flow control parameter includes at least one of the following: a pulse width modulation (PWM) valve drive signal for a valve of the flow control system, an inlet pressure measured by a pressure sensor of the flow control system, or a flow rate of the mobile phase measured by a flow sensor of the flow control system.

23. The computer-readable medium of claim 18, wherein: The flow control data indicates a disturbance in the measured value of the flow control parameter; and Determining that the injection was unsuccessful includes determining that the perturbation did not behave as expected.

Images