This invention relates generally to the systems and methods for monitoring conditions of at least one article. In particular, the invention relates generally to the systems and methods for monitoring conditions of blades for turbines. Furthermore, the invention relates generally to the systems and methods for monitoring conditions of gas turbine blades for turbines where the system and methods can detect defects, and predict failures of gas turbine blades using sensors, such as non-contact sensors.
It is known to monitor and determine a condition of a blade, for example of blade tip deflections; using a variety of non-contact sensing technology. Further, these methods and systems may also monitor turbine blade tip vibration using estimation algorithms. In these conventional methods and systems, blade tip deflection magnitudes can be an indication of the blade cracks. The methods and systems can relate blade tip vibrations to high cycle fatigue and potential blade failure.
A single algorithm may not be robust enough by itself to address blade deflection behaviors associated with cracks. Therefore a combination of algorithms may be desired to provide algorithm output signals, or blade health features, into a diagnostic system that uses multiple inputs to build confidence and accuracy in the final estimate of blade health.
A system for monitoring a condition of an article comprises a controller; at least one sensor for detecting a characteristic of the article; a signal processor for processing signals from the at least one sensor; a feature extractor that can extract at least one of a range of article conditions from the output from the signal processor and that can evaluate at least one of a range of article conditions, the feature extractor providing feature extractor output to the controller; an operation detector receiving data of detected features of the elements being monitored, the operation detector providing output to the controller; a central system storing historical data about the condition of an article, the off-line processor providing output to the controller, wherein the controller analyzes the output from the feature extractor, the operation detector and the central system can provide a system output of the condition of the article
A method for monitoring a condition of an article comprises providing a controller; detecting a characteristic of an article; processing signals detected of the characteristic of an article; extracting at least one of a range of article conditions from the output from the processed signal and evaluating at least one of a range of article conditions, providing feature extractor output to the controller; receiving data of detected features of the elements being monitored, the operation detector providing output to the controller; storing historical data about the condition of an article; providing the historical data about the condition of an article and providing the historical data about the condition of an article output to the controller. The method further comprising outputting a condition of the article being monitored.
These and other features, aspects, and advantages of the present invention will become better understood when the following detailed description is read with reference to the accompanying drawings in which like characters represent like parts throughout the drawings, wherein:
As used herein, an element or step recited in the singular and proceeded with the word “a,” “an,” or “one” (and especially, “at least one”) should be understood as not excluding plural said elements or steps, unless such exclusion is explicitly stated. Furthermore, references to “one embodiment” (or to “other embodiments”) of the present invention are not intended to be interpreted as excluding either the existence of additional embodiments that also incorporate the recited features or of excluding other features described in conjunction with the present invention. Moreover, unless explicitly stated to the contrary, embodiments “comprising” or “having” an element or a plurality of elements having a particular property may include additional such elements not having that property.
A local data acquisition system, as embodied by the invention, can be capable of reducing the raw blade vibration data by progressively increasing compression ratios, storing more highly granular data around an anomalous change in a blade health feature, and have the capability to upload the data to a remote system for long term monitoring and diagnostics (M&D). Blade features or compressed vibration data from the local data acquisition system can be sent over standard networks to the central system.
Further processing of blade features can be run on the central system to trend key features relating to blade health, as embodied by the invention. These functions include correlation to other related turbine parameters gathered from other turbine monitoring systems and turbine controller, trending individual or combined features to look for meaningful changes with reference to pre-established defect thresholds, and generating alarms for personnel to analyze further and escalation to customers for potential inspections of the turbine. Alarming is accomplished in a variety of ways, including emails, phone calls, and text messages.
The central system, as embodied by the invention, also stores the results of field inspections of turbine blades to update false positive and false negative rates of the blade health diagnostic algorithms, allowing continuous improvement of the blade health monitoring system over time. Risk models are updated based on the field inspections, enabling fine-tuning of turbine inspection intervals and confidence values associated with the M&D system alarms.
Accordingly, with respect to
Raw data of at least one characteristic of the article from at least one sensor 102 can be processed in real-time to generate a set of blade features. The at least one sensor 102 may comprise one or more sensors, but in
A logic or signal processor 103 (hereinafter “signal processor”) then processes the signal(s) from sensor 102. The signal processor 103, as embodied by the invention, can be provided as any conventional processor. For example, and in no way limiting of the invention, the signal processor 103 may comprise any appropriate high-powered solid-state switching device. As illustrated, the signal processor 103 can be a computer. However, this is merely exemplary of an appropriate high-powered signal processor, which is within the scope of the invention. For example but not limiting of the invention, the signal processor 103 can be implemented as a single special purpose integrated circuit, such as an ASIC, having a main or central processor section for overall, system-level control, and separate sections dedicated performing various different specific combinations, functions and other processes under control of the central processor section. It will be appreciated by those skilled in the art that the signal processor 103 can also be implemented using a variety of separate dedicated or programmable integrated or other electronic circuits or devices, such as hardwired electronic or logic circuits including discrete element circuits or programmable logic devices, such as PLDs, PALs, PLAs or the like. The signal processor 103 can also be implemented using a suitably programmed general-purpose computer, such as a microprocessor or microcontrol, or other processor device, such as a CPU or MPU, either alone or in conjunction with one or more peripheral data and signal processing devices. In general, any device or similar devices on which a finite state machine capable of implementing the flow charts, can be used as the signal processor 103. A distributed processing architecture can be provided for enhanced data/signal processing capability and speed.
The signal processor 103 can process signal(s) from one or more of the sensors 102 both in time and frequency domains. Therefore, the signal processor 103, as embodied by the invention, then sends its output to a feature extractor 104. The feature extractor 104 can extract at least one of a range of article conditions, such as but not limited to, a range of blade features from the output from the signal processor 103 and can also evaluate at least one of a range of article conditions. These features from the feature extractor 104 comprise, but are not limited to, features such as static blade tip bending, blade untwist, blade radial extension, and blade tip vibratory amplitudes and frequencies.
These exemplary features from the feature extractor 104 can then be sent to a controller 106. Additionally, the controller 106 can receive output or signals from a machine operating state detector 105 that detects operating characteristics of the machine or element being monitored, such as speed, load, and other miscellaneous pressures and temperatures associated with a gas turbine. The output or signals from the state detector 105 and from the feature extractor 104 can be used for diagnostics and prognostics of detected features of the elements being monitored. The system output or signals at output 107 from the controller 106 can be used in a variety of ways, such as but are not limited to, model-free trending features over time, and model-based comparison of actual features to expected monitored signatures. The system output 107, as embodied by the invention, can provide output provided in a hierarchical output from simple to complex output, as described hereinafter.
The controller 106 can comprise any appropriate solid-state switching device. As embodied by the invention, the controller 106 can be a computer. In the illustrated embodiment, controller 106 can be implemented as a single special purpose integrated circuit, such as ASIC, having a main or central processor section for overall, system-level control, and separate sections dedicated performing various different specific combinations, functions and other processes under control of the central processor section. It will be appreciated by those skilled in the art that controller 106 can also be implemented using a variety of separate dedicated or programmable integrated or other electronic circuits or devices, such as hardwired electronic or logic circuits including discrete element circuits or programmable logic devices, such as PLDs, PALs, PLAs or the like. The controller 106 can also be implemented using a suitably programmed general-purpose computer, such as a microprocessor or microcontrol, or other processor device, such as a CPU or MPU, either alone or in conjunction with one or more peripheral data and signal processing devices. In general, any device or similar devices on which a finite state machine capable of implementing the flow charts, can be used as the controller 106. In a particular embodiment, the controller 106 can be a data acquisition system located in the vicinity of the sensors in a power plant, thereby providing a remote access system, as embodied by the invention.
The signal processor 103, as embodied by the invention, will extract and send to feature extractor 104 at least two properties of the blade tip for each blade 202, as it passes under the sensors 102. These properties or output 107 include, but are not limited to, the circumferential offset of at least one of the leading or trailing edge from a nominal position, the average of the leading and trailing edges, and radial clearance between the blade tip 201 and casing 203 (the turbine components are illustrated schematically for illustration purposes) from the sensor 102.
Using this information, several features can be computed as output 107, as embodied by the invention. For example, the features include static blade tip bending, blade untwist, blade radial extension, and blade tip vibratory amplitudes and frequencies. These features can be indications of the condition or “health” of a blade (or other monitored element), where the health can be analyzed, processed, or otherwise used to determine the condition of the blade. For example, and not intended to limit the invention in any manner, the indication in a blade may represent a bend, crack, or missing section, the indications having been caused by foreign object damage (FOD), low and high-cycle fatigue or corrosion on a blade 202.
Furthermore, other complex time, and frequency domain analyses can be conducted by the system 100. For example, through appropriate analysis, such as but not limited to, known algorithms and processing, including without limitation Fourier analysis, finite element analysis, fracture mechanics algorithms, 3-D analysis, and the like, can yield static deflection and dynamic deflection properties, such as blade tip vibratory amplitudes and frequencies.
Output 107 can be extracted and used to provide diagnostic outputs, such as a designation of an “indication” 301 for a blade 202 at progressively different, such as increasing, levels of specificity. The indication, as embodied by the invention, can comprise, but is not limited to, a blade fault, including the detuning 302 of the blade 202; blade tip deflection 303, such as but not limited to, dynamic tip deflection; blade extension 304, such as but not limited to, static blade extension; blade twist 306, such as but not limited to, static blade twist; and blade bend 307, such as but not limited to, static blade bend; or combinations thereof. As embodied by the invention, the indications or other features extracted above can be used to provide diagnostic outputs by applying diagnostic algorithms in controller 106 to provide the levels of diagnostic outputs. This feature of the system 100 is illustrated in
In
For each system output 107 (hereinafter “output”), the blade health monitoring system 100, as embodied by the invention, can provide an associated confidence value, which can be assigned to the diagnostic output 107. The confidence value is based on an aggregation approach as described hereinafter.
The diagnostic output and associated confidence value, as embodied by the invention, in the blade health monitoring system 100 can be implemented at at least two levels. A first level is a data based approach. In the data based approach, as embodied by the invention, extracted output is trended over time, and statistically significant changes are identified or flagged as possible indications of an “anomalous” blade condition. A second level for the output confidence value and associated diagnostic output value, as embodied by the invention's system 100, incorporates a model-based fault diagnosis and feature aggregation capability. This model-based fault diagnosis and feature aggregation comprises comparing a feature value with an expected value for that feature from a previously stored, predetermined, or an a priori model of a blade 202. If the feature magnitudes match within a pre-defined, predetermined margin of error, an indication can be provided. The model predictions may be used as a guideline for relative deflections that can be expected in various static and dynamic vibratory modes of a turbine, or a turbine blade 202.
Feature aggregation, as embodied by the invention, refers to a progressive accumulation of evidence from simple to complex, supported by a priori knowledge from models and laboratory tests. These can provide a confidence value associated with the diagnostic announcement. The confidence value of a diagnostic increases as the number of supporting fault related indications for a give blade increase, thereby reducing the probability of a false alarm. This methodology can allow for flexibility in the configuration and processing of sensors 102 of the blade health monitoring system 100, as embodied by the invention. As the blade health monitoring system 100 is developed with more knowledge of blade diagnostics, more features and historical trends of those feature scan be stored in the central system or logic 101 and can be generated from the same sensor data. Additionally, these features can be generated from the same sensor data and can be added into the aggregation process, to provide even more enhanced confidence in the blade health diagnosis, as embodied by the invention. At enhanced levels of the feature aggregation hierarchy as embodied by the invention, a priori knowledge from field investigations of failed blades, as well as finite element models, can be used to determine whether or not a feature value is valid. Therefore, the aggregation of the blade health monitoring system 100, as embodied by the invention, does not accept feature values that could be interpreted as outliers.
The blade health monitoring system 100, as embodied by the invention, can also have the capability to remotely monitor the health of the article or machine, as illustrated in
While the invention has been described in terms of various specific embodiments, those skilled in the art will recognize that the invention can be practiced with modification within the spirit and scope of the claims.