Patent Application
20220128522
This patent application claims priority to European Application No. 20204158.8, filed Oct. 27, 2020, the contents of which are expressly incorporated by reference in their entirety, including any references contained therein.
The present disclosure pertains to training a neural network processor for providing diagnostic information of a controlled liquid chromatography pump unit. The present disclosure further pertains to manufacturing a controlled liquid chromatography pump unit having an auto-diagnostic facility. The present disclosure still further pertains to a controlled liquid chromatography pump unit having an auto-diagnostic facility. The present disclosure still further pertains to using a controlled liquid chromatography pump unit having an auto-diagnostic facility. The present disclosure still further pertains to manufacturing a system comprising a controlled liquid chromatography pump unit and a diagnostic unit. The present disclosure still further pertains to a system comprising a controlled liquid chromatography pump unit and a diagnostic unit. The present disclosure still further pertains to using a system comprising a controlled liquid chromatography pump unit and a diagnostic unit.
Liquid chromatography is an analytical method that takes advantage of the difference in affinity of substances with both a mobile and a stationary phase. The mobile phase in liquid chromatography consist of a fluid carrying various substances. The stationary phase is a column of porous material, through which the fluid is driven at high pressure. Liquid chromatography is widely used in hospitals, universities and various industries, both for analytical and for production applications. An example of the use of liquid chromatography is the demonstration of doping use in sports. In order to demonstrate the presence of the doping agent among the variety of substances excreted in the urine by the body, the doping agent, which has its own molecular characteristics, must be separated from the other substances. This is done using chromatographic techniques. Because each substance has its own ‘adhesion force’ to the stationary phase, they are carried along with the mobile phase faster or slower. In this way the doping agent is separated from the other substances.
For a proper performance of the method it is mandatory that the pump delivers the mobile phase at a highly accurate and precise flow rate to the column, to avoid undesired variations in the retention time. In a typical liquid chromatography pump unit this is accomplished by using two mutually cooperating pistons. The liquid chromatography pump unit comprises a controller that drives a pair of motors such that when the first one of the pistons presses the fluid into the column, the other one of the pistons sucks fluid from the reservoir. Therewith a stable pressure and flow rate can be achieved, provided that no pump faults occur. Nevertheless, even minor defects of the pump may cause deviations and/or variations in pressure and/or flow rate that render the liquid chromatography unreliable. It is often difficult to identify the cause of the deviations and/or variations. Accordingly, there is a need to facilitate a diagnosis, so that the pump can be promptly serviced and/or repaired to mitigate out-times of the apparatus.
It is an object of the present disclosure to provide means that enable a more efficient diagnosis of a controlled liquid chromatography pump unit.
According to the first aspect of the present disclosure, a controlled liquid chromatography pump unit having an auto-diagnostic facility is provided. For example a diagnostic facility is part of the controlled liquid chromatography pump unit. The auto-diagnostic facility may be activated by a control input from the operator, or may be activated automatically, for example upon startup or regularly, e.g. after every 100 hours of operation.
The controlled liquid chromatography pump for providing a controlled supply of a fluid retrieved from an inlet to an outlet, comprises a pump controller, at least one sensor and a trained neural network processor. The pump controller has control outputs connected to the liquid chromatography pump for providing control signals to the liquid chromatography pump in accordance with a pump controller state.
The pump controller drives the pump in a periodic manner, by periodically repeating a driving signal pattern. The pump controller may provide an output signal indicative for its operation state, also denoted as pump controller state signal. In an embodiment the pump controller state signal indicates that the pump controller is in a phase i/n of the driving period, wherein the driving period is partitioned into n portions (e.g. 32 portions) and i is an integer in the range of 1 to n. The pump controller state signal may for example be updated at the onset of every 1/n-th portion of the driving period. Additionally or alternatively, in case the pump controller operates according to a state sequence, the pump controller state signal may for each point in time indicate the specific state of the state sequence assumed by the pump controller at that point in time.
At least one sensor comprises a sensor output for providing a sense signal indicative for an operational characteristic of the pump. The trained neural network processor is connected to the sensor output of the at least one sensor and to at least one control outputs of the pump controller to receive one or more signals indicative for said pump controller state and to provide in response thereto a diagnostic output signal. In an embodiment the pump unit has a first pump section and a second pump section, each having a respective pressure chamber with a piston driven by a respective motor in response to the control signals of the pump controller. The at least one sensor may comprise sensors for measuring a pressure prevailing in several positions in a fluid channel defined by the pump and sensors for providing an indication of the piston positions, e.g. provided as encoders coupled to the pump motor. Also flow sensors may be contemplated. However, flow sensors are preferably omitted to keep costs at a modest level. The trained neural network processor is connected to the outputs of these sensors. Moreover, the trained neural network processor is connected to one or more outputs of the pump controller, e.g. the control outputs to receive signals indicative for the supplied one or more pump control signals. Alternatively or additionally it may be connected to outputs of the pump controller to receive one or more signals indicative for the pump controller state. It has been found that in this way a suitably trained neural network processor is capable to provide a proper diagnostic output signal which facilitates the detection of various pump faults that may in practice occur, such as a primary seal leakage, a secondary seal leakage, an inlet check valve leakage, an outlet check valve leakage, a flow path blockage, a pressure deviation and a pressure ripple.
Alternatively, or additionally the one or more diagnostic output signals may be indicative for a need of maintenance to avoid an expected occurrence of a pump fault in the (near) future. In this way it can be avoided that the pump fault actually does occur by a timely maintenance. The pump can continue to function within specifications and its operation has to be interrupted only during the execution of the maintenance. Therewith the duration of the non-operational state is reduced as compared to the situation that only the actual occurrence of a fault is signaled. On that case the duration of the non-operational state is not only determined by the maintenance time, but also includes the waiting time between the moment that the fault is detected and the moment that the maintenance operator has arrived.
In an embodiment, the controlled liquid chromatography pump unit comprises a unidirectional fluent path between said inlet and said outlet, a first fluent path section therein being defined between a first unidirectional valve and a second unidirectional valve and a second fluent path section therein being defined between the second unidirectional valve and the outlet. Therein a first piston pump with a first pump chamber communicates with the first fluent path section and a second piston pump with a second pump chamber communicates with the second fluent path section. The control signals of the pump controller are to drive the first and the second piston pump so that they mutually cooperate to maintain a constant flow. In an exemplary configuration the first and the second piston pump provide for a mutually parallel fluid paths from a common inlet to a common outlet. In this example the pumps operate in counter phase. I.e. when a first one of the pistons presses the fluid into the column, the other one of the pistons sucks fluid from the reservoir. In another exemplary configuration the first and the second piston pump are arranged in series. In this case the operation of the first and the second piston pump is not necessarily in counterphase during a full operational cycle. For example, while the second pump supplies the fluid from its chamber to the outlet, the first pump may first suck the fluid from the inlet, and pressurize the fluid already before the second pump has finished supplying.
In an example of this embodiment, the controlled liquid chromatography pump unit further comprises a n-ary pump section with an n-ary pump having n inputs and a single output coupled to the inlet of the unidirectional fluent path. The value n is an integer greater or equal than 2, for example four. Therewith the fluid can be provided with a plurality of solvents.
In an embodiment of the controlled liquid chromatography pump unit the trained neural network processor comprises a Multilayer Perceptron (MLP) cooperating with a feature extraction component that preprocesses the combination of input signals. The preprocessed input signals for example comprise a minimum value, a maximum value, and a mean value of each of the original input signals.
In another embodiment of the controlled liquid chromatography pump unit the trained neural network processor comprises a combination of a set of Convolutional Neural Networks (CNN) and a pair of Long Short-Term Memory (LSTM) layers. The presence of the CNNs obviate the use of a separate feature extraction and contrary to standard region based neural networks (RNNs), LSTMs are capable of learning long-term dependencies.
Alternatively, according to a second aspect a system is provided that comprises a controlled liquid chromatography pump unit and a diagnostic unit. In this embodiment the controlled liquid chromatography pump unit comprises a liquid chromatography pump, a pump controller, at least one sensor, and a pump interface. The pump controller of the pump unit in this system has control outputs connected to the liquid chromatography pump for providing control signals to the liquid chromatography pump in accordance with a pump controller state. The at least one sensor comprises a sensor output for providing a sense signal indicative for an operational characteristic of the pump. The pump interface is coupled to at least one sensor to receive said sensor output and is coupled to the pump controller to receive a combination of input signals that comprise the drive signals for driving the pump sections and one or more signals that are indicative for the pump controller state.
In the system according to the second aspect, the diagnostic unit comprises a trained neural network and a diagnostic unit interface. The diagnostic unit interface is configured to be communicatively coupled to the pump interface to enable the trained neural network processor to provide a diagnostic output indicative for a state of the controlled liquid chromatography pump unit on the basis of the combination of input signals.
Hence, in this alternative embodiment, the trained neural network processor is not part of the controlled liquid chromatography pump unit but is part of a separate diagnostic unit. Therewith a single trained neural network processor can be used for diagnosing a plurality of controlled liquid chromatography pump units. I.e. a diagnostic unit may be used in a maintenance department of a research institute for periodically checking the controlled liquid chromatography pump units of said institute, or for diagnosing on a case to case basis. In an embodiment the diagnostic unit may be coupled with its diagnostic unit interface to the pump interface by a wired connection, e.g. by directly engaging a diagnostic unit interface connector with a pump interface connector. Either one or both of the connectors may include a cable. Alternatively, communication between the pump interface and the diagnostic unit interface may take place via a wireless connection. The connection whether wired or wireless may be private or public, e.g. a local network or a public network. In some embodiments the diagnostic unit is used by a service provider that remotely performs the diagnosis for various clients, e.g. on regular basis or on a case to case basis. It is noted that the communication will mainly involve a transfer of the signals received by the pump interface from the pump controller and the at least one sensor, but typically the communication will also involve a transfer of signals from the diagnostic unit to the pump interface.
In some embodiments, the pump interface is configured to record one or more control signals of the pump controller and the sense signal of the at least one sensor during a period of time and transfers said recorded signals to the diagnostic unit interface.
It is noted that the controlled liquid chromatography pump unit as used in the system may be one of various embodiments, for example as described for the version of the controlled liquid chromatography pump unit with auto-diagnostic capability it may have two controlled pump sections. Also a n-ary input pump section may be comprised as described above.
Furthermore, various options are available for the diagnostic unit. In an exemplary embodiment thereof, the trained neural network processor comprises a Multilayer Perceptron (MLP) cooperating with a feature extraction component that preprocesses the combination of input signals. The preprocessed input signals for example comprise a minimum value, a maximum value, and a mean value of each of the original input signals. In another embodiment thereof, the trained neural network processor comprises a combination of a set of Convolutional Neural Networks (CNN) and a pair of Long Short-Term Memory (LSTM) layers.
A method of training a neural network processor for providing diagnostic information of a controlled liquid chromatography pump unit is claimed as a third aspect of the present disclosure. The trained neural network processor obtained with this method is suitable for application in the controlled liquid chromatography pump unit with auto-diagnostic facility according to the first aspect, as well as in the system according to the second aspect that comprises a controlled liquid chromatography pump unit and a diagnostic unit.
The claimed method according to the third aspect comprises an execution of the following sequence of steps:
a) simulating a pump controller generating one or more pump control signals in accordance with a pump controller state;
b) simulating a pump fault simulation signal representative for the presence or absence of a pump fault;
c) executing a simulation model of a pump unit, the simulation model providing one or more simulated sensor signals representing operational parameters of the simulated pump unit in response to the one or more simulated pump control signals, the simulation model selectively simulating a pump fault in response to the pump fault simulation signal;
d) supplying the neural network processor with a combination of input signals comprising the one or more simulated sensor signals and one or more signals indicative for said pump controller state and/or indicative for the supplied one or more simulated pump control signals;
e) the neural network processor computing one or more output signals in response to the supplied combination of input signals;
f) computing a loss function by comparison of a diagnostic state as indicated by the one or more output signals and a diagnostic state as indicated by the pump fault simulation signal;
training the neural network processor by feeding back a loss computed with the loss function;
wherein during one or more executed sequences the pump fault simulation signal indicates that a predetermined pump fault is to be simulated and during one or more executed sequences the pump fault simulation signal indicates that the predetermined pump fault is to be absent in the simulation.
Typically the neural network processor is trained to recognize a plurality of pump faults, such as the pump faults referred to above, e.g. a primary seal leakage, a secondary seal leakage, an inlet check valve leakage, an outlet check valve leakage, a flow path blockage, a pressure deviation and a pressure ripple. For each of the pump faults for which the neural network processor is to be trained, the simulation model is configured by the pump fault simulation signal to simulate an operation of the pump unit with that pump fault and is configured to provide one or more simulated sensor signals representing operational parameters of the simulated pump unit with that pump fault. During the training the neural network processor is provided with a combination of input signals comprising the one or more simulated sensor signals as well as one or more signals indicative for said pump controller state and/or indicative for the supplied one or more simulated pump control signals. In response to the supplied combination of input signals, the neural network processor computes one or more output signals that represent an estimated diagnostic state. For example, each output may be indicative for a particular pump fault. Subsequently a loss function is computed by comparison of an estimated diagnostic state as indicated by the one or more output signals and a diagnostic state as indicated by the pump fault simulation signal. Then the neural network processor is trained by feeding back a loss computed with the loss function. One option to perform feeding back is by back propagation. Typically, for each type of pump fault, as well as for the case that there is no pump fault the sequence of steps is repeated to improve the quality of the estimated diagnostic state. Noise may be introduced in one or more signals of the combination of input signals supplied to the neural network processor for a more robust training. Also the neural network processor may be trained to recognize a plurality of pump faults.
These and other aspects are described in more detail with reference to the drawing. Therein:
Like reference symbols in the various drawings indicate like elements unless otherwise indicated.
In general the pump controller 2, will use one or more sensors 63, 64, 66, 67, 68 to calculate the drive signals S24, S25. However, the exact operation of the pump controller is not the subject of this disclosure and therefore the sensory inputs of the controller are omitted.
The controlled liquid chromatography pump unit 1 of
In operation of the controlled liquid chromatography pump unit 1, the pump controller 2 drives the two pistons 42, 52. When one piston presses the fluid into the column, the other piston sucks fluid from the reservoir. In normal circumstances, therewith a desired flow rate can be accurately and precisely maintained, which is essential for a performing a proper chromatography measurement. Various circumstances however may in practice prevent the controlled liquid chromatography pump unit 1 from a proper operation.
As illustrated in
As noted above a proper operation of the controlled liquid chromatography pump unit 1 is essential to enable reliable chromatography measurements. For this purpose it is important that a failure can be rapidly identified, diagnosed and that the controlled liquid chromatography pump unit can be repaired, preferably by local personnel, so that inactivity is minimized. This is achieved in that the controlled liquid chromatography pump unit 1 has an auto-diagnostic facility including the trained neural network processor 7. In the embodiment shown, the trained neural network processor 7 comprises a respective output for issuing a respective diagnostic output signal 7a, 7b, 7c, 7d, 7e, 7f and 79, which each are indicative for the presence of the pump faults A-G referred to above. The output signals 7a-7g may be of a binary nature, e.g. indicating the diagnosed absence or presence of the corresponding pump fault. Alternatively the output signals may have a higher number of signal levels. For example the signal level may indicate the diagnosed probability or the severity of a fault with a value selected from a value range, wherein the minimum of the range signifies that absence of the fault is diagnosed and the maximum of the range signifies that it is highly likely that the fault is presence or that the fault is severe.
The presence of the trained neural network renders it possible to provide reliable diagnostics of the controlled liquid chromatography pump unit 1 even in the absence of flow sensors data. Therewith cost of material are modest. Whereas the trained neural network processor 7 in the first place issues a diagnostic output indicative of the nature of a fault, the diagnostic output may further comprise information to assist the local operator to efficiently repair the pump where possible, e.g. by specifying which parts need to be replaced, and which procedural steps are to be taken. If the local operator is not capable to handle a particular fault, the diagnostic information may provide information for a service team.
Of course, it is even more preferable if the occurrence of a pump fault can be avoided. In some embodiments the diagnostic output facilitates preventive maintenance. Therewith it can be avoided that operation of the pump has to be discontinued at an unfavorable point in time. For example the seal leakage A may be indicated as the amount of fluid leakage in mL/s. With such an output range, the auto-diagnostic facility may also be used for predictive maintenance, i.e. with the amount of fluid leakage and the increase of this amount over time, a proper time for replacement of the piston seal may be suggested. In the example shown, a further output is provided with signal 7o, with which it can be explicitly indicated if the diagnosis reveals that no fault was detected at all. By way of example the signals 7a-7g, 7o may be provided to drive a respective LED indicator at a housing of the controlled liquid chromatography pump unit 1. It is not necessary that the trained neural network processor 7 is permanently active. Maintenance personnel or an operator of the controlled liquid chromatography pump unit 1 may for example deliberately activate the trained neural network processor 7, e.g. by a control button 74 in case a diagnosis is requested. Whereas in this example respective outputs are provided that each provide a diagnostic output signal 7a-7g, 7o, it is alternatively possible that the trained neural network processor 7 is trained to provide a single diagnostic output signal having respective signal levels, each indicative for a particular type of fault.
As observed above, the one or more diagnostic output signals may be indicative for a need of maintenance to avoid an expected occurrence of a pump fault in the (near) future. In this way it can be avoided that the pump fault actually does occur by a timely maintenance. The pump can continue to function within specifications and its operation has to be interrupted only during the execution of the maintenance.
It is not necessary that a diagnostic facility is integrated in the controlled liquid chromatography pump unit.
Whereas
In the examples described with reference to
Neural networks are trained, using a large set of measurements of pumps operating under normal conditions and with faults such as primary seal leakage, secondary seal leakage B and others. For new pumps, pumps that have been brought to the market recently, such data sets are generally not available. In this disclosure these data sets are generated using simulations.
In a step S1 a pump controller (e.g. the pump controller 2 of
In step S2 a pump fault simulation signal PFC is simulated that is representative for the presence or absence of a pump fault, for example one of the pump faults A-G as discussed with reference to
In step S3 a simulation model of a pump unit is executed. The simulation model simulates the operation of the pump unit (e.g. the pump unit 3, 4, and 5 in
Step S1, Step S2 and Step S3 may be sequentially simulated in time instances where the S1 and S2 form the input of S3. The simulated sensor signals and/or other variables of S3 may be fed back into S1 and S2 after which the next sequence starts.
In step S4 a combination of input signals comprising the one or more simulated sensor signals PS is supplied to the neural network processor. The combination of input signals further comprises one or more signals OS indicative for the pump controller state and/or one or more signals indicative for the supplied one or more simulated pump control signals PC. In the example shown in
In step S5, the neural network processor computes one or more output signals NO in response to the supplied combination of input signals. The combination of input signals comprises at least the one or more simulated sensor signals. In addition the combination of input signals comprises either the one or more signals OS indicative for the pump controller state or the one or more simulated pump control signals PC or the pump controller state signals OS and the simulated pump control signals PC. As noted the neural network processor to be trained may be either a separate computational unit, or may be part of a common computational unit. The one or more output signals NO are a tentative diagnostic indication for the condition of the simulated controlled liquid chromatography pump unit. During the execution of the training method the quality of the output signals as a diagnostic indication gradually improves.
In step S6 a loss function is computed by comparison of a diagnostic state as indicated by the one or more output signals NO and a diagnostic state as indicated by the pump fault simulation signal PFC.
In step S7 the neural network processor is trained by feeding back a loss computed with the loss function. This can for example be achieved by a back propagation computation.
The steps S1-S7 can be repeated one or more times for various conditions indicated by the pump fault simulation signal PFC. Typically for each condition indicated by the pump fault simulation signal the sequence of steps is performed a plurality of times, therewith introducing noise in the input signals of the neural network processor to avoid over fitting. Also the repetitions will typically include a plurality of sequences for the case that the pump fault simulation signal PFC indicates the absence of a pump fault. The repetitions for a pump fault simulation signal PFC indicating a particular pump fault, combination of pump faults or absence of pump faults do not need to immediately succeed each other, but simulations for various conditions may be alternated.
The method can be discontinued once it is known, or can be assumed that the neural network processor is sufficiently trained to function as a trained neural network processor 7 in a controlled liquid chromatography pump unit 1 with auto-diagnostic facility, as shown in
After a sufficient number of cycles a trained neural network processor 7b is obtained as shown in
In one exemplary embodiment, as shown in
An example of an LSTM can be seen in
The terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention. As used herein, the singular forms “a”, “an” and “the” are intended to include the plural forms as well, unless the context clearly indicates otherwise. It will be further understood that the terms “comprises” and/or “comprising,” when used in this specification, specify the presence of stated features, integers, steps, operations, elements, and/or components, but do not preclude the presence or addition of one or more other features, integers, steps, operations, elements, components, and/or groups thereof. Further, unless expressly stated to the contrary, “or” refers to an inclusive or and not to an exclusive or. For example, a condition A or B is satisfied by any one of the following: A is true (or present) and B is false (or not present), A is false (or not present) and B is true (or present), and both A and B are true (or present).
| Number | Date | Country | Kind |
|---|---|---|---|
| 20204158.8 | Oct 2020 | EP | regional |
