Method of monitoring operation of a machine and computer implemented monitoring system implementing the method

Abstract

A method for monitoring operation of a machine includes providing a function that links a control parameter to an operating parameter; operating the machine and measuring the control parameter and the operating parameter while the machine is operating; calculating the control parameter by way of the function based on the measured operating parameter; comparing the measured control parameter to the calculated control parameter and generating an alert when they do not match. The method also includes defining a fitting time window and adjusting the function based on the measured operating parameter and the measured control parameter measured through the fitting time window.

Description

[0001] Cross-references to related applications This patent application claims priority to European patent application no. 20195343.7, filed on 9 September 2020, the entire disclosure of which is incorporated herein by reference. Technical Field

[0002] The present invention relates to a method for monitoring the operation of a machine and a computer-implemented monitoring system for implementing the method.

[0003] The machine can be part of a power plant used for generating electrical power, such as a gas turbine, a steam turbine, or a generator like a turbine generator or a hydro generator. The machine can also be a power plant used for generating electrical power. In the following text, gas turbines will be referred to in particular. Background Technology

[0004] During operation, machines such as gas turbines may suffer damage that can affect their operation.

[0005] In the event of major damage such as the detachment of a complete turbine or compressor blade, the machine's safety system can stop the machine from operating to protect its integrity and allow for necessary repairs.

[0006] In the event of minor damage that does not trigger the intervention of safety systems, such as if only a portion of the compressor blades flies away or a seal is damaged, allowing the machine to continue operating more or less unaffected, the machine will continue to operate but the damage will worsen, in some cases resulting in medium- or long-term major damage requiring extended downtime for repair.

[0007] The first approach to handling these minor breaches is to monitor changes in the average value of some parameters. This approach is highly dependent on the size of the time window in which the average value of the parameters is measured, because a short window will provide many false alarms, while a long window will in many cases not issue any alarms; this method is therefore unreliable.

[0008] A second approach to handling minor damage relies on artificial intelligence and machine learning, and involves monitoring parameters to identify patterns that indicate a problem has occurred based on a database of known machine failures. This method is computationally intensive and requires a large amount of reliable data on machine malfunctions. Summary of the Invention

[0009] Aspects of the invention include providing a method for monitoring the operation of a machine and a computer-implemented monitoring system for implementing the method, which is reliable and requires limited computing resources and data.

[0010] These and other aspects are achieved by providing a monitoring system implemented by means of the method according to the appended claims and a computer implementation. Attached Figure Description

[0011] Further features and advantages will become more apparent from the description of preferred, but not exclusive, embodiments of the method and system, which are illustrated in the accompanying drawings by way of non-limiting example, wherein: Figure 1 Showing a gas turbine; Figure 2 The timeline with the fitting time window is shown. Detailed Implementation

[0012] system Referring to the accompanying drawings, which illustrate machines such as gas turbine 1; even though gas turbines are specifically described below, this method can be used to monitor other types of machines, such as generators (i.e., turbine generators or hydro generators) or steam turbines or other machines; the machines can also be complex arrangements of multiple machines, such as power plants for generating electrical power.

[0013] The gas turbine 1 has a compressor 2, a combustion chamber 3 and a turbine 4; different arrangements are possible as are known, and for example, the gas turbine may have a second combustion chamber downstream of the first combustion chamber, wherein a high-pressure turbine and / or an air diluter are located between the two combustion chambers.

[0014] The compressor inlet is typically equipped with VIGV 8 (Variable Inlet Vane), which is an adjustable guide vane that regulates the amount of air allowed to enter the compressor, and the compressor and turbine have rotor blades and guide vanes.

[0015] Gas turbine 1 typically has a control system and a safety system; these two systems can be implemented by separate or identical circuit systems.

[0016] The control system receives data inputs (such as ambient pressure, temperature, humidity, required power, etc.) and supplies control data for driving the gas turbine (such as VIGV position, fuel mass flow supply, etc.).

[0017] The safety system also receives input data, which may be the same as or different from the data received by the control system, and trips the gas turbine if these input data show irregular and abnormal values ​​indicating that damage has occurred in the gas turbine.

[0018] During operation, gas turbines can suffer minor damage that does not trigger safety systems, and therefore, after such damage occurs, gas turbine operation continues essentially in the same manner as before. For example, such minor damage may include the detachment of turbine blades or a portion of the compressor, or damage to seals in the combustion chamber. Even if such damage does not cause gas turbine failure, it initiates a degradation process that typically leads to gas turbine failure over time. Figure 1 Reference 6 indicates minor damage to compressor rotor blade 5 caused by a blade that flew off.

[0019] The gas turbine is advantageously equipped with a computer-implemented monitoring system capable of detecting minor damage that, while not causing the gas turbine to trip, still results in irregular operation. Therefore, in addition to the control and safety systems, a monitoring system is provided and can be implemented using the same or different circuitry as the control and / or safety systems.

[0020] In particular, computer-based monitoring systems can detect small irregularities in the operation of gas turbines, while safety systems can detect large irregularities (such as when an intact blade flies off, thereby damaging the downstream compressor or turbine stage and / or components in the casing or combustion chamber due to backfire).

[0021] The computer-implemented monitoring system includes: a first sensor 10 for measuring one or more control parameters; and a second sensor 11 for measuring one or more operating parameters of the gas turbine.

[0022] Depending on the needs, control parameters and operating parameters can be different. Typically, control parameters indicate the operating conditions of the gas turbine, while operating parameters indicate the boundary conditions that determine the operation of the gas turbine.

[0023] Control parameters include, for example, the pressure at the outlet of compressor 2 and / or the pressure at the outlet of combustion chamber 3 and / or the temperature at the outlet of compressor 2 and / or the temperature at the outlet of combustion chamber 3 and / or the flame temperature and / or the gas turbine output power, etc.

[0024] Operating parameters include, for example, VIGV position (VIGV position indicates the mass airflow through the gas turbine) and / or ambient humidity and / or ambient temperature and / or gas turbine output power, etc.

[0025] Naturally, for the same machine, the control parameters and operating parameters are not the same; however, for different machines, they can be selected as needed. For example, output power can be selected as a control parameter or an operating parameter.

[0026] The computer-implemented monitoring system also includes an elaborating unit 15, which is connected to a memory 12. The elaborating unit is configured to run program code for performing the methods of the present invention, and the memory 12 is configured to store data measured by the first sensor 10 and the second sensor 11 and the results calculated by the elaborating unit 15.

[0027] In a preferred embodiment, a computer or other electronic device defines the interpreter unit 15 and the memory 12.

[0028] The interpretation unit 15 is connected in sequence to the output device 16, which is arranged to make the interpretation results available. The output device can be any type of known machine-to-human interface, such as a monitor, printer, etc.

[0029] method Methods for monitoring machine operation include providing one or more functions that link one or more control parameters to one or more operating parameters.

[0030] In the first example, the function is at least a polynomial function, and the control parameters and operation parameters are linked via one or more coefficients. In this case, the function can generally be defined as: in: -CP is a general control parameter. -OP1, OP2, and OP3 are operation parameters. -a, b, c, d, e, f… are coefficients that link the operating parameters to the control parameters.

[0031] Typically, for a general number of m control parameters and n operation parameters, the function can be expressed in matrix form as follows: in: -a1, 1-am, n are coefficients that link the operating parameters OP1-OPn to the control parameters CP1-CPm. -OP1-OPn are operation parameters. -CP1-CPm are control parameters.

[0032] The coefficients can be defined by operating the machine, measuring control parameters and operating parameters, and by selecting appropriate coefficients, so that the operating parameters provide control parameters in the form of functions.

[0033] A specific example of a polynomial function could be the following: in: - The operating parameters are: VIGV indicating the VIGV location (i.e., the opening); ET indicating the ambient temperature; - The control parameters are: PCO, which indicates the air pressure at the compressor outlet; TCO, which indicates the air temperature at the compressor outlet; TOCC, which indicates the hot gas temperature at the combustion chamber outlet; and POT, which indicates the hot gas pressure at the turbine outlet.

[0034] Polynomial functions can advantageously represent any machine or combination of machines.

[0035] Alternatively, the function can be defined via simulation software for a machine such as a gas turbine. As is known, the operation of a gas turbine (or a typical combination of machines or a machine) can be simulated (online or offline) by dedicated software that provides outputs based on input signals in the form of gas turbine state parameters, compressed air, and / or hot gas flowing through the gas turbine.

[0036] The monitoring method of the present invention can utilize such software and can provide operating parameters as input signals to analog software to obtain control parameters as output signals.

[0037] Simulation software can advantageously reproduce machine behavior across the entire spectrum of operation, and can also follow potentially “normal” abrupt changes in control parameters (i.e., those that correspond to normal machine operation and are expected to be such changes), which might not be possible using other functions.

[0038] When the function is defined via simulation software (preferably via adjustment coefficients), the function can be written as: in -Tf1…Tfn: These are adjustment coefficients that account for the possibility of mismatch between the data calculated by the model and the actual data; ideally, these coefficients are equal to 1 and therefore the model perfectly reproduces the behavior of the gas turbine. - f (OP1…OPn): Represents the model, i.e., the function f ( ) (as a model) receives operation parameters OP1…OPn as input and provides control parameters CP1…CPm as output; -OP1…OPn: These are the operation parameters provided to the model as input; -CP1…CPm are control parameters.

[0039] In this case, as in the case of polynomial functions, the coefficients can be obtained by operating the machine, measuring the operating parameters and control parameters, and providing the operating parameters as part of the model.f The input of () is limited by calculating control parameters and selecting appropriate coefficients, so that the model... f The calculated control parameters match (i.e., are equal to) the measured control parameters.

[0040] Using models has several advantages over using polynomial functions, such as providing model validation because if the adjustment coefficients are constants equal to 1, the model perfectly reproduces the behavior of the gas turbine; in addition, it is generally simpler to manage the adjustment coefficients of the model-based approach than the polynomial-based approach because the adjustment coefficients are ideally 1 (i.e., they are known in advance under ideal conditions).

[0041] Once the function is defined, the gas turbine can be operated, and control and operating parameters can be measured simultaneously with the gas turbine's operation, depending on the monitoring method.

[0042] Based on the measured operating parameters, the control parameters are calculated using functions.

[0043] The control parameters calculated in this way are the control parameters expected under the condition of “regular” operation of the gas turbine (i.e., under the condition that the gas turbine operation has no impact on the operation of minor damage), the amount of which is too small to be detected by the safety system, but may have a huge impact in long-term operation.

[0044] To monitor minor damage that may affect the machine, even if the minor damage does not trigger the safety system, the measured control parameters are compared with the calculated control parameters (i.e., the control parameters calculated using a function), and an alarm is generated when the measured control parameters and the calculated control parameters do not match.

[0045] The matching of the measured control parameter and the calculated control parameter is determined when (i) the difference between the measured control parameter and the calculated control parameter is below a given threshold and / or (ii) the curves of the measured control parameter and the calculated control parameter do not intersect each other within a given time interval.

[0046] In different examples of the method, any one or two criteria can be used to determine whether the measured control parameter matches the calculated control parameter. For example, criterion (i) can be used to ensure that the measured control parameter does not differ too much from the calculated control parameter; criterion (ii) can be used to check that the measured control parameter follows, i.e., reproduces the trajectory of the calculated control parameter. Even with a large margin in each of criteria (i) and (ii), criteria (i) and (ii) can be used together to check for correct operation to avoid unexpected false alarms, while simultaneously ensuring the detection of faults.

[0047] The fitting time window can be defined, and the coefficients linking the control parameters to the operating parameters can be adjusted based on the operating parameters and control parameters that have been measured through the fitting time window, so that the coefficients can always correctly reproduce the control parameters based on the operating parameters.

[0048] Advantageously, the fitting time window can be shifted over time, allowing the function to be adjusted continuously or periodically, thus taking into account losses due to normal aging and avoiding the generation of inappropriate alarms.

[0049] In different embodiments, the fitting time window can be shifted forward continuously or periodically over time; the method can then be implemented in real time to perform real-time monitoring of the gas turbine.

[0050] The method may also include ( Figure 2 Define an operating time window OTW that is shifted in time after the fitting time window FTW; and use a function that has been adjusted in conjunction with the given fitting time window FTW throughout the entire operating time window OTW.

[0051] For example, the fitting time window FTW can be one week long and can be periodically shifted forward one week each week, such that the fitting time window FTW occurs in the week preceding the operating time window, which can also be one week long.

[0052] In different examples, the fitting time window FTW can be continuously shifted forward; in this example, the fitting time window can have a length of one week, starting 168 hours before the actual implementation time of the method.

[0053] Additionally, a black time window (BTW) can be provided between the operating time window (OTW) and the fitting time window (FTW). During this black time window (BTW), the control parameters and operating parameters are not sampled, or if they are sampled, they are not used. This can, for example, separate the fitting time window from the operating time window, thus preventing the variation in the measurements of the control and operating parameters from being confused with measurement variations due to machine aging or signal noise.

[0054] Additionally, longer or shorter fitting time windows can be defined. Therefore, longer and shorter control parameters can be calculated based on the longer or shorter fitting time windows as indicated above.

[0055] In this scenario, a longer fitting time window is removed from the actual operating time, while a shorter fitting time window is closer to the actual operating time.

[0056] As an example, a longer fitting time window can be anticipated a week before the actual operating time; alternatively, a longer fitting time window can shift over time according to the actual operating time or remain constant while the gas turbine is operating through a longer operating time window; in this case, the start or end of the longer fitting time window can be appropriately a week before the start of the longer operating time window.

[0057] Similarly, a shorter fitting time window can be predicted one hour before the actual operating time; in addition, a shorter fitting time window can shift over time according to the actual operating time or remain constant while the gas turbine is operating through a shorter operating time window; in this case, the start and end of the shorter fitting time window can be timely one hour before the start of the shorter operating time window.

[0058] The use of a longer fitting time window that is closer to the actual operating time enables the model to accurately predict gas turbine operation (and thus calculate control parameters based on input operating parameters) under different operating conditions (e.g., partial load, full load, etc.).

[0059] The use of a shorter fitting time window allows the actual operation of the machine to be monitored using the most recently fitted model (e.g., partial load, full load, or other models). Therefore, by utilizing a shorter fitting time window, even in the case of small signal steps, sudden changes in gas turbine operation can be monitored due to a very good fitting model, as it adjusts within a short time window and is close to the actual operating time, thus accurately reproducing the operation of the gas turbine.

[0060] The use of both longer and shorter fitting time windows allows variations caused by gas turbine operation (e.g., from full load to partial load) to occur, which can cause deviations from the model with the short fitting time window to be identified as normal operation from the model with the long fitting time window. Therefore, false alarms can be avoided.

[0061] The measured control parameter is therefore compared with the calculated longer and shorter control parameters, and an alarm is generated when (i) at least one of the calculated longer and shorter control parameter sets does not match the measured control parameter, or (ii) neither of the calculated longer and shorter control parameter sets matches the measured control parameter.

[0062] The method is advantageously implemented based on recorded data (i.e., data collected in the past) or data continuously measured on the operating gas turbine (a real-time implementation of the method).

[0063] The method of the present invention is particularly advantageous because it provides reliable results, yet is not computationally necessary.

[0064] In addition, the method requires a small amount of data to be implemented, and advantageously, a small amount of data refers to the correct operation of the machine, not its failure. Therefore, collecting such data is simple and fast, even in the case of a small machine fleet or a single machine.

[0065] Naturally, the described features can be provided independently of each other. In fact, the materials and dimensions used can be arbitrarily selected according to needs and existing technology.

Claims

1. A method for monitoring the operation of a machine, comprising: a. Provide at least one function that links at least one control parameter to at least one operation parameter; b. Operate the machine, and while the machine is operating, measure the at least one control parameter and the at least one operating parameter; c. Calculate the at least one control parameter using the at least one function based on at least one measured operating parameter; d. Compare at least one measured control parameter with at least one calculated control parameter, and generate an alarm when they do not match; e. Define the fitting time window; f. Adjust the at least one function based on the at least one measured operating parameter and the at least one measured control parameter measured through the fitting time window; g. Define an operation time window, which is shifted in time after the fitting time window; as well as h. Use one or more functions adjusted based on a given fitting time window throughout the entire operation time window; i. Define longer and shorter fitting time windows in time; j. Calculate at least one longer and at least one shorter control parameter based on the longer and shorter fitting time windows; k. Compare the at least one measured control parameter with at least one calculated longer and shorter control parameter; l. An alarm is generated when at least one of the calculated longer and shorter control parameters does not match the at least one measured control parameter. The longer fitting time window and the shorter fitting time window shift over time.

2. The method according to claim 1, characterized in that... Repeat steps b through f.

3. The method according to any one of the preceding claims, characterized in that... A match is determined when at least one measured control parameter is matched with at least one calculated control parameter under the following conditions: - The difference between the at least one measured control parameter and the at least one calculated control parameter is below a given threshold; and / or - Within a given time interval, the trajectories of at least one measured control parameter and at least one calculated control parameter do not intersect each other.

4. The method according to claim 1, characterized in that... Between the operation time window and the fitting time window, a black time window is provided, during which at least one control parameter and at least one operation parameter are not sampled or are not used if they are sampled.

5. The method according to claim 1, characterized in that... The alarm is generated when both the calculated longer and shorter control parameters do not match the at least one measured control parameter.

6. The method according to claim 1, characterized in that... The method is implemented during the real-time operation of the machine.

7. The method according to claim 1, characterized in that... The at least one function that links the at least one control parameter to the at least one operating parameter is a polynomial function.

8. The method according to claim 1, characterized in that... The at least one function is defined via the machine's simulation software.

9. The method according to claim 1, characterized in that... The minimum function is defined by an adjustment coefficient.

10. The method according to claim 1, characterized in that... The machine is a gas turbine, generator, steam turbine, or power plant used for generating electrical power.

11. A computer program having program code that, when executed, performs the method according to claim 1.

12. A computer-implemented system for monitoring the operation of a machine (1), comprising: - At least a first sensor (10) is used to measure at least the control parameters of the machine (1); - At least a second sensor (11) is used to measure at least the operating parameters of the machine; - Memory (12), the memory (12) is arranged to store data measured by the first sensor (10) and the second sensor (11); - An interpreter unit (15), which is connected to the memory (12) and arranged to receive data from the memory (12), the interpreter unit having code thereon for implementing the method of claim 1.

13. The computer implementation system according to claim 12, characterized in that... An electronic device defines the interpreting unit (15) and the memory (12).

14. The computer-implemented system according to claim 12, characterized in that... The machine is a gas turbine, generator, steam turbine, or power plant used for generating electrical power.

Images