The present invention relates generally to machine condition monitoring and more particularly to determining pattern rules for use in machine condition monitoring.
Machine condition monitoring (MCM) is the process of monitoring one or more parameters of machinery, such that a significant change in the machine parameter(s) is indicative of a current or developing condition (e.g., failure, fault, etc.). Such machinery includes rotating and stationary machines, such as turbines, boilers, heat exchangers, etc. Machine parameters of monitored machines may be vibrations, temperatures, friction, electrical usage, power consumption, sound, etc., which may be monitored by appropriate sensors. The output of the sensors may be in the form of and/or be aggregated into a sensor signal or a similar signal.
Generally, a condition is a comparison of the machine parameter to a threshold. For example, a machine parameter value may be compared with an equality and/or inequality operator, such as <, =, >, ≠, ≡, ≦, ≧, etc., to a threshold value. Therefore, a condition signal is a signal based on the machine parameter values (e.g., a plurality of machine parameter values grouped as a discrete or continuous signal) and a condition signal pattern is a portion (e.g., sub-set) of the condition signal.
Machine condition monitoring systems generally use a number of rules, referred to as a rule base, to define the machine parameters to be monitored and critical information (e.g., indicative of a condition change) about those machine parameters. In some cases, hundreds of sensors monitor and/or record these machine parameters. The output of the sensors (e.g., sensor signal, sensor estimate, sensor residue, etc.) may then be used as the input to one or more rules. Rules must be correctly and intelligently designed to properly detect faults, but minimize improper indicators of faults (e.g., false alarms).
In general, simple rules are constructed as indicative conditional logical operations (e.g., if-then statements). The input of a rule, the “if”, is a condition as described above (e.g., if machine parameter A>threshold B) and the output of the rule, the “then”, is a fault (e.g., then fault type 1). Conditions may be composite by concatenating multiple conditions (e.g., with AND, OR, etc.) to create one input. Rule bases may be improved using a persistence measure, which is a duration of the condition. Persistence measure-based rules use information in a time range in contrast to the single time of simple rules and/or individual times of concatenated simple rules. Persistence measure-based rules may provide greater utility than simple rules and/or concatenated simple rules, but are limited in that they check the same condition at each time within the time range.
Many prior rule bases rely on human experts to manually create and maintain large amounts of rules. Manual rule creation is a time consuming process that requires human estimation of complex signal patterns. Further, some signal patterns indicative of faults are highly complex and cannot be captured with the rules described above. Accurately describing complex symptoms of faults is extremely complicated and, in many cases, intractable for a human using conventional methods of creating rules.
Therefore, alternative methods and apparatus are required to create rules in machine condition monitoring.
The present invention provides methods of machine condition monitoring and fault detection by creating pattern rules. Pattern rules are created by comparing a condition signal pattern to a plurality of known signal patterns and determining a machine condition pattern rule based at least in part on the comparison of the condition signal pattern to one of the plurality of known signal patterns. A matching score based on the comparison of the condition signal pattern to one of the plurality of known signal patterns as well as a signal pattern duration is determined. The machine condition pattern rule is then defined for nonparametric condition signal patterns as a multipartite threshold rule with a first threshold based on the determined matching score and a second threshold based on the determined signal duration. For parametric signal patterns, one or more parameters of the signal pattern are determined and the machine condition pattern rule is further defined with a third threshold based on the determined one or more parameters.
These and other advantages of the invention will be apparent to those of ordinary skill in the art by reference to the following detailed description and the accompanying drawings.
The present invention generally provides methods and apparatus for machine condition monitoring using pattern rules.
Machines 102 may be any devices or systems that have one or more monitorable machine parameters, which may be monitored by sensors 104. Exemplary machines 102 include rotating and stationary machines, such as turbines, boilers, heat exchangers, etc.
Sensors 104 are any devices which measure quantities and convert the quantities into signals which can be read by an observer and/or by an instrument as is known. Sensors 104 may measure machine parameters of machines 102 such as vibrations, temperatures, friction, electrical usage, power consumption, sound, etc. The output of sensors 104 may be in the form of and/or aggregated into a condition signal as depicted in
In some embodiments, pattern detection module 106 and/or pattern rule module 108 may be implemented on and/or in conjunction with one or more computers, such as computer 600 described below with respect to
All signal patterns have a parameter T, which is the duration of the pattern. Signal patterns are categorized as parametric signal patterns or nonparametric signal patterns. Parametric signal patterns have a predefined shape that can be described by a set of parameters. Exemplary parametric signal patterns are shown in
Though not depicted, any appropriate parametric patterns may be used. Such parametric patterns include higher-order polynomial patterns (e.g., y=mx2+dx+b, etc.), exponential patterns, cosine patterns, etc. Generally, in signal patterns 202 and 302 as well as signal patterns with other parameters, the parameter sets may be referred to as signal parameters S.
In step 504, known signal patterns are stored at pattern detection module 106. Known signal patterns include parametric signal patterns, such as signal pattern 202 and signal pattern 302, as well as nonparametric signal patterns, such as nonparametric signal pattern 402. Any appropriate parametric signal patterns may be stored. Nonparametric signal patterns indicative of fault or other significant conditions may also be stored at pattern detection module 106. In some embodiments, such nonparametric signal patterns are automatically detected and stored. In alternative embodiments, nonparametric signal patterns are identified by a user and entered into (e.g., selected by or otherwise denoted) pattern detection module 106.
Parametric signal patterns may be stored by storing their relevant signal parameters S. Nonparametric signal patterns may be stored using time and/or frequency templates. Such signal patterns may be represented by ZT=[z1, z2, . . . , zT], where zi is the signal value at the i-th data point ant T is the signal pattern duration as described above with respect to
In step 506, a condition signal pattern is received. Herein, a condition signal pattern is a signal pattern for which a pattern rule is to be determined. In at least one embodiment, the condition signal pattern is received at the pattern detection module 106. In the same or alternative embodiments, the condition signal pattern is a signal pattern received from sensors 104 that is indicative of a fault condition. Accordingly, the condition signal pattern may be a parametric or nonparametric signal pattern. In some embodiments, a user may designate the received condition signal pattern as a known fault and may submit the condition signal pattern to pattern detection module 106. The condition signal pattern may be represented by XT=[xt−T+1, xt−T+2, . . . , xt] where xt is the value of the signal (e.g., signal 200, 300, 400, etc.) at a time t.
In step 508, the condition signal pattern is compared to known signal patterns stored in step 504. The condition signal pattern may be compared to one or more parametric signal patterns and nonparametric signal patterns. Additionally, the duration T of the condition signal pattern and/or the known signal pattern may be stretched and/or compressed to match each other to facilitate the comparison.
Any appropriate comparison measure may be used and a matching score G may be determined. An individual matching score G may be determined for each comparison of the condition signal pattern to a known signal pattern. Matching scores G are the best values obtained using all available comparison measures. That is, the comparison measures are optimized to present the best possible fit of the condition signal pattern to the known signal patterns. In some embodiments, a user may select a comparison measure. For example, an average Euclidean distance of the condition signal pattern to the known signal pattern may be used. Such a distance may be calculated as
Alternatively, an average correlation measure may be used as
The matching score G is thus an indication of a correlation, or match, based on the comparison measure. In embodiments where an average Euclidean distance or other similar distance measure is employed, the optimal match is the minimum matching score G. In embodiments where an average correlation measure or other similar measure is employed, the optimal match is the maximum matching score G.
In step 510, a signal pattern duration is determined. The comparison measures of step 508 are normalized by T such that they are insensitive to the variable durations of T, as discussed above. At each time, XT is compared with ZT using an appropriate comparison measure (e.g., a Euclidean distance, a correlation, etc.). The duration T of the known signal pattern may be varied to coincide with the optimal (e.g., maximum or minimum, as appropriate) comparison measure. That is, the duration T is varied to allow the comparison of the condition signal pattern to each known signal pattern to achieve the most optimal correlation. By keeping the duration T of the incoming condition signal pattern intact while varying only the known signal pattern duration T, a fast Fourier transform or another appropriate transform may be employed to scan the whole incoming condition signal pattern in a very short time. For nonparametric signal patterns when the duration T is not the same as the original T, downsampling (e.g., reducing the sampling rate of the signal), interpolation, and/or other appropriate methods may be used to “find” signal values at non-existing data points.
In step 512, the optimal known signal pattern is selected. Based on the matching score determined in step 508 and the signal pattern duration determined in step 510, the condition signal pattern is compared to each of a plurality of known signal patterns and the known signal pattern that most closely matches (as evidenced by matching score G and/or signal pattern duration T) may be selected.
In step 514, a determination is made as to whether the known signal pattern is a parametric (P) or nonparametric (NP) signal pattern. If the known signal pattern is a parametric signal pattern, the method passes to step 516 and an optimal parameter set S is determined. In some embodiments, a standard least square method may be used to find an optimal matching score G. In alternative embodiments, a gradient-based optimization method may be used to search for an optimal matching score G. Of course, any appropriate method of finding an optimal matching score G may be used. The parameter set S corresponding to the solution with the optimal matching score S may be considered as the optimal parameter set S.
If the known signal pattern is a nonparametric signal, the method passes to step 518 and a machine condition pattern rule is determined by pattern rule module 108. The machine condition pattern rule is determined, in step 518, using the signal pattern duration T and the matching score G. The machine condition pattern rule is thus a multipartite threshold rule with a first threshold based on the determined matching score and the second threshold based on the determined signal pattern duration. The pattern rule is defined as a multi-input conditional logical rule with a duration threshold as one input and a matching score threshold as another input. For example, using a Euclidean distance measure as described above, a pattern rule may be defined as “If signal duration T>threshold A AND matching score G<threshold B, then fault type 1 occurs.”
After the optimal parameter set S is determined in step 516, a machine condition pattern rule is determined in step 520 by pattern rule module 108. The machine condition pattern rule is determined, in step 520, using the signal pattern duration T, the matching score G, and the parameter set S. The machine condition pattern rule is thus a multipartite threshold rule with a first threshold based on the determined matching score, a second threshold based on the determined signal pattern duration, and a third threshold based on the one or more parameters of parameter set S. The pattern rule is defined as a multi-input conditional logical rule with a duration threshold as one input, a matching score threshold as another input, and a parameter set as still another input. For example, using a correlation measure as described above, a pattern rule may be defined as “If signal duration T>threshold A AND matching score G<threshold B AND slope>m, then fault type 2 occurs.”
Method steps 506-520 may be repeated as appropriate to determine multiple pattern rules. That is, following determination of pattern rules in steps 518 and/or 520, the method 500 may return control to step 506. These pattern rules may be stored after steps 518 and/or 520 in a rule base (not shown) in step 522.
In step 524, a machine condition signal is received from sensors 104 at pattern detection module 106 or another pattern rules processing location. The machine condition signal may comprise a machine condition signal pattern as described above with respect to
In step 526, a duration of the machine condition signal pattern is determined and the received machine condition signal pattern is compared to at least one known signal pattern. Such a duration determination may be based on a user input and/or may be based at least in part on the signal values. That is, the duration may be determined based on the changes to the signal values that indicate machine condition changes. The received machine condition signal pattern is compared to at least one known signal pattern. Such a comparison may similar to the comparison of step 508 described above and may include a determination of a matching score G.
In step 528, a determination is made as to whether the received machine condition signal pattern is a parametric or nonparametric signal pattern. If the received machine condition signal pattern is a parametric signal pattern, the method passes to step 530 and a parameter set S of the received machine condition signal pattern is determined. If the received machine condition signal pattern is a nonparametric signal pattern, the method passes to step 532.
Based on the determination of the duration of the machine condition signal pattern and the matching score G, nonparametric rules in the rule base are used to detect a fault condition in step 532. Similarly, based on the determination of the duration of the machine condition signal pattern and the matching score G in step 526 and the parameter set S in step 530, parametric rules in the rule base are used to detect a fault condition in step 534. In steps 532 and 534, the signal pattern duration T, matching score, and, in the case or parametric signal patterns, the parameter set S, are input to the pattern rules stored in method step 522 to detect a fault condition. In this way, a fault condition is detected if the machine condition signal pattern satisfies one or more properties of the determined machine condition pattern rule.
The method ends at step 536.
Computer 600 contains a processor 602 that controls the overall operation of the computer 600 by executing computer program instructions, which define such operation. The computer program instructions may be stored in a storage device 604 (e.g., magnetic disk, database, etc.) and loaded into memory 606 when execution of the computer program instructions is desired. Thus, applications for performing the herein-described method steps, such as pattern rule creation, fault detection, and machine condition monitoring, in method 500 are defined by the computer program instructions stored in the memory 606 and/or storage 604 and controlled by the processor 602 executing the computer program instructions. The computer 600 may also include one or more network interfaces 608 for communicating with other devices via a network. The computer 600 also includes input/output devices 610 (e.g., display, keyboard, mouse, speakers, buttons, etc.) that enable user interaction with the computer 600. Computer 600 and/or processor 602 may include one or more central processing units, read only memory (ROM) devices and/or random access memory (RAM) devices. One skilled in the art will recognize that an implementation of an actual controller could contain other components as well, and that the controller of
According to some embodiments of the present invention, instructions of a program (e.g., controller software) may be read into memory 606, such as from a ROM device to a RAM device or from a LAN adapter to a RAM device. Execution of sequences of the instructions in the program may cause the computer 600 to perform one or more of the method steps described herein, such as those described above with respect to method 500. In alternative embodiments, hard-wired circuitry or integrated circuits may be used in place of, or in combination with, software instructions for implementation of the processes of the present invention. Thus, embodiments of the present invention are not limited to any specific combination of hardware, firmware, and/or software. The memory 606 may store the software for the computer 600, which may be adapted to execute the software program and thereby operate in accordance with the present invention and particularly in accordance with the methods described in detail above. However, it would be understood by one of ordinary skill in the art that the invention as described herein could be implemented in many different ways using a wide range of programming techniques as well as general purpose hardware sub-systems or dedicated controllers.
Such programs may be stored in a compressed, uncompiled, and/or encrypted format. The programs furthermore may include program elements that may be generally useful, such as an operating system, a database management system, and device drivers for allowing the controller to interface with computer peripheral devices, and other equipment/components. Appropriate general purpose program elements are known to those skilled in the art, and need not be described in detail herein.
The foregoing Detailed Description is to be understood as being in every respect illustrative and exemplary, but not restrictive, and the scope of the invention disclosed herein is not to be determined from the Detailed Description, but rather from the claims as interpreted according to the full breadth permitted by the patent laws. It is to be understood that the embodiments shown and described herein are only illustrative of the principles of the present invention and that various modifications may be implemented by those skilled in the art without departing from the scope and spirit of the invention. Those skilled in the art could implement various other feature combinations without departing from the scope and spirit of the invention.
This application claims the benefit of U.S. Provisional Application No. 60/911,577 filed Apr. 13, 2007, which is incorporated herein by reference.
Number | Date | Country | |
---|---|---|---|
60911577 | Apr 2007 | US |