Monitoring structures
This invention is concerned with both an apparatus and a method for the monitoring of structures. It relates, in particular, to structure- condition remote monitoring apparatus, and to such apparatus which is for the remote detection of incipient failures in a structure composed of a framework of large members joined together.
Offshore oil and gas exploration and production Companies maintain large open-lattice structures that are mostly underwater. To extend the design life of such a structure there is required some method to monitor it so as to give advance warning of faults in its structural members well before any such structural member should fail. This method needs: to assess the whole structure on a regular and frequent basis; to give warning before failure of a member is complete; to operate in a manner that keeps costs to a minimum; and to be reliable, not missing any member failure and not giving false alarms. The invention proposes such a method.
There already exist several methods to monitor offshore structures, some of which are actually in use. For example, most offshore steel structures have hollow structural members designed to be watertight, and obviously detecting the flooding of a member can give early indication of a crack. Unfortunately, the detection of water inside such members involves expensive and hazardous diver intervention, and so reduces the frequency at which such inspections are carried out. And the alternative of fitting permanent sensors to every member is very costly.
Other methods that have been used or proposed for use include the following:-
Growing cracks are known to emit noise in the range of 100kHz, and acoustic emission is a method of listening for such noise. A large number of sensors has to be deployed for this method to have any chance of working.
Water waves subject offshore structures to excitation. The response of the structure to the excitation can change if a member cracks. The frequency input from the waves is generally below 1Hz, and local member structural response is well above that. The force input cannot be measured, so the only parameters to compare are frequency and damping, not magnitude or phase of response. The overall structure frequency response in the lowest modes may be closer to the excitation frequency, but the method can be insensitive even to complete severance of a single member. Offshore structures are deliberately designed to have redundant numbers of members; a single member can be removed without the structure collapsing. Such redundancy makes the response to water wave excitation highly insensitive to failure of individual members. The method is thus of limited use.
Ultrasonic reflection techniques that are used on land to assess pipelines and similar structures can be used underwater. The presence of water is an advantage, as it improves the coupling between the transducers and the steel. The high frequencies used (over 100kHz) give good resolution in detecting defects but limit the range because of rapid attenuation of the waves. Every weld has to be individually tested, and the steelwork has to be carefully cleaned by divers. This is a lengthy, dangerous and expensive operation.
Ultrasonic refraction wave techniques, sometimes called Lamb waves, work in the 50kHz to 1MHz range, well beyond the human audible range. The waves travel along the wall of cylindrical structures with a range of a few metres. The possible operating coverage along the structure can be ten or twenty metres. The high frequency gives the opportunity for good defect resolution. The input excitation is in the form of a few cycles of a single frequency pulse to reduce the effect of dispersion of the waves. The frequency of the pulse is chosen to optimize the detection. The method could be sensitive to changes in the mass supported by the structure. The input pulse has to be exactly repeated so that the response waveforms can be time-averaged to reduce the effect of background noise. Again, this is a hazardous and expensive operation involving divers, although fewer access points
are required than is the case with ultrasonic reflection techniques.
The apparatus and method proposed by the present invention have considerable advantages over existing suggestions. These advantages are achieved by introducing into the structure to be tested, at several points thereon, vibration waves in the acoustic frequency range, and then monitoring the response to these waves simultaneously at a limited number of points on the structure. The method involves separating out the induced and resulting compression waves and bending waves, and determining the impedance spectrum between each input and each measurement point (the ratio of the Fourier transform of force input at one point to the Fourier transform of acceleration measured at another point is the impedance spectrum between the two points). From the inverse Fourier transform of the impedance spectrum there is obtained the response due to a unit impulsive force - known as the unit impulse response - in the time domain. There may be several paths between the two points along which waves may travel; a defect such as a crack that comes into being along one of those paths will alter the measured unit impulse response (UIR), and thus tracking the compression wave UIR and bending wave UIR between several points allows any such defect to be located on the structure.
Changes in the supported mass can also cause changes in the measured unit impulse response. Since such changes are not of interest in this context, the invention provides means of preventing them.
In one aspect, therefore, the invention provides apparatus usable to detect the origin, location and growth of defects in a large structure composed of a framework of members, which apparatus comprises: a plurality of forcing means mountable on the structure and which can introduce force into the structure in the acoustic frequency range, each actuator being associated with a load sensor to measure the force it introduces; a plurality of response measurement means mountable on the structure and which can measure the bending wave and compression wave response at any cross-section of the structure; control means to control the sequential action of the forcing means and the time history of force they introduce;
recording means to capture the measured forces and responses, and store these; computing means to form the frequency-domain impedance spectrum ratio of forces and responses, to correct the impedance ratios or any changes in supported mass, and to transform the frequency-domain impedance spectrum ratios into time-domain unit impulse responses; and comparison means to compare the latest set of unit impulse responses with a previously-measured set, and to relate any changes between the two sets to the growth of defects on the structure.
Most preferably the apparatus also includes alarm means for providing a warning to the Operator of the existence of a potential failure of a member, and the location of the potential failure point.
In a second aspect the invention provides a method, using the thus-defined apparatus of the invention, to detect the origin, location and growth of defects in a large structure, in which method: the forcing means and the response measurement means are suitably distributed around the structure, at forcing points and response points respectively, and activated to produce and record bending and compression waves which are stored for future recall together with the applied forces; the bending and compression wave impedance spectra between every forcing point and every response point are calculated by taking the ratio of the Fourier transform of force input at one force point to the Fourier transform of vibration measured at the response point, and the wave impedances are corrected if necessary for changes in the mass supported by the structure; the wave impedance spectra are converted into the time domain by an inverse Fourier Transform, to provide what are referred to as unit impulse time responses; and the various unit impulse time responses are compared with original values obtained on the intact structure, and any changes are related to the size and location of defects in the structure.
And optionally, of course, a warning is provided of any defects that are growing with time, and an indication given of their location.
The force may be input by various means, including both mechanical and electromechanical devices. In the latter case the time history of the force input - the way it changes over time - can conveniently be chosen to be a pulse, or a sinusoid whose frequency sweeps through the acoustic range, or even a random signal. Devices which input a continuous wave are referred to herein as "shakers"; their input/output can be a sine wave that rapidly and steadily increases in frequency from a few hertz up to a few kilohertz (such an excitation is known as a "chirp", from the sound it makes). Various piezo-electric and piezo-resistive materials such as PZT (lead zirconate titanate) and PVDF (polyvinylidene difluoride) can be used as electromechanical devices to excite the structure in the chosen frequency range. These have the attraction of being robust and relatively cheap to apply, but the difficulty with them lies in calibrating them so that the force is known. Other force input devices include hammers (suitably instrumented with force sensors to measure impulses introduced by each hammer blow).
The forcing means are preferably arranged in groups at each relevant loading point on the structure - for example, three or four forcing means are conveniently arranged symmetrically around the chosen component (such as leg or a bracing member, beam or girder). Moreover, advantageously at each loading point the forcing means actuators in the group are so disposed that the resultant force is axial to the loaded member.
The response measurement means may take any convenient form, but the response is most conveniently measured by the use of accelerometers (arranged in groups) producing an electrical output (though accelerometers that produce a variation in a light signal passed down optical fibre cable can be employed). One or more response measurement means can be placed at a given cross-section of the structure, their responses then combined to separate out the bending
and compression waves. This separation can be done either in situ or remotely (by post processing).
Notionally, any force introduced into the structure for the purpose of the application of the invention can be a vibration of almost any frequency. Nevertheless, by choosing to limit the frequency of the generated waves to the human audible acoustic range - from just below 20kHz down to around 100Hz - which are less significantly attenuated than are very short wavelength ultrasound waves, the invention extends the reach of the waves to hundreds of metres instead of merely tens of metres. However, the penalty paid for lower attenuation, and therefore greater reach, is poorer resolution of defects (for the wavelength of waves within this frequency range is of much the same order as the size of some of the defects to be found, and so the waves may travel past the smaller cracks without significant interaction). Nevertheless, bearing in mind the size of the structural member used in, say, oil rigs - a leg or other structural member may be a metre across, and a hundred metres or so long - the invention's apparatus is able to detect cracks extending over at least twenty per cent of the member cross- section (that is, about 20cms long). It will not detect small defects of the order of size that typical ultrasonic methods will find - a few centimetres or so - but the prime objective is to detect cracks which are large enough to be a threat to structural integrity and early enough to be able to repair them before complete severance occurs.
A large number of measurements can be repeated to average out random background noise.
Structures that are a few hundred metres in height may be able to be excited by forcing means - shakers in particular - that are all above the water line. This will have the advantage of reducing the cost of their installation. In these instances the excitation would be near the top of each of the main members, and there will be more response measuring (accelerometer) groups than forcing means (actuator) groups. Taller structures, or structures with several gross changes in cross- sectional area down the main members, may need to be excited at some additional points below the water line. These additional points of excitation are likely to be on the main members. With large members it will be difficult to excite compression waves with a single forcing means.
In principle this does not matter for suitable excitation can be achieved simply by mounting three or more forcing means at one cross-section at equal distances around the member's circumference. Each forcing means is excited in turn, and the sum of the impedances formed from each in turn will be the impedance due to a compression wave alone. Alternatively, it is feasible to excite all the forcing means at one location together, and in phase, so that the net result is that of an excitation along the centre line of the pipe.
The response and force points on the structure are reciprocal, so that impedances can equally well be obtained by reversing the force and response points. The implication is that it is possible to have the response points above the water line, and use remotely operated vehicles (ROV) to apply forces underwater. The limitation in this approach is that the forces are less well controlled, and since good control is a preferred aspect of the invention, this reverse configuration is not preferred.
The number of response points will depend upon the design of the structure, but it is expected that only of the order of one tenth of the members will need to have sensors mounted on them. Numerical analysis is performed to decide on the required density of coverage of the structure with response points to ensure defects can be located with the required accuracy.
At each location for measurement of response there will be a group of response measurement means - sensors - chosen so as easily to separate out the bending and compression waves. For example, there could be four accelerometer sensors diametrically opposite each other at one cross-section and measuring in an axial direction along the member. The addition of all four responses will subtract out the bending waves, leaving only the compression wave response. The subtraction of opposite accelerometers will remove the compression waves, leaving only the bending wave response in the two planes. The use of both types of wave is important because their response to a fracture will differ, which increases the information available when seeking to identify the location of the fracture.
The term "impedance spectrum" has been defined as the ratio of the Fourier transform of force input at one point to the Fourier transform of resulting acceleration measured at another point. If the resulting frequency-domain impedance spectrum is transformed back into the time domain the result is the response to a unit impulse. Any variation in magnitude, phase or frequency content of the input force excitation is removed by the operation. Either the unit impulse response can be obtained in repeated excitations, and the result averaged in time, or the frequency-domain impedance spectrum from repeated excitations can be averaged In the frequency domain before transforming into the time domain. The consequence will be that random noise will be averaged out. The unit impulse response consists of the pulse that travels along the direct path from excitation to response, followed by a succession of pulses that travel by other paths. If one of the paths should be removed by complete severance of a member then the corresponding arrival pulse from that path will disappear. Equally, if the member is not severed but instead is fractured part way through then the corresponding arrival pulse will be reduced in magnitude, and there will be a new arrival pulse corresponding to the new path formed by reflection from the fracture, and its time of arrival is then known.
The purpose of the structure is to support equipment whose weight is borne by the structure. If the mass of that equipment should vary significantly then the measured impedances can be altered. It is proposed to remove the effect of such changes in supported mass by installing, at points as close as possible to the junction between the supported mass and the main structure, additional forcing means and response measurement means groups. These form units that can be actively controlled, as described in "Active Control of Vibration" by C.R.Fuller, S.J.Elliot 6k P.A.Nelson, Academic Press 1996, to ensure that the motion in these response measurement groups is held at zero. In doing so, the impedance at these points is being constrained to be that of a perfect reflector, and so the influence of the mass beyond the reflector is no longer observed in response measurement units on the main structure.
The active control can be applied in real time, while other forcing means groups are operated, or else virtual active control can be applied.
With virtual active control the additional forcing means groups and additional response measurement groups are operated separately in the same manner as all the others. It can then be calculated what drive signal at the additional forcing means groups would hold the response at the additional response measurement groups to zero. These additional responses are then added to the other measured responses.
The invention involves the use of: control means to control the sequential action of the forcing means and the time history of force they introduce; recording means to capture the measured forces and responses, and store these; of computing means, to form the frequency- domain impedance spectrum ratio of forces and responses, to correct the impedance ratios for any changes in supported mass, and to transform the frequency- omain impedance spectrum ratios into time-domain unit impulse responses; and of comparison means, to compare the latest set of unit impulse responses with a previously-measured set, and to relate any changes between the two sets to the growth of defects on the structure. These control, recording, computing and comparison means, and the way they operate, and now described in more detail.
In operation the shakers in each actuator group are each driven in turn by a power amplifier. The power amplifier is excited by a wave function generator that preferably produces a sine wave that rapidly and steadily increases in frequency from a few hertz up to a few kilohertz (the wave function generator may instead produce a pulse whose width is of the order of a few milliseconds). The control of these actuator groups consists of switching between each group to ensure that each operates in turn and with sufficient delay between each to ensure that all responses to the previous actuation have completely died away. This control is to ensure that each transmitted wave function occurs at the same instant that the response measurements are taken; this synchronisation is important for eliminating small timing errors in the subsequent analysis. The sequential operation of the actuator groups and synchronous excitation and response measurement can be taken care of by a standard PC interface card on a computer, as can the wave function generation. Every input force and every response is measured and converted into digital format using a proprietary analog- to-digital converter with an anti-alias filter prior to the conversion to
digital form.. The digital signals are converted by Fourier transform into the frequency domain. The impedance at every response point relative to every input forcing point is formed as the ratio of input force to response. The impedances formed as a result of repeated excitations at the same point are averaged to reduce background noise. These averaged impedances are inverse-Fourier transformed back into the time domain to obtain the unit impulse response which is recorded in digital format on the hard disc of the same computer that generated the excitations.
Finally, the invention makes use of alarm means for providing a warning to the Operator of the existence of a potential failure of a member, and the location of the potential failure point.
A complete set of unit impulse responses is stored at some reference time, perhaps early in the life of the structure. Subsequent unit impulse responses are stored at later dates. The reference impulse responses are compared with the subsequent ones by taking the differences. The power of these difference signals is formed by taking the mean square of the difference signal, and checked in the computer to assess if it exceeds a minimal level that might be expected as a result of measurement errors and noise in the system. If the minimal level is exceeded then that will be taken as an indication of failure in the structure, and a message produced on the computer screen. Further timing analysis of the difference signals, and a path computation on the rig layout, will indicate the location of the failure. This will also be given as an alert on the computer screen.
An embodiment of the invention is now described, though by way of illustration only, with reference to the accompanying diagrammatic Drawings in which:
Figure 1 shows how in a semi-infinite continuous homogeneous medium one sub-surface sensor and two surface excitations can identify the position of a defect;
Figure 2 shows some example response measurements;
Figure 3 shows the overall structure standing in water with attachments of forcing means (shaker) groups and response measurement means (accelerometer) groups;
Figure 4 shows the attachment of three electromechanical shakers 1 to the top of a main member of the structure;
Figure 5 shows the attachment of accelerometers 4 to one of the members of the structure;
Figure 6 shows the attachment of an additional forcing means (actuator) group and an additional response measurement means group to a structural member near the junction between the supported mass and the main structure; and
Figure 7 shows in block diagram form the various components utilised for this invention.
To illustrate how detection of such changes in pulse arrivals at only a small number of points on the structure can determine the exact location of the defect, consider the method applied to a semi-infinite continuous medium such as the earth, and illustrated in Figure 1 (in which is shown the surface (6) of the earth (23). Two surface excitation points (5,32) and a single sub-surface sensor (9) can reduce the indeterrninacy in the location of a defect (7) within the body to just two
possible positions. A second sub-surface sensor, or a third excitation point, can then give a unique location of the defect.
The damage point time of arrival is known, and for one excitation point 5 and one response point 9 the damage must lie on an ellipse (8) formed with the excitation point 5 and response point 9 as foci and the path length known from the time of arrival. The location is then known to be on that ellipse. With a second excitation point 32 a second ellipse (30) can be drawn, and the damage point (31) will lie where the two ellipses intersect.
A real open-lattice structure is more complicated, having wraparound symmetries and other complications, but the principle remains the same.
The means whereby the complete severance of a member can be detected from the unit impulse response is illustrated in Figure 2, taken from tests on a scale model structure in which is shown the undamaged unit impulse response (heavy line 11), the response (dotted line 12) after a cut around 50% of the perimeter, the response (dashed line 13) after a cut around 90% of the perimeter, and the response (light line 14) after complete removal of the member, on the path of the first direct arrival pulse. The first direct arrival pulse progressively reduces in magnitude, and disappears when completely removed. The next arrival pulse is more complex, due to a bending path, and progressively changes as the damage is increased. The third arrival is a second direct path, and progressively changes with increasing damage.
A specific application of the invention is now described with reference to Figures 3, 4, 5 and 7.
In Figure 3 is shown the overall structure standing in water (21) with attachments of forcing means shaker groups (15) and response measurement means accelerometer groups (17); the shaker groups have coincident accelerometer groups.
The top mass (16) is supported by the cross-braced structure, which has main leg members (as 19), which are usually the most significant members, and close to vertical. There are usually horizontal members (as 18) and other diagonal cross-braces (as 20). There will
usually be other horizontal members and cross-braces in the unrepresented third dimension of the Figure.
The whole structure is in this case supported within the seabed (22). The shaker groups 15 are close to the top of the main members, and above the water line. The accelerometer groups 17 are mounted on the main members below every level of horizontal braces. In real life the exact distribution of shaker groups and accelerometer groups will depend on the exact geometry of the structure.
In Figure 4 is shown (in elevation and in section) the attachment of three electromechanical shakers (as 1: typically a Gearing & Watson GWV46 inertial shaker) to the top of a main member 19 of the structure. The shakers illustrated are inertial, and so react against their own mass; they act on the main member 19 through a stinger (3) pushing against a pad (2) on the member. Three shakers 1 are shown equally disposed around the member 19, so that equal in-phase forces can be applied to generate a compressive wave in the member. At the point of input 2 of each force is also measured the response (with an accelerometer 29; typically a Monitran 1100 series device).
Figure 5 shows (in elevation and in section) the attachment of accelerometer sensors (4) to one of the members 19 of the structure. The sensors 4 are shown in a group of four equally disposed around the circumference of the member 19. This arrangement allows the addition of all sensor responses to obtain the in-line component of acceleration, and the difference of opposite pairs of sensors to obtain the rotational acceleration about two planes at right angles to each other.
In Figure 7 is shown how the shakers 1 are driven, and how the response accelerations 4 are combined with input forces to form the response impedances.
In operation the shakers in each actuator group are each driven by a power amplifier (36; typically a Harrison Information Technology X150). The power amplifiers 36 are themselves excited by a sine wave generator (35; typically a Hewlett Packard 3314A function generator) that rapidly and steadily increases in frequency from a few hertz up to a few kilohertz. Such an excitation is known as a chirp, or swept sine wave.
The data from every input force inputter 1 and every response measurement means 29,4 is converted into digital format and
Fourier-transformed into the frequency domain (fft) by a converter device (37; typically a National Instruments DAQ-1200 analog-to-digital [adc] converter PCMCIA card) that forms the input device to a standard computer (such as the Compaq Armada 1530) and using a standard interface driver control software (such as National Instruments Lab View). The impedance (38) at every response point relative to every input force is formed as the ratio of input force to response.
The impedances formed as a result of repeated excitations at the same point are averaged (39) to reduce background noise. These averaged impedances are inverse-Fourier transformed (ifft) back into the time domain (41) to obtain the unit impulse response, which is recorded in digital format on the hard disc of the computer (40). Subsequent analysis (42) takes place in the computer to check if any changes have occurred in the unit impulse responses. If changes have occurred then this fact, and the location of the failure, are indicated on the alarm (43) which may be a message on the computer screen.
The impedances are corrected for any change in supported mass 16 in the following manner. As shown in Figure 6 (in both elevation and section), additional forcing means actuator (33) and response measurement means response (34) groups are provided at, or close to, the points of attachment of the support structure to the supported mass 16. The additional actuator groups 33 are driven at the same time as the actuator groups 15 in such a manner that there is no net motion at the additional response groups 34 (this constrains the cross- sections at points of attachment of the support structure to the supported mass 16 to be perfect reflectors). Thus, since the points of attachment to the mass 16 are kept stationary, changes to the mass 16 can have no influence on the behaviour of the structure.
By using directional response groups 33 at the points of attachment of the support structure to the supported mass 16 it is possible to hold the impedance of the cross-section at the point of attachment of the support structure to the supported mass 16 to be any other constant value, of which a perfect absorber would be one choice. Directional response groups have two groups of sensors in close proximity. This increases the amount of hardware and complexity of the system. Using directional additional support groups is not a preferred method.
It is not strictly necessary to drive the additional forcing means actuator groups 33 at the same time as the actuator groups 15. By operating each actuator in each additional actuator group 33 in the same manner as other actuator groups 15, and measuring the response in each response measurement means sensor of each additional sensor/response group 34 in the same manner as the other response groups 17, it may be determined what drive signals would have cancelled all motion at the additional response groups 34. By determining the motions produced at all other response groups 17 by these required additional drive signals and summing these with the motions produced by the actuator groups 15, the same result of active control can be achieved by post-processing.
How to achieve this active control by post-processing is now illustrated as follows. Suppose Hu(w) and Hto(w) represent the frequency domain impedances at response point i due to applied unit impulses Iι(w) and Io(w) at points 1 and 0 respectively. Point 1 is taken to be the location of the main shaker group and point 0 is taken to be the location of the additional shaker group that is closer to the supported top mass of the structure. Then a force
at location 0 will produce a response at the same location that equals the response at this location due to a unit force at location 1. So the difference of these two responses would imply that location 0 did not move. Therefore the corrected impedances at every point on the structure would be
The inverse Fourier transform of this corrected impedance hcι o (t) will be the true unit impulse response that is independent of changes in top mass.
The impedances formed as a result of three or more excitations in a group on the same member at one cross-section are averaged to obtain the impedance due to a force input in-line with the main member. So the impedances formed at point 4 due to excitations at 1, 25 and 26 are averaged. The impedances formed at four response points at one cross-
section are averaged to form the impedance corresponding to a wave reaching that cross-section in-line with the member at the response group. Calling these the compressive impedances, the impedances at points 4, 24, 27 and 28, for example, can be averaged to form the compressive impedance at cross-section 4-24-27-28. The impedances of pairs of sensors in the same group that are opposite each other are subtracted to form the impedance due to bending waves reaching that cross-section. Respective responses from the other pair of sensors in the group are subtracted to form the bending wave impedance about the other axis. So, the impedances at points 4 and 24, for example, are subtracted to form the bending wave impedance about the line orthogonal to 4-24, and the impedances at points 27 and 28, for instance, are subtracted to form the bending wave impedance about the line orthogonal to 27-28.
The impedance spectra are Fourier-transformed back into the time domain to form the unit impulse response functions. The time of arrival and magnitude of waves in the impulse response function are compared with corresponding quantities in the impulse response functions taken originally on the structure when it was presumed to be undamaged. The changes in these quantities in the various impulse response functions are used to detect the onset of failure and its location based on the known paths through the structure. The measure of this damage could, for example, be the total power in the difference signal, with zero power indicating no damage at all.
Warning is given to the operator in the event that any failure is detected. Subsequent to failure being detected and repaired, a new set of reference undamaged impulse responses are taken from which to make later comparisons.