Systems and methods for real-time processing of signals from a multi-element transducer array

Abstract

Systems and methods for real-time processing of signals from a multi-element transducer array for detecting transient elastic waves originating from unknown locations in a body, which may propagate in a dispersed manner. The systems and methods allow real-time combination and analysis of the signals. The systems and methods include two time coherence circuits for reliably determining a small time shift that puts all signals into a coherent time base, and then combining the signals to achieve a composite waveform with an increased signal-to-noise ratio compared to any single element.

Description

Technical Field

[0001] The present invention relates to detecting transient elastic waves in a subject using a transducer, detector, transceiver or receiver array. Background Art Summary of the invention

[0002] Provided herein is a system and method for real-time processing of signals from a transducer array for detecting transient elastic waves originating from unknown locations in a subject, which transient elastic waves may propagate in a dispersed manner. The system and method allow real-time combination and analysis of signals, including decisions about storage when new data is received. The method described herein provides a method for designing a detector array and for real-time processing of signals, given the constraints of the subject being measured, determining whether to store the set of information while receiving a new set of information for processing in a real-time environment, ensuring greater system sensitivity. The method described herein includes a method that results in determining or minimizing small time shifts, the method placing all signals in a coherent time base, and then combining the signals to achieve a composite waveform with an increased signal-to-noise ratio compared to any single element.

[0003] In one aspect, the present disclosure describes a method comprising: determining in real time whether to store in a computer memory a first set of samples of a plurality of signals from a multi-element transducer array coupled to a body of material under test in a real-time processing environment. The first set of samples represents a first time range and ends with a first set of boundary samples, which are subsequently also processed together with a second set of samples of a plurality of signals representing a second time range, the second time range having a necessary overlap at the end of the first time range, thereby generating overlapping multiple processed samples, including samples processed with the first set of samples and samples processed with the second set of samples. The overlapping multiple processed samples are sufficient to capture the slowest measurement wave of interest within the body under test in the multi-element transducer array. The slowest measurement wave of interest is sensed by a first portion of the multi-element transducer array during the first time range and by a second portion of the multi-element transducer array during the second time range. Determining in real time whether to store the first set of samples also includes calculating a plurality of corresponding delay times of each of the plurality of signals relative to a predetermined reference signal in the plurality of signals, so that the time shift of the plurality of corresponding delay times modifies the plurality of signals to be coherent with the time base of the predetermined reference signal.

[0004] The calculation of the corresponding delay times may also be performed by a second time coherence circuit configured to calculate a second set of multiple time delays between a first reference wideband signal and the remaining wideband signals in the plurality of wideband signals for comparison with the first set of time delays to evaluate the time coherence calculation. The second time coherence circuit may use the same or a different algorithm for calculation.

[0005] The real-time determination also includes combining at least one of the time-shifted plurality of signals with a reference signal in the plurality of signals to generate a real-time combined signal from the plurality of signals. The real-time determination also includes evaluating the real-time combined signal over the first set of samples for values ​​of the real-time combined signal exceeding a predetermined threshold. The real-time determination also includes instructions to store the first set of samples in a computer memory if the real-time combined signal exceeds a predetermined threshold in the real-time combined signal in the first set of samples.

[0006] On the other hand, the present disclosure further describes a system comprising a body of a material under test, the body of the material under test being suitable for detecting propagating transient elastic vibrations having a slowest moving wave component, the propagating transient elastic vibrations being transmitted from the transient elastic wave deformation information source through the body of the material to a multi-element transducer array. The system also includes a maximum spacing of the multi-element transducer array coupled to the body, the spacing being predetermined by the computational and storage limitations of the hardware and its ability to determine the coherence of the signals from the multi-element transducers while maintaining the required positive overlap to ensure increased sensitivity. The system also includes a receiver circuit for processing a plurality of signals received from the multi-element transducer array. The receiver circuit also includes a circuit for determining a time delay between a first reference signal and other signals in the plurality of signals. The receiver circuit also includes a combining circuit for generating a combined signal based on the time-shifted signal. The receiver circuit also includes a threshold detection circuit for detecting in real time whether an output value of the combining circuit exceeds a threshold when the value of none of the plurality of received signals exceeds the threshold.

[0007] Other embodiments and features of the present disclosure will become apparent from the accompanying drawings and the detailed description that follows. BRIEF DESCRIPTION OF THE DRAWINGS

[0008] Figure 1 shows an exploded view of an exemplary transducer array described herein, including associated electronics and channels for each transducer element;

[0009] Figure 2 An example diagram depicting sample overlap is shown for determining in real time whether to store a first set of samples;

[0010] Figure 3A , Figure 3B , Figure 3C shows the angular dependence of the phase velocity of transient elastic vibrations during transmission along the circumferential direction, the axial direction, and at 45° to the axial direction of an exemplary test composite container;

[0011] Figure 4 The effect of source orientation of the vibration source relative to the transducer array is shown;

[0012] Figure 5The additional effect of the source orientation of the vibration source relative to the transducer array is shown;

[0013] Figure 6 shows details of an exemplary transient elastic wave measured by a single element from a source at a parallel angle of incidence;

[0014] Figure 7 shows the signals detected from a parallel oriented pencil lead break source to a transducer array and the digitally summed and shifted signals as described herein;

[0015] Figure 8 shows the signal detected from a vertically oriented pencil lead break source to the transducer array and the digitally summed and shifted signal as described herein;

[0016] Fig. 9 The time differences calculated for different angles of incidence for summing signals in an exemplary linear array of four elements are shown;

[0017] Fig.10 shows the increase in bending mode amplitudes for alternative summing possibilities of a measured subject compared to the digital real-time summing method described herein;

[0018] Fig.11 shows the conversion of channel signals into temporal coherence calculated using three inter-element time shift (Δt) determination algorithms according to one embodiment of the present invention;

[0019] Fig.12 A table is shown summarizing the Fig.11 The calculated inter-element time shift (Δt) and the increase in signal response when all elements are time shifted the indicated amount and digitally summed compared to the best responding element (channel 1).

[0020] Embodiments are illustrated by way of example and not limitation in the figures of the accompanying drawings in which like references indicate similar elements. DETAILED DESCRIPTION

[0021] The following patent description and drawings are illustrative and should not be construed as limiting. Many specific details are described to provide a thorough understanding. However, in some cases, well-known or conventional details are not described to avoid ambiguous descriptions. References to one or an embodiment in this disclosure are not necessarily references to the same embodiment; moreover, such references mean at least one. References to "one embodiment" or "embodiment" and the like in this specification mean that a particular feature, structure, or characteristic described in conjunction with the embodiment is included in at least one embodiment of the present disclosure. Phrases such as "in one embodiment" and the like that appear in different places in the specification do not necessarily all refer to the same embodiment, nor are they necessarily separate or alternative embodiments that are mutually exclusive of other embodiments. In addition, various features are described that may be exhibited by some embodiments but not by other embodiments.

[0022] Figure 1 An exemplary transducer array further described herein is shown having four transducer elements 11 and four separate data channels for transmitting data for processing and / or storage. The transducer array is disposed between an interface portion 10 and an array housing 6, which may have a removable cover 1 and may be pressurized through a port 4. The housing 6 contains electronics, for example, as shown on a circuit board 7. The interface portion 10 may be covered or disposed by an optional barrier 3 (e.g., a watertight barrier) that is retained by a frame 2 and fasteners 5 for creating a flexible fluid-filled bladder. In certain measurement scenarios (e.g., when coupling to a rough or highly irregular surface is desired, or when automated sensor placement is required), it is not possible to couple a conventional sensor with a rigid interface between the component under test and the piezoelectric element; in these cases, a flexible fluid-filled bladder can be coupled to the component under test. By filling the flexible bladder with a fluid (wherein only expansion wave propagation is supported), the fidelity of the measured waveform can be acceptably maintained with adequate coupling.

[0023] The array housing 6 shows a like number of ports for relaying data from the circuit board 7 for further processing. In another embodiment, several or all data channels may also be multiplexed together onto a single port or channel for transmission purposes.

[0024] However, regardless of how the signals and data are transmitted from the transducer through the electronics, in order to accomplish the real-time processing described herein, the received data samples must be processed by methods that do not lose data. As further described herein, operating these methods in real time without loss of data requires overlap of sample processing and, in fact, overlap of samples in the system.

[0025] In other embodiments, the multi-element transducer array may be configured in two dimensions. This will make it easier to unambiguously obtain source direction information in the full 360° field.

[0026] Figure 2 An example graph depicting sample overlap is shown for determining in real time whether to store a first set of samples. The disclosed method utilizes a specifically configured array to process signals from the array in real time in a manner that improves the signal-to-noise ratio (SNR) of the received signal to exceed that of any single channel. The graph shows the positive sample overlap required to detect an event through the array without the possibility of missing an event that may exceed the thresholds described and required herein, thereby ensuring real-time performance and data collection.

[0027] A single exemplary event is shown as it would be detected by each transducer of the array, and the single event is captured by sufficient positive sample overlap. This positive sample overlap shows the result of a set of engineering decisions used to implement the systems and methods described herein. The waves shown in the single exemplary event represent the slowest moving wave component in the frequency band of interest of the subject under test as determined by the dispersion relation (or body wave propagation velocity, as the case may be), the wave mode of interest, and the transducer array geometry. Each of these aspects of this real-time determination will be further defined in the corresponding embodiments herein.

[0028] The disclosed method allows for determinations to be made in real time based on the material being measured and the specific transducer array characteristics and geometry. When making these decisions, it is necessary to fully understand the events that may be missed due to the events arriving at the array portion in one set of samples and arriving at other portions of the array in another set of samples. For example, real-time decisions can be made for the storage, transmission and / or further processing of each set of N samples in the real-time environment described herein so that these decision-making processes do not lose or omit data and / or events. These decisions allow for advanced real-time triggering decisions that help improve the signal-to-noise ratio (SNR) of the array compared to the SNR of the individual elements of the transducer array. These decisions include processes for detecting a set of time differences for coherently combining data, combining the data into a composite signal, and analyzing the signal before making a decision to store the information and starting processing the next set of received data.

[0029] The sample overlap shown in the figure represents the amount of time that samples can be processed by the analysis process herein (e.g., including decisions made regarding storage) on a first set of samples while a second set of samples is being received. The overlap in time between the first and second sets of samples ensures that for a given sampling rate, the slowest moving wave components must be detected in the same set of samples in order to enhance the increase in array SNR.

[0030] Thus, the boundary sample group is processed with the first set of samples and the second set of samples. As used herein, the terms boundary sample group and sample overlap may be used interchangeably to describe the sample group processed with the first set of samples and the second set of samples, with no gaps in samples for real-time processing. The real-time processing requirement may be expressed as a positive overlap of processing time, specifically, for a positive overlap, the available processing time is more than the processing time required to process (e.g., analyze, combine, store) a set of samples (including the associated boundary sample group) while receiving another set of samples for processing.

[0031] The timing requirements of real-time processing require positive sample overlap, as shown in the figure. This positive overlap creates the ability to capture an exemplary event within the array and process the entire event within the time represented by a set of samples. In this way, the transducer array can appropriately capture transient elastic waves within the subject from events at any angle relative to the array. The maximum overlap required for real-time processing occurs at the event that records the maximum time difference between events arriving at each transducer, measured by the slowest moving wave of the transient elastic waves from events within the frequency band of interest. Therefore, the positive sample overlap or processing time overlap required for real-time processing can be calculated from the separation of the transducers receiving the transient elastic waves within the subject, the sampling rate, and the speed of the slowest moving wave of the expected event to be detected.

[0032] As further described herein, signal combination by the methods herein requires that at least two signals detected from an event (and typically more, e.g., four, nine, sixteen, etc. signals) must be combined in order to maintain the described gains in the effective signal-to-noise ratio of the transducer. These gains may be necessary to detect certain events above the noise level or to increase the spacing of the sensors while still maintaining adequate coverage of the structure, as further described herein. Thus, the methods described herein may include a system having specific features for ensuring that real-time processing can be achieved without loss of data.

[0033] For example, Figure 2As shown, the initial N samples representing 13 seconds of monitoring time (t=0-13 seconds) are designated as the boundary sample group for its last 3 seconds (from 10-13 seconds) before the second group of N samples begins (e.g., 13 seconds on the graph). Therefore, in the exemplary embodiment shown in the figure, in order to process the initial N samples in real time, the samples representing from 0 seconds to 13 seconds must be stored in physical memory and the first time coherence determined while receiving the second N samples for its own processing (digitization, physical memory storage and time coherence determination). In order to ensure real-time increase in the SNR of the combined signal from the array, the computing power of the hardware must be such that the time coherence of the set of N samples and its corresponding boundary set is determined, and the trigger decision is made before the available memory overflows (if the threshold is exceeded, the data is transferred to the PC for storage, or if no trigger occurs, the physical memory space is cleared).

[0034] A second independent temporal coherence may also be determined to calculate a plurality of time delays between a first reference wideband signal and remaining wideband signals in the plurality of wideband signals for confirming or determining that the algorithm for calculating the temporal coherence is inconsistent.

[0035] In order to make the evaluation more robust, the second temporal coherence determination may be performed by utilizing an alternative coherence determination algorithm, ie an algorithm different from the first temporal coherence determination.

[0036] The first and second sets of calculated time delays can be compared to each other to evaluate the temporal coherence calculations. For example, the comparison of the first and second sets of time delays can determine the consistency or inconsistency of the temporal coherence calculations of the first set of time delays. The use of more than one calculation of the time delays can improve the coherence of the combined signal, provide greater accuracy in the triggering decision to pass data on storage if a threshold is exceeded, or clear physical storage space if a trigger does not occur.

[0037] Shown in the figure is a set of boundary samples representing the time from 10 seconds to 13 seconds. This boundary sample set is processed with both the initial set of N samples and the second set of N samples representing the time from 13 seconds to 23 seconds. The process is repeated thereafter, for example, forming a second set of boundary samples between the second set of N samples using samples from 20-23 seconds and the third set of N samples representing data from 23-33 seconds. This second set of boundary samples will be processed with the second set of N samples as well as the third set of N samples.

[0038] In this way, if Figure 2As shown, although the event arrives at the top channel after 13 seconds and after the first set of N samples (and the first set of boundary samples) are completed, the event can still be captured by analyzing the second set of N samples that also include the first set of boundary samples. In some cases, analyzing only the first set of N samples and the boundary samples will miss the event. For example, in some cases, if the first N samples and the boundary samples do not contain enough events so that the sum of the signals described herein does not cause the combined signal to exceed the threshold, the event may be missed.

[0039] As another example, consider transient elastic vibrations propagating in a plate structure with a calculated dispersion relation. In addition, consider a source with a dominant out-of-plane source orientation that preferentially excites bending modes. From the dispersion relation of the plate / material property combination, it is determined that the slowest moving component of the lowest frequency bending mode of interest in the test is 0.8 mm / μs. The final information required to determine the necessary time overlap is the array geometry and the inter-element spacing. For this example, consider a four (4) element 1D array with 15.9 mm inter-element spacing. From this information, it is determined that in the limiting case (relative to the parallel source orientation of the 1D array), the maximum distance a wave will travel through the array will be from element four to element one (or vice versa). From the velocity of the slowest moving wave component (0.8 mm / πs), it is determined that a minimum time overlap of greater than or equal to 59.625 μ is required to ensure that there are no gaps in the data acquisition, thereby achieving a real-time increase in the array SNR.

[0040] The timing and sample sizes discussed herein with respect to this figure are exemplary and may vary significantly in various applications based on the details of the transducer array monitoring the subject under test, the wave speed, and the frequencies of interest. For example, while certain setup and processing constraints may be illustrated by an inter-element spacing of 15.9 mm or by a sample overlap representing a boundary time frame of 3 seconds, other applications may have different spacings or much shorter boundary times, as calculated herein based on transducer size and / or characteristics of the elastic waves being measured.

[0041] A detailed discussion of the sampling frequency for selecting an embodiment is not included in this disclosure, as these techniques are well discussed in digital signal processing work. In general, the sampling rate must be high enough so that the highest frequency component of interest meets the Nyquist frequency criterion and is not eliminated by the anti-aliasing filter.

[0042] Figure 3A , Figure 3B , Figure 3C The angular dependence of the phase velocity of transient elastic vibrations during transmission along the hoop direction, the axial direction, and at 45° to the axial direction of an exemplary test composite container is shown. Figure 3A A set of dispersion relations for elastic waves traveling along the hoop direction of the composite pressure vessel are shown. Figure 3B A set of dispersion relations for elastic waves traveling along the axial direction of the composite container is shown. Figure 3C A set of dispersion relations for elastic waves traveling at 45° to the axial and circumferential directions of a composite pressure vessel is shown. Other measured bodies may have different bulk ultrasound or bulk seismic wave velocities, and the transducer array may use these different wave velocities, as further described herein, to determine the appropriate array geometry and number of transducers and to determine the location of the event. FIG. 3A to FIG. 3C It can be clearly seen that the mode velocity depends strongly on the propagation direction of the mode in the measured body. The dispersion relation can include different wave modes of elastic waves traveling through the measured body (e.g., extensional wave modes, bending wave modes). These various characteristics of transient elastic waves can be used to design transducer arrays and to modify the methods of this article (e.g., the maximum number of lags to be considered in the time shift algorithm, or the number of calculations required for the temporal coherence determination) for real-time analysis and storage decisions.

[0043] Techniques in active phased array ultrasonic testing (PAUT) can use an inverse time delay applied to the detection signal of each element according to the focusing law to time-shift all detected waveforms into the zero time register, and then add the resulting time-shifted coherent waveforms to produce a synthetic amplitude scan (or A-scan). Conventional phased array sensing provides PAUT with a significant improvement in the sensitivity (and SNR) of the received signal. The increase in detection sensitivity using the sum of the signals from the phased array in the PAUT is a simple effort, where the bulk ultrasound waves propagate from a known focus because ultrasound waves are non-dispersive (i.e., the velocity is constant for all frequencies), and the focusing law can be used to set the appropriate time offset for each element to shift the waveform of each element back into the zero register. In the case of passive waveform measurements where the source is broadband (e.g., modal acoustic emission (MAE) testing, seismology, etc.), the wave propagation can be dispersive and the source originates from an unknown position relative to the array, and a set of closed-form time delays are not applicable to time-shift the waveform from multiple elements to coherence. However, using the methods described in this article to analyze the signals in real time, SNR gains similar to those achieved in PAUT applications can be achieved for transient elastic waves measured in situations where the closed form time reversal law does not apply (i.e., broadband sources, dispersive wave propagation, and / or unknown source origin).

[0044] The composite pressure vessel tested allowed for precise geometry and structural measurements to be made in order to establish the real-time capabilities of the methods described herein for real-time processing. These tests performed on the composite pressure vessel confirmed that the methods were able to correctly determine the appropriate time differences in real time despite the dispersion, multi-mode, attenuation, and other characteristics of the acoustic waves propagating therein. For other subjects under test with broadband sources, dispersive wave propagation, or sources from unknown locations, making them unsuitable for traditional time reversal techniques, these subjects can use the techniques described herein to analyze the signals in real time based on their ability to determine the appropriate time differences without having to fit a set of closed-form time differences to each detected event.

[0045] The composite material in the composite test container provides a viscoelastic medium for measuring transient elastic waves. The polymer matrix in the composite material can be composed of a viscoelastic resin so that attenuation increases with frequency. Exemplary composite materials can attenuate higher frequencies more times than lower frequencies. Some analyses of these subjects may require examination of the frequencies in the subject, and as used herein, the term "measured frequency" or "frequency of concern" refers to the frequency that needs to be analyzed or used for a specific analysis being performed (e.g., correctly identifying a damage mechanism originating from within a microstructure, or identifying the location of an event in a subject).

[0046] Furthermore, each of these transient elastic waves may have a separate travel speed through the medium, which results from the dispersion of the transient elastic waves representing the event and creates an additional requirement to maintain positive overlap. Therefore, for transducer array design purposes, the frequencies of interest will be analyzed for their "slowest moving wave component," used herein to mean the portion of the transient elastic wave having the slowest velocity in the target material. As further described herein, the slowest moving wave can be used to determine several desired parameters of the transducer array and the required computational power of the time-shifting hardware using the methods herein.

[0047] Furthermore, in the presence of material anisotropy, the slowest speed may be different in different directions. Therefore, the array structure must take into account the characteristics of the incident wave and the need to capture the entire event in a boundary sample set that may be processed with two sets of surrounding samples.

[0048] As further described herein, during operation of the systems and methods herein, processing time and acceptable overlap of samples can be controlled to adjust the operation and detection of events within the subject being tested. For example, while processing time may fluctuate between events, an additional process of monitoring the desired and actual overlap available and used for each detected event can be employed to monitor the methods herein.

[0049] Figure 4 and Figure 5The effect of alternating orientation of a transient vibration source relative to a transducer array is shown. A transient vibration source (e.g., an event, material break, etc.) induces transient elastic waves in a measured body. The source will produce a broad spectrum of vibration frequencies, each of which may have a transmission velocity based on the elastic medium and geometry. The method herein employs a transducer array measuring the measured body such that the slowest moving wave propagating from the source will be detected by each element within at least the overlap time or relevant sample overlap available during processing.

[0050] The methods described herein provide real-time processing of data sets from signals having overlap between data sets consisting of the maximum time difference between arrival of an event and completion of each transducer element in the array. In one embodiment, the array structure and processing are adapted so that events from any location within the subject under test can be captured by the array with positive overlap, regardless of propagation angle, capturing the slowest moving waves and allowing combination of the entire event, even from a source angle that maximizes detection delays between elements in the array (e.g., aligned parallel to the axis of a 1D array).

[0051] In some embodiments, more overlap is available due to computational processing efficiencies greater than required by the physical array structure, the slowest moving transient elastic wave component, and the time range of the relevant events. In an alternative embodiment, the array structure and processing are adapted to adjust the overlap between angles from zero overlap to positive overlap.

[0052] like Figure 4 As shown, a transient elastic wave source emanating from an angle perpendicular to the linear (1D) array (i.e., at a 90° angle) will deliver the transient elastic wave to the transducer array with minimal delay between elements. For example, if the source or event is far enough from the transducer array, a planar assumption can be made for the wave, and the arrival time differences between the transducer array elements will be close to zero.

[0053] Or, if Figure 5 As shown, a transient elastic wave source emitted from an angle that is in line with a linear (1D) transducer array (i.e., zero degrees, parallel to the axis of the array) has a maximum delay time between the transducer signals of the array. For example, the source wave incident on the first element will need to be matched with a delay of an integer multiple of the delay of each consecutive element in the array based on the position of the transducer that produces the corresponding signal. As further described herein, other angles may produce delays, which require different delay times to make the signal reach temporal coherence. These different angles can produce different SNR increases in the combined signal based on the combination efficiency and the ability of the described method to correlate and eliminate noise, as described herein.

[0054] Figure 6Details of exemplary transient elastic vibration waves measured by individual elements (1-4) from a source at a zero degree angle (i.e., the waves propagate in a direction parallel to the array) are shown. The incident wave with a specific velocity is based on the dispersion relation and comes from a direction that is in line with (i.e., parallel to) the axis of the transducer array. This direction maximizes the time difference between the signals received by each transducer, while there will also be other masking distortions (e.g., due to dispersion, attenuation, electronic noise) between the received transient elastic waves. As further described herein, the time difference between the received signals can be determined in real time to combine the signals into coherence with a specified element in the array (e.g., element 2, element 3), thereby increasing the effective SNR in real time and allowing enhanced processing steps (e.g., normalization, real-time storage decisions). By making each signal coherent with the center element of the array (e.g., element 2 or element 3 in a linear (1D) 4-element linear array), processing requirements can be reduced by making minor adjustments to each signal. The following figure includes measured signals from a linear (1D) array of four elements, including a sum signal generated in real time, ensuring the method described herein.

[0055] Figure 7 Signals detected from a pencil lead break source at 0° relative to the transducer array and digitally summed and shifted signals as described herein are shown. In this embodiment, the source location is the same as shown in the previous figure, emphasizing the time difference between the signals. As shown, the wave generated by the induced event (i.e., pencil lead break) is first detected on channel 4 and then detected on channel 3-1 in reverse order of sequence. As further described herein, the combination of the signals (e.g., time-shifted and summed signals) shown in the fifth trace at the bottom shows a combined signal with an amplitude greater than each of the component signals. In another embodiment, the sum of the signals can be scaled, normalized, or otherwise modified to produce a better combined signal.

[0056] As shown in the combined signal, the signal-to-noise ratio of the individual signals is increased via the processing described herein (i.e., correlation, time-difference calculation, combination), and this SNR gain of the combined signal is described in detail in the following figures. The SNR gain depends on the angle of incidence of the wave from the source on the array, the subject under test (e.g., its dispersion and other characteristics), and the frequency content of the transient elastic wave from the source, but is shown to be achievable and robust using the methods described herein.

[0057] In one embodiment, the time difference calculation for the combination is generated by determining the maximum value of the cross-correlation product between the signal from a predetermined reference element (e.g., element 2) and the signals from other individual elements. However, there are other time-shifting algorithms and / or calculations that may be applicable to the current real-time environment that may produce an acceptably accurate time difference to accomplish the process described herein. Thus, one or more of several existing algorithms and / or calculations may be suitable for generating the combined signal used by the methods and systems described herein.

[0058] Cross-correlation is a mathematical construct that provides a measure of the similarity of two time series as a function of the lag of the sampling points of one signal relative to the other signal. The following equation provides the mathematical definition of the mth cross-correlation coefficient of two signals x and y, each of length n,

[0059]

[0060] By time-shifting the two sequences by multiple lags, the offset between the two sequences can be determined by the point where the two signals are most similar or the waveforms exhibit maximum coherence (i.e. From a conceptual point of view, Figure 4 and Figure 5 shows how each of the four individual elements of a phased array will receive the wavefield at slightly offset times from each other as the stress wave propagates through the medium. By using cross-correlation, the waveforms captured by all four elements in the array can be shifted to appear as if they were all detected simultaneously. As an example of determining an upper bound on the number of calculations required for a given array, consider the worst case scenario with respect to the time difference between the elements (i.e., parallel wave incidence), e.g. Figure 6 As shown. Figure 6 It can be seen that the wavefield is incident on element 4 first and on element 1 last. Assuming that the slowest moving component of the flexure mode of most engineering materials (i.e., the lowest frequency of interest) has a velocity of no more than 0.5 mm / μs in most practical MAE inspection scenarios, the equation establishing the maximum amount of time (or hysteresis) required to shift the waveform into register can be written as

[0061]

[0062] Where n is equal to the number of components to be shifted, dx is the spacing or pitch between components, and c fis the slowest moving bending mode speed. In this example and orientation, shifting all waveforms to element 2 requires n=2 (t=63.6 or 318 lags if sampled at 5MS / s), while shifting to elements 1 or 4 requires n=3 (t=95.4 or 477 lags if sampled at 5MS / s). Therefore, to minimize the number of required calculations and without loss of generality, all waveforms are shifted to element 2, which will be referred to as the reference element hereafter. Although the computational requirements of the time-shifting algorithm can be reduced by choosing a favorable reference element, the maximum width (or length, depending on the orientation) of the number of array elements to be covered must be considered when determining the sample overlap requirement.

[0063] Figure 8 The signal detected from a pencil lead break source at 90° relative to the transducer array is shown along with the digitally time-shifted and summed signal as described herein. In this embodiment, the source is located at 90° to the axis of the array, thus allowing the transient elastic wave to arrive at each transducer nearly simultaneously. However, the methods described herein can still be used to increase the SNR, as shown in the combined signal trace.

[0064] As in other embodiments, this embodiment can use the same time difference calculation and processing techniques as described for any other event that emits a transient elastic wave from any source location. For clarity, in most embodiments, the systems and methods described herein do not know anything about the source location of the wave, so the same determination and calculation will be necessary for each set of samples received. In other embodiments, some information about the source location is known, and adjustments can be made to the methods and systems herein based on that information. Therefore, the amplification and triggering decisions for each channel must take into account the changes in receiving the wave from different angles.

[0065] Fig. 9 The time differences calculated for different incident angles for summing the signals in an exemplary linear array of four elements are shown. As expected, for sources with perpendicular (90°) orientation, the time shift between elements is relatively low and increases steadily as the source orientation approaches the parallel (0°) source position.

[0066] Furthermore, the time differences shown in the figures are all determined with the reference element being an internal element of the linear array (element 2). In several embodiments, time differences are calculated for internal elements of the array, minimizing the time differences and processing power requirements between sample groups. Furthermore, the required time differences for summing the signals may be different for a particular element based on the material or subject being measured, and calculating those differences in real time after receiving the signal from the event allows for small but important changes in the time differences to optimize the real-time detection and SNR increase aspects of the systems and methods herein.

[0067] Fig.10The increase in bending mode peak amplitude for an alternative summing possibility for a subject under test is shown compared to the digital real-time summing method described herein. Each row in the figure provides results comparing direct analog to digital time shifting and then summing from a different propagation angle. The reported summation gain obtained via analog summation provides a modest SNR gain on a single channel, however, it does not increase or consistently increase SNR sensitivity like the disclosed digital shift and sum signal. The disclosed combined signal provides SNR gains reported in phased array modal acoustic emission (PA-MAE) systems.

[0068] The increase in SNR is important to the methods and systems described herein, but the coherence of the SNR increase is also an important factor in the system design to achieve the goal of providing complete coverage of the subject under test while minimizing the number of arrays that must be placed on the structure. By increasing the SNR, the number of transducer arrays required can be reduced by increasing the effective dynamic range of the system.

[0069] Therefore, real-time processing requirements must be met to prepare the triggering scheme, and the analog summation techniques shown cannot provide the benefits of the methods and systems herein because they both provide a small SNR gain and provide a varying SNR gain depending on the propagation angle. Because the systems and methods herein are used with arrays to detect waves at any propagation angle, the variation in SNR increase further complicates the process of triggering the detected event.

[0070] To further improve the time offset, a second time coherent circuit may be employed to determine the time delay. Fig.11 Shown by Figure 6 The signal captured by the four element array shown and the signal after digital summation and shifting. In this embodiment, the source position relative to the array is Figure 6 The same as shown, emphasising the time differences (delays) between the signals. There may also be further time coherence circuitry to provide further redundancy. This provides the possibility of redundant determination of inter-element time shifts to coherence and enables a determination whether the similarity algorithm has failed on a particular data set and a "voting scheme" as to which set of time shifts are accepted for coherence determination.

[0071] Depend on Figure 6 The array captured in Fig.11 The waveforms in can be analyzed by using three unique inter-element time shift (Δt) determination algorithms, i.e., inter-element cross-correlation, Akaike Information Criterion (AIC), and fixed threshold crossing. Then, the signals of all elements are converted to be time coherent with element 3. Fig.11 The top four (curves) show individual signals from each element, while the Fig.11The bottom graph shows the waveforms being digitally converted to time coherence and summed. In this summed curve, the signal is time shifted by the delay determined by the cross-correlation algorithm.

[0072] Fig.12 A table is shown summarizing the Fig.11 The curve in Figure 1 shows the calculated element-to-element time shift (Δt) and the increase in signal response compared to the best responding element (channel 1) when all elements are shifted by the indicated amount and digitally summed.

[0073] The present invention discloses the inter-element time shift (Δt) calculated according to the three algorithms mentioned above, namely, inter-element cross-correlation, Akaike Information Criterion (AIC), and fixed threshold crossing. In addition, the table shows the increase in signal response when all elements are time-shifted by the indicated amount and digitally summed compared to the best responding element (element 1). Fig.12 It can be seen that for the dataset, the determination of the cross-correlation and AIC time-shifts is very consistent, while the determination of the fixed threshold crossing is not.

[0074] Through such determination, a voting scheme can be established among multiple time-shifting algorithms to identify whether a given data set has caused a particular algorithm to fail, and enable the PA-MAE system to "trust" that the time shifts determined for a given data set are consistent with each other (within a certain tolerance).

[0075] Furthermore, the voting scheme may include variable weighted importance of different algorithms. As an example, if the calculated cross-correlation coefficient is above a threshold indicating a high similarity between the signals and a probability of detecting a real event, the cross-correlation algorithm may receive a larger weight in the vote. Conversely, if the cross-correlation coefficient is below a threshold indicating that the signals are not highly similar, the cross-correlation determined time shift algorithm may receive a smaller weight in the vote.

[0076] Therefore, the methods and systems described herein must be adapted to measure transient elastic waves to provide an appropriate SNR boost at many or all propagation angles. The methods and systems allow for receiving SNR-enhanced signals from unknown source locations with dispersive properties in the material being measured, while still allowing the system to operate in real time.

[0077] In some embodiments, arrays using the systems and methods herein may be able to produce these SNR gains covering many angles of the source location, thereby allowing a smaller array to monitor a larger body under test. For example, a pressure vessel may be maintained within a shipping container and the array may be attached only to the accessible portion, yet still be able to apply these SNR gains to sensing events originating from other portions of the pressure vessel. By taking into account the dispersion properties in the material under test, the methods herein allow an array with designed spacing to process transient elastic waves from anywhere on the body of the material under test in real time, including when a portion of the body under test is inaccessible. For example, these methods and systems can be used to require a smaller number of sensor arrays and / or arrays that can be used on the material under test without physically contacting a portion of the body of the material.

[0078] The patent description and drawings are illustrative and should not be construed as limiting. It is clear that many modifications and variations may be made to the embodiments by those skilled in the art without departing from the spirit of the novel technology of the present disclosure. Although specific parameters, device configurations, component parameters, and thresholds may have been disclosed, other reference points may also be used. These modifications and variations do not depart from the broader spirit and scope of the present disclosure, and the examples cited herein are illustrative rather than limiting.

Claims

1. A system for real-time processing of signals from a multi-element transducer array, the system comprising: a multi-element transducer array (11) for attachment to a body of material under test, the body being configured to transmit a plurality of broadband signals representing transient elastic broadband vibrations propagating in the body of material under test and containing frequencies in a frequency band of interest of the transient elastic broadband vibrations, the frequency band of interest including the slowest moving wave component of the transient elastic broadband vibrations; wherein the multi-element transducer array is configured to have a maximum distance between the most distant elements of the multi-element transducer array, the maximum distance being small enough to capture a transit of a wave of the slowest moving wave component of the transient elastic broadband vibration within a transit time window; wherein the transit time window is less than a positive processing overlap time, the positive processing overlap time comprising a time of a broadband signal that can be processed in real time by the system in a first time window and a second time window immediately following the first time window of the broadband signal; a receiver circuit configured to process the plurality of broadband signals received from the multi-element transducer array, the receiver circuit further comprising: a first time coherence circuit configured to calculate a first set of a plurality of time delays between a first reference wideband signal in the plurality of wideband signals and other wideband signals in the plurality of wideband signals, so that when the plurality of wideband signals are respectively time-shifted by a plurality of time delays, a plurality of time-shifted wideband signals that are time-coherent with each other are generated; a second time coherence circuit configured to calculate a second plurality of time delays between the first reference wideband signal and remaining wideband signals of the plurality of wideband signals for comparison with the first plurality of time delays to evaluate a time coherence calculation, a combining circuit configured to generate a combined wideband signal based on a sum of the plurality of time-shifted wideband signals; and a threshold determination circuit configured to determine which values ​​of the combined wideband signal to store based on the value of the combined wideband signal being above the threshold when none of the plurality of time-shifted wideband signals has a value exceeding a threshold; and A memory is configured to receive data from the receiver circuit and store the value determined by the threshold determination circuit.

2. The system according to claim 1, wherein: The second time coherence circuit uses a different coherence determination algorithm than the first time coherence circuit.

3. The system according to claim 2, wherein: A comparison of the first set of time delays and the second set of time delays determines whether the temporal coherence calculation of the first set of time delays is consistent or inconsistent.

4. The system according to claim 1, wherein: The multi-element transducer array can be configured in two dimensions so that source direction information in a full 360° field can be unambiguously obtained.

5. A method for real-time processing of signals from a multi-element transducer array, the method comprising: attaching a multi-element transducer array to a body of material under test, the body being configured to transmit a plurality of broadband signals representing transient elastic broadband vibrations propagating in the body of material under test and containing frequencies in a frequency band of interest of the transient elastic broadband vibrations, the frequency band of interest including the slowest moving wave component of the transient elastic broadband vibrations; wherein the multi-element transducer array is configured to have a maximum distance between the most distant elements of the multi-element transducer array, the maximum distance being small enough to capture a transit of a wave of the slowest moving wave component of the transient elastic broadband vibration within a transit time window; wherein the transit time window is less than a positive processing overlap time, the positive processing overlap time comprising a time of the broadband signal that can be processed in real time by the system in a first time window and a second time window immediately following the first time window of the broadband signal; processing the plurality of broadband signals received from the multi-element transducer array, comprising: In a first time coherence circuit, a first set of time delays between a first reference wideband signal in the plurality of wideband signals and other wideband signals in the plurality of wideband signals are calculated so that when the plurality of wideband signals are respectively time-shifted by the plurality of time delays, a plurality of time-shifted wideband signals that are time-coherent with each other are generated; calculating, in a second time coherence circuit, a second set of time delays between the first reference wideband signal and remaining wideband signals in the plurality of wideband signals and comparing the first set of time delays to the second set of time delays to evaluate a time coherence calculation, generating a combined wideband signal based on a sum of the plurality of time-shifted wideband signals; and determining which values ​​of the combined wideband signal to store based on the value of the combined wideband signal being above the threshold when none of the plurality of time-shifted wideband signals has a value exceeding a threshold; and Values ​​determined to be above the threshold are stored in a memory.

6. The method according to claim 5, wherein: The second set of time delays is calculated using a coherence determination algorithm that is different from an algorithm used to calculate the first set of time delays.

7. The method according to claim 5, wherein: Comparison of the first set of time delays and the second set of time delays determines agreement or disagreement with temporal coherence calculations of the first set of time delays.

8. The method according to claim 5, wherein: The multi-element transducer array can be configured in two dimensions so that source direction information in a full 360° field can be unambiguously obtained.

Images