Functional actuator-sensor path optimization in structural health monitoring system

This application claims the benefit of U.S. Provisional Patent Application Ser. No. 60/912,112, filed Apr. 16, 2007, the entire disclosure of which is incorporated herein by reference.

This invention relates generally to structural health monitoring (SHM) systems. More specifically, this invention relates to a methodology for selecting and optimizing functional actuator-sensor paths in such systems.

When a large area of a composite structure is monitored by a SHM system, a network of a large number of transducers may be used for that purpose. The transducers may be acoustic wave emitters (“actuators”) and/or detectors (“sensors”). These transducers may form combinations of actuator-sensor paths that are many times more numerous than the total number of transducers in the array. It is very common for a transducer network to produce hundreds to thousands of signals in each scan of the plurality of possible paths.

Because of the differences that may occur among the path lengths and the paths due to the surrounding geometry and the possible diversity of sensor installation, the signals of different paths typically vary in quality. Signals of some paths may be too weak to provide reliable information about a region of the structure through which the acoustic wave signal passes. The signal energy and the optimal frequencies along different paths are typically also very different.

To ensure the quality of the signals and the accuracy of the subsequent signal processing for damage detection, it is desirable to pre-process the network signals and optimize the operational excitation-detection conditions. However, in systems with a large number of such signals, it may be impractical to perform this pre-processing manually. An automatic pre-processing method is therefore needed.

In accordance with the present disclosure, a method is provided for automatically pre-processing signals obtained in a network of transducers of a SHM system.

In one embodiment, the pre-processing method processes detected acoustic wave signals received at a plurality of sensors, wherein the signals are characterized at least by amplitude and frequency. The method includes an auto-gain process, in which a transmit and a receive gain amplification is determined for each path for the frequency of the signal, and a frequency selection process for selecting the frequency providing the most sensitivity (e.g., amplitude and dynamic range).

The novel method further includes removing from the network those paths for transmitting and receiving signals in which the signals are disposed below a selected set of parameter threshold values, including signal strength and/or signal-to-noise ratio (SNR).

A better understanding of the above and many other features and advantages of the novel method of the present disclosure may be obtained from a consideration of the detailed description of some example embodiments thereof below, particularly if such consideration is made in conjunction with the several views of the appended drawings, wherein like elements are referred to by like reference numerals throughout.

FIG. 1 is a process flow diagram of an example embodiment of a method for optimizing actuator-sensor path signals in a SHM system in accordance with the present disclosure;

FIG. 2 is a functional block diagram of an example embodiment of a sensor array data acquisition and processing system in accordance with the disclosure;

FIG. 3 is a process flow diagram of an example embodiment of an adaptive method for determining optimal transmit and receive conditions in a SHM system in accordance with the disclosure; and,

FIG. 4 is a process flow diagram of an example embodiment of a method for frequency selection for an array of actuator-sensor paths in a SHM system in accordance with the disclosure.

FIG. 1 is a process flow diagram of an example embodiment of a method 100 for optimizing actuator-sensor path signals of a SHM system, which may be executed the first time a sensor network is used, or periodically thereafter when the system is calibrated. This method includes, at step 10 , 1) specifying a sensor network, including the total number and 2) the respective location coordinates of all of the transducers applied to a structure that is to be inspected and monitored by the system. Initial step 10 also includes specifying 3) candidate actuation signal frequencies for each actuator-sensor path of the network and 4) a maximum allowable signal propagation distance for transmission of acoustic wave signals in or on the surface of the structure.

At step 20 of the example method 100 , physical interrogation of the network of sensors is begun to select an initial set of paths between sensors on the structure. The initial selection of paths is made on the basis of the maximum signal propagation distance provided at step 10 . That is, only those signal paths between respective actuator-sensor pairs of the array having a length less than or equal to the maximum allowable signal propagation distance are selected for optimization.

At step 30 of the exemplary method 100 , a gain determination analysis is then performed for each selected path, including the removal of electromagnetic interference (EMI) cross-talk from the signals, and the selecting of an optimum amplification gain (both in transmit and receive). Different paths may have the same or different amplification gains.

At step 40 , a determination is made of which of the paths of the initial set are “functional,” i.e., which of the paths have signals of sufficient amplitude and SNR to be useful for monitoring of the structure. Criteria for selecting functional paths include specifying a minimum threshold for signal strength and SNR. The functional path selecting step 40 then removes all other paths, i.e., those deemed to be non-functional, from the set of selected paths.

After the functional paths of the system are determined, at step 50 , a frequency determination analysis is performed for each functional path to determine at what frequency the actuator-sensor path monitoring is to be effected. This analysis includes selecting the most sensitive frequency response for each path. Different paths may have the same or different frequencies.

At step 60 of the method 100 , an output to a configuration file is provided, including 1) the specification of the functional paths, 2) the corresponding optimal gains (i.e., transmit and receive) of the paths, and 3) the optimal operating frequencies for each path.

As those of skill in the art will appreciate, in each of the gain determination step 30 , the functional path selection step 40 and the optimal frequency determination step 50 of the example method 100 of FIG. 1 , an evaluation of the signal quality (Q) of the detected signal is strongly indicated. One or more signal features may require measuring to describe the signal quality. Such measuring may include, in gain analyzing, a measured maximum signal amplitude after cross-talk is removed; a measured total energy in the whole signal of the acoustic pulse corresponding to the “first arrival” signal (i.e., the signal corresponding to a direct path transmission between the actuator and the sensor for a selected path); a measured or estimated SNR (whether an amplitude, energy or power ratio); and, a measure of “robustness” to environmental change, wherein, for example, such robustness may be defined as R=1/variance of the measured signal energy while an environmental effect, such as temperature, changes within a specified range. For example, if signal energy measured on a selected path remains within a specified range, the variance of the energy measured will be below a selected maximum threshold value, and the robustness R will exceed a minimum threshold value, indicating that the signal quality Q is satisfactory over the range of environmental variation. The signal energy may also be required, either individually or in correspondence with satisfying the robustness threshold R, to satisfy a specified SNR threshold.

The following discussion of the gain determination analysis step 30 of the example method 100 of FIG. 1 is made with reference to FIG. 2 , which is a functional block diagram of an example sensor array data acquisition and processing system 200 . The sensors of the array may comprise, for example, acoustic wave transducers, such as piezoelectric lead-zirconate-titanate (PZT) ceramic transducers. The acquisition and processing system 200 of FIG. 2 may include a signal receiver filter 205 and receiver gain block 210 . The receiver filter 205 and receiver gain 210 provide analog filtering and variable gain to received signals before the signals are digitized. The range of variable gain supplied by the gain block may be, for example, 0 dB to 40 dB. A transmit gain block includes a programmable waveform generator 215 , which provides an arbitrary waveform to a power amplifier 220 capable of outputting high voltage pulses to the PZT transducers. The power amplifier 220 may have a fixed gain. Thus, the transmit gain may be determined by the programmable waveform generator. Alternatively, the power amplifier 220 may have variable, programmable gain.

The receiver filter 205 and gain block 210 and the transmit gain block (i.e., the waveform generator 215 and power amplifier 220 ) may be applied in a coordinated manner to determine an optimal transmit gain and receive gain condition, where the optimal condition may be specified by requiring that the received RF signal envelope input to the analog to digital converter ADC 225 following the analog receiver filter 205 and gain block 210 does not saturate the ADC 225 .

Filter/Gain Initialization

As illustrated in FIG. 3 , in one embodiment, an adaptive method 300 for determining the optimal transmit and receive conditions for each signal path at each actuation frequency includes initializing the receive filter 205 , the transmit gain and the receive gain 210 . Based on selecting an initial transmit frequency (step 305 ), the receive filter 205 may be initialized (step 310 ) to a matched filter characteristic, where the initial transmit frequency may be selected one by one from the candidate frequencies (step 10 of method 100 above) for the respective signal path. The transmit gain 210 may be set initially at the maximum possible value (step 315 ). Alternatively, transmit gain may be set at a user selected value. The receive gain may be set at a desired selected value (step 320 ), e.g., a gain intermediate the minimum and maximum gain values.

Receive Amplifier Gain Setup

In FIG. 3 , an initial or desired variable gain is input to set the analog receive amplifier to a selected receive gain (step 320 ).

Noise Level Determination

Following initializing of the receiver filter (step 310 ) and transmitter gain (step 315 ) and receiver gain (step 320 ) above, the method 300 proceeds to determine the noise level in the receiver section of the acquisition and processing system 200 of FIG. 2 . A transmit switch SW T 230 may be turned off while the receive block switch SW R 235 is on (step 325 ), i.e., the receiver block is in communication with a receive transducer for a selected path. The received signal is noise, and may be digitized and stored to a memory for analysis (step 330 ). The maximum received digitized signal amplitude is defined (step 335 ) as the maximum filtered, amplified and digitized received system noise for the receive gain setting.

Signal Level Determination

With the receive switch SW R 235 on, the transmit switch SW T 230 may be turned on (step 340 ), i.e., the transmit block is placed in communication with a transmit transducer, and an acoustic signal is transmitted for detection by the receive transducer and receive block of the system 200 . The received analog filtered and gain amplified signal may be digitized and stored to the memory for analysis (step 345 ).

Preferably, cross-talk is removed from the signal (step 350 ). Cross-talk may be caused by EMI from direct radiation associated with the high voltage that may be required for actuators to excite the acoustic wave signals. The acoustic wave signals may be transmitted as packets, or pulses. Cross-talk may significantly affect the accuracy of detecting the arrival of acoustic pulses at the sensor transducers, and consequently, also affect the detection of damage. Therefore, removal of cross-talk from the signal is an important step.

Since the time of arrival of cross-talk is substantially instantaneous, various methods of time gating may be used for accomplishing this step. For example, in a digitized signal waveform, assuming the signal pulse length (taking into account the ring-down from the transducer) is shorter than the shortest time of arrival, digital signal processors, such as may be found in digital oscilloscopes, may be set to null all values of the digital signal waveform from the time corresponding to the trigger point to a time equal to or greater than the pulse length. Requiring the signal pulse length (and ring-down time) to be less than the shortest time of arrival insures that the cross-talk signal will not overlap the elastic wave signal, and cross-talk removal is thereby more easily facilitated.

The signal level SIG may be defined as the maximum received signal amplitude after removing the cross-talk corresponding to the current selected receive gain value. The SIG level may result in saturation of the input to the ADC 225 if it exceeds a maximum limiting value, SIG MAX . If the received signal level SIG exceeds SIG MAX , the receive gain may be reduced, e.g., by reducing the variable gain level. If the receive gain is already reduced to the minimal value and SIG still exceeds SIG MAX , then the transmit gain may be reduced until SIG is at an acceptable level.

In the example method 300 , SIG is also compared to the receiver noise level to compute a signal-to-noise ratio (SNR). Both SIG and SNR may be determined (step 355 ) and the results used to adjust transmitter and receiver gains, in the following manner. If SIG exceeds SIG MAX (a “Yes” result in decision block 360 ) the receiver gain can be reduced (step 365 ), and a new signal transmitted. Recording of a new acoustic signal may be obtained by continuing at step 345 . If SIG is less than SIG MAX (a “No” result in decision block 360 ), then the signal may be determined not to have reached a saturation level, and is then tested for SNR (decision block 370 ). If the SNR does not exceed a minimum selected value, e.g., SNR MIN =20 dB (a No result in decision block 370 ), the transmitter gain can be increased (step 375 ) and the new acoustic signal obtained by continuing at acoustic signal record step 345 . If the SNR is greater than SNR MIN (a Yes result in decision block 380 ), the example method 300 ends and the gain and frequency parameters for the selected path are stored (step 380 ).

The transmitter and receiver gain may now be considered correctly set, and initializing may proceed with selecting functional paths for all the candidate frequencies by comparing the signal with a minimum threshold for signal strength and/or SNR.

Determining Optimal Frequency

Since different paths may be sensitive to different actuation frequencies, it is desirable that different actuation frequencies be used for different paths. Alternatively, using an identical actuation frequency for all paths is computationally more efficient. A reasoning process in damage detection may allow use of both frequency selections depending on user specification or a threshold requirement.

FIG. 4 is a process flow diagram of an example embodiment of a method 400 for selecting operating frequencies for an array of actuator-sensor paths in an example SHM system. Let N be the total number of paths 1, . . . n, . . . N selected (step 405 ). Let Q be a measure of signal quality. Let {f 1 , f 2 , . . . , f M } (or, for convenience, denoted by {f m }) be a set of M candidate frequencies for each actuator-sensor acoustic transmission path (step 410 ), where the number of candidate frequencies M for each path may be different depending on the selection of functional paths (step 40 of method 100 of FIG. 1 ). As indicated above, examples of Q that may be used to select frequency include, but are not limited to, a measured total signal energy, a measured or estimated signal-to-noise ratio, or a measure of robustness to environmental parameter changes.

As illustrated in FIG. 4 , a pseudo-code for selection of different best suitable frequencies for different paths is:

Alternatively, a single and identical frequency may be selected for all paths. For example, corresponding to each candidate frequency f m , compute the total path number K m for which f m was selected as the best suitable actuation frequency (step 450 ). Let {K 1 , K 2 , . . . , K M } be the set of these numbers. The maximum value in the set of {K 1 , K 2 , . . . , K M }, i.e., the largest value of K m , may be found (step 455 ). Denote it by K max and the corresponding frequency by f max . If K max /N is greater than a threshold (e.g., 90%) (a Yes result in decision block 465 ), then f max is selected (step 470 ) as the best suitable frequency for all the paths. Otherwise, use different actuation frequencies for each path as selected above (step 475 ).

Although the methods of the present invention have been described and illustrated herein with reference to certain specific example embodiments thereof, it should be understood by those of skill in this art that a wide variety of modifications and variations may be made to them without departing from the spirit and scope of the invention, as defined by the claims appended hereafter and their functional equivalents.