Clamp release tool and method

Abstract

A method of operating a pre-opened pre-positioned clamp, the clamp being operable between an open and a closed position. The method comprises releasing the clamp from the open position using a tool head (201), making measurements of a movement of the tool head resulting from the releasing, recording data relating to the measurements, determining a frequency content of the data, and, on the basis of the determined frequency content, determining whether the clamp has moved to the closed position.

Claims

1. A method of operating a pre-opened pre-positioned clamp, the clamp being operable between an open and a closed position, and the method comprising: releasing the clamp from the open position using a tool head; making measurements using an accelerometer and a gyroscope of a movement of the tool head resulting from the releasing; recording data relating to the measurements; determining a frequency content of the data; and on the basis of the determined frequency content, determining whether the clamp has moved to the closed position.

2. The method of claim 1, wherein making measurements of a movement comprises measuring movement in respect of all three planes of motion, and rotation in respect of pitch, roll, and yaw.

3. The method of claim 1, wherein the movement comprises vibration.

4. The method of claim 1, wherein the frequency content comprises content concerning one or more frequencies that are characteristic of the clamp having moved to the closed position.

5. The method of claim 1, further comprising, on the basis of the measured movement, determining that a trigger event has occurred.

6. The method of claim 5, wherein the determining that a trigger event has occurred comprises detection of a significant oscillation, wherein a significant oscillation comprises at least one of an acceleration of more than a threshold, the threshold being greater than 500 ms.sup.?2, and a rate of change of acceleration of more than a threshold, the threshold being greater than 500 ms.sup.?3.

7. The method of claim 6, wherein the trigger event comprises detection of at least four consecutive significant oscillations.

8. The method of claim 5, wherein the recorded data comprises data defining movement preceding the trigger event and data defining movement following the trigger event.

9. The method of claim 1, wherein determining whether the clamp has moved to a closed position comprises determining a change in orientation of the clamp release tool.

10. The method of claim 1, wherein determining whether the clamp has moved to a closed position comprises calculating a plurality of parameters of, or derived from, the frequency content of the data and using those parameters to evaluate whether the clamp has moved to a closed position, the plurality of parameters comprising at least one of the following: a correlation coefficient, a diversity index, a kurtosis measure, a skewness measure, the magnitude of the maximum peak in the frequency domain, the frequency of the maximum peak in the frequency domain, and the magnitude and/or frequency of at least one other frequency peak.

11. The method of claim 1, wherein determining whether the clamp has moved to a closed position comprises calculating a plurality of parameters of, or derived from, the recorded data and using those parameters to evaluate whether the clamp has moved to a closed position, the plurality of parameters comprising two or more different parameters from a group consisting of: a correlation coefficient, a diversity index, a count of significant oscillations, an average rate of change of the significant oscillations, a rate of change of acceleration, a rate of change of angular velocity, a kurtosis measure, a skewness measure, a maximum value of one or more of acceleration, angular velocity, rate of change of acceleration, rate of change of angular velocity, a minimum value of one or more of acceleration, angular velocity, rate of change of acceleration, rate of change of angular velocity, an average value of one or more of acceleration, angular velocity, rate of change of acceleration, rate of change of angular velocity, and a first quartile and/or a third quartile value of one or more of acceleration, angular velocity, rate of change of acceleration, rate of change of angular velocity.

12. The method of claim 1, wherein determining whether the clamp has moved to a closed position comprises applying one or more rules, the one or more rules defining the characteristics of a movement deemed to be one corresponding to the clamp moving to the closed position.

13. The method of claim 1, wherein determining whether the clamp has moved to the closed position comprises evaluating the output of a sliding window algorithm, the sliding window algorithm comprising calculating a ratio of an amplitude of a first frequency band to an amplitude of a second frequency band.

14. The method of claim 13, wherein the sliding window algorithm comprises calculating at least ten ratios.

15. The method of claim 1, wherein determining whether the clamp has moved to the closed position comprises a positive detection of a movement corresponding to the clamp moving to the closed position.

16. A clamp attachment tool for operating a pre-opened pre-positioned clamp, the clamp being operable between an open position and a closed position, and the tool comprising: a tool head for releasing a clamp from the open position; inertial measurement electronics comprising an accelerometer and a gyroscope arranged to make measurements of a movement of the tool head; and signal processing electronics, configured to record data relating to the measurements, determine a frequency content of the data, and, on the basis of the determined frequency content, determine whether the clamp has moved to the closed position.

17. The clamp attachment tool according to claim 16, wherein the clamp is suitable for clamping a hose to a spigot.

18. The clamp attachment tool of claim 17, wherein: the clamp in the open position is held at an expanded diameter, greater than an external diameter of the hose; and the clamp in the closed position is positioned around the hose at substantially the same diameter as the external diameter of the hose; and the clamp is resiliently deformable, such that the clamp in the closed position exerts a compressive force on the hose, clamping the hose to the spigot.

19. A method of manufacturing signal processing electronics for a clamp attachment tool, the tool being suitable for operating a pre-opened pre-positioned clamp, the tool comprising a tool head arranged to engage with the clamp and inertial measurement electronics comprising an accelerometer and a gyroscope arranged to take measurements of a movement of the tool head, the clamp being operable between an open position and a closed position, and the method comprising: providing signal processing electronics, arranged to receive a signal corresponding to the measurements; programming the signal processing electronics to record data relating to the measurements, determine a frequency content of the data and, on the basis of the determined frequency content, determine whether the clamp has moved to the closed position.

20. The method of claim 19, further comprising operating a machine learning algorithm to determined one or more rules for classifying a detected movement as either corresponding to the clamp having moved to the closed position or as not, the one or more rules defining the characteristics of a movement deemed to be one corresponding to the clamp moving to the closed position; and wherein determining whether the clamp has moved to the closed position comprises applying the one or more rules.

Description

DESCRIPTION OF THE DRAWINGS

(1) Embodiments of the present invention will now be described by way of example only with reference to the accompanying schematic drawings of which:

(2) FIG. 1 shows a perspective view of a typical pre-opened pre-positioned clamp of the prior art;

(3) FIG. 2 shows a perspective view of a clamp release tool according to a first embodiment of the invention;

(4) FIG. 3 shows a block diagram of an electronics package of a clamp release tool of FIG. 2;

(5) FIG. 4 shows a flow chart illustrating the steps of a method of operating a pre-opened pre-positioned clamp according to a second embodiment of the invention; and

(6) FIG. 5 shows a flow chart illustrating the steps of a method of manufacturing a clamp release tool according to a third embodiment of the invention.

DETAILED DESCRIPTION

(7) FIG. 2 shows a perspective view of a Pre-Opened Pre-Positioned (POPP) clamp release tool 200 according to a first embodiment of the invention.

(8) Clamp release tool 200 comprises a tool head 201 and a handle 203. Tool head 201 comprises three prongs, each of which is arranged to be received in recess 107 of POPP clamp 100 such that tool head 201 is suitable for releasing a POPP clamp 100 from the open position. Alternative embodiments of the invention may comprise a different number of prongs such as one or two prongs, or possible more prongs. Incorporating the three prongs enables clamp release tool 200 to be engage with clamp 100 from three different angles, providing an operator of clamp release tool 200 a number of potential angles for using the tool. This is particularly beneficial when clamps are used in confined spaces, as is often the case for automotive applications. Clamp release tool 200 further comprises an electronics package 300 contained within handle 203.

(9) FIG. 3 shows a block diagram of electronics package 300. Electronics package 300 comprises an inertial measurement unit 301 and is mounted to tool head 201 such that a movement of tool head 201 results in a corresponding movement of electronics package 300, and therefore of inertial measurement unit 301.

(10) Inertial measurement unit 301 comprises three accelerometers and three gyroscopes, arranged such that inertial measurement unit 301 can detect movement, and particularly vibration, in respect of all three planes of motion and in respect of pitch, roll, and yaw. Inertial measurement unit 301 can therefore be said to comprise a six-axis gyroscope. Alternative embodiments of the invention may comprise fewer (i.e. one or two) accelerometers and/or fewer (i.e. one or two) gyroscopes. Inertial measurement unit 301 is configured to transmit a signal corresponding to the detected movement to a signal processing unit 303 of electronics package 300. The signal from inertial measurement unit 301 can be said to comprise inertial data of inertial measurement unit 301. In this exemplary embodiment, inertial measurement unit 301 detects movement and generates a full set of inertial data, comprising data from each of the accelerometers and gyroscopes, at a rate of 6.667 kHz. The inertial data therefore comprises data defining an acceleration of each of the accelerometers in the respective directions of detection and data defining an angular rate of each of the gyroscopes about their respective axes of detection (speed of rotation about each such axisi.e. angular velocity).

(11) Signal processing unit 303 is configured to process the inertial data to determine whether a detected movement corresponds to the successful release of a POPP clamp. To do so, signal processing unit 303 is configured to monitor the inertial data for a trigger event. The trigger event comprises a movement containing at least four significant oscillations, wherein a significant oscillation is one that corresponds to a rate of change in acceleration in excess of 100 gs.sup.?1 (i.e. 980 ms.sup.?3). The monitoring therefore comprises checking the inertial data for four consecutive oscillations which meet the following criteria:
Abs(ACX[i]?ACX[i?1])>0.147 ms.sup.?2(1)
Where: ACX=acceleration in the x-direction.

(12) Signal processing unit 303 is further configured to, in response to detection of the trigger event, log inertial data corresponding to 0.45 seconds of time. Specifically, signal processing unit 303 is configured to log inertial data for the 0.27 seconds immediately preceding the trigger event (pre-trigger inertial data) and for the 0.18 seconds immediately following the trigger event (post-trigger inertial data). In this exemplary embodiment, inertial data is acquired at a rate of 6.667 kHz, therefore the preceding 0.27 seconds corresponds to 1800 logged data points and the following 0.18 seconds corresponds to 1200 logged data points. The signal processing unit therefore logs six arrays, each of 3000 data points, and corresponding to one array for each of the three accelerometers and three gyroscopes.

(13) Signal processing unit 303 is further configured to calculate, on the basis of the logged inertial data, a change in orientation of clamp release tool 200. This is calculated by use of the following equations:

(14) $\begin{array}{cc}\mathrm{RollGY}\left[j\right]={?}_{i=0}^{i=j}\mathrm{GYX}\mathrm{dt}=\underset{i=0}{\overset{i=j}{.Math.}}\frac{\mathrm{GYX}\left(i\right)}{\mathrm{dt}}& \left(2\right)\\ \mathrm{PitchGY}\left[j\right]={?}_{i=0}^{i=j}\mathrm{GYY}\mathrm{dt}=\underset{i=0}{\overset{i=j}{.Math.}}\frac{\mathrm{GYY}\left(i\right)}{\mathrm{dt}}& \left(3\right)\\ \mathrm{Yaw}\left[j\right]={?}_{i=0}^{i=j}\mathrm{GYZ}\mathrm{dt}=\underset{i=0}{\overset{i=j}{.Math.}}\frac{\mathrm{GYZ}\left(i\right)}{\mathrm{dt}}& \left(4\right)\\ \mathrm{RollAC}\left[j\right]={\tan }^{-1}\left(\frac{\mathrm{ACX}\left[j\right]}{\mathrm{ACZ}\left[j\right]}\right)& \left(5\right)\\ \mathrm{PitchAC}\left[j\right]={\tan }^{-1}\left(\frac{\mathrm{ACY}\left[j\right]}{\sqrt{{\mathrm{ACZ}\left[j\right]}^2+{\mathrm{ACZ}\left[j\right]}^2}}\right)& \left(6\right)\end{array}$
Where: RollGY=roll as determined from the integral of the gyroscope data; PitchGY=pitch as determined from the integral of the gyroscope data; Yaw=yaw as determined from the integral of the gyroscope data; GYX=gyroscope data for rotation about the x-axis; GYY=gyroscope data for rotation about the y-axis; GYZ=gyroscope data for rotation about the z-axis; RollAC=roll as determined from the accelerometer data; PitchAC=pitch as determined from the accelerometer data; ACX=accelerometer data for acceleration in the x-direction; ACY=accelerometer data for acceleration in the y-direction; and ACZ=accelerometer data for acceleration in the z-direction.

(15) A data fusion algorithm is used to provide estimates of roll and pitch, as the results of the integration of GYX and GYY are liable to drift over time. The final estimates of roll and pitch are therefore calculated by use of the following equations:
Roll[j]=0.82*RollGY[j]+0.18*RollAC[j](7)
Pitch[j]=0.82*PitchGY[j]+0.18*PitchAC[j](8)
Where: Roll=roll as determined from the data fusion; and Pitch=pitch as determined from the data fusion.

(16) The change in orientation of clamp release tool 200 comprises a factor that may be evaluated in determining whether a detected movement corresponds to a successful release of a POP clamp.

(17) Signal processing unit 303 is further configured to determine the frequency content of the logged inertial data. Determining the frequency content comprises performing a Fourier transform on the logged inertial data, for example a Discrete Fourier Transform (DFT) or a Fast Fourier Transform (FFT). The Fourier transform is performed on the 128 data points following the trigger event, for each of the accelerometers and gyroscopes.

(18) Releasing a POPP clamp induces a vibration in the clamp release tool. A successfully released POPP clamp induces vibrations with a specific characteristic frequency distribution, comprising a single dominant frequency and a number of harmonic frequencies. This specific frequency distribution is therefore characteristic of the clamp having moved to the closed position, and is indicative that the clamp has been successfully released. The frequency distribution may comprise frequency components relating to a movement of protrusion 105 of clamp 100. The frequency distribution may comprise frequency components relating to a vibration of prong 109 of clamp 100. POPP clamps come in a range of sizes, each of which vibrates with a different characteristic frequency distribution. There are therefore a number of different characteristic frequency distributions that can indicate the successful release of a POPP clamp.

(19) The frequency content is processed by use of a sliding window algorithm. The data is first grouped into a finite number, for example 64, of bins. In this exemplary embodiment, each bin corresponds to a 52 Hz frequency range. Two windows, each of which are four bins wide in this exemplary embodiment, are then applied to the data, with the first window comprising the first four bins and the second windows comprising the second four bins. The mean of the first window is divided by the mean of the second window and the result stored. The second window is then advanced by one bin, such that it now comprises the fifth to ninth bins, and the process is repeated. Once the second bin has reached the final four bins, the first window is advanced by one bin, such that it now comprises the second to fifth bins, and the second window is moved to comprise the four bins following the first window. This process is repeated until both windows reach the end of the data set. The output of this algorithm (the DFT sliding window algorithm data) comprises a factor that may be evaluated in determining whether a detected movement corresponds to a successful release of a POPP clamp.

(20) The particular frequency bin that corresponds to the maximum peak in the frequency content also comprises a factor that may be evaluated in determining whether a detected movement corresponds to a successful release of a POPP clamp.

(21) Signal processing unit 303 is further configured to determine one or more quantities related to the arrays of logged inertial data.

(22) A first exemplary quantity comprises the Pearson Correlation Coefficient (PCC). Signal processing unit 303 is configured to calculate the PCC of pairs of the arrays of pre-trigger inertial data. PCC provides a measure of the correlation of the two sets of data, where a PCC of +1 indicates total positive linear correlation, 0 indicates no correlation, and ?1 indicates total negative correlation. The PCC of arrays X and Y are calculated by use of the following equation:

(23) $\begin{array}{cc}\mathrm{PCC}=\frac{n\underset{0}{\overset{n}{.Math.}}\mathrm{XY}-\left(\underset{0}{\overset{n}{.Math.}}X*\underset{0}{\overset{n}{.Math.}}Y\right)}{\sqrt{\left(n\underset{0}{\overset{n}{.Math.}}{X}^2-\left(\underset{0}{\overset{n}{.Math.}}X*\underset{0}{\overset{n}{.Math.}}Y\right)\right)*\left(n\underset{0}{\overset{n}{.Math.}}{Y}^2-\left(\underset{0}{\overset{n}{.Math.}}X*\underset{0}{\overset{n}{.Math.}}Y\right)\right)}}& \left(9\right)\end{array}$

(24) PCC is calculated for each possible pair of accelerometers (i.e. x and y, y and z, and x and z) and for each possible pair of gyroscopes (i.e. roll and pitch, pitch and yaw, and roll and yaw).

(25) A second exemplary quantity comprises the number of significant oscillations in the inertial data. In embodiments of the invention, the quantity may comprise the number of significant oscillations in, for example only in, inertial data relating to the roll of clamp release tool 200. Signal processing unit 303 is configured to determine the number of significant oscillations, by searching the arrays of inertial data for elements for which the difference from the preceding element exceeds a pre-determined threshold. Signal processing unit 303 is further configured to determine the average magnitude of those differences that exceed the threshold.

(26) A third exemplary quantity comprises the skewness of an array. Signal processing unit 303 is configured to calculate the skewness of each of the arrays of inertial data, by use of the following equation:

(27) $\begin{array}{cc}\mathrm{Skewness}=\frac{\mathrm{Mean}-\mathrm{Median}}{\mathrm{Mean}+\mathrm{Median}}& \left(10\right)\end{array}$

(28) A fourth exemplary quantity comprises the kurtosis of an array. Signal processing unit 303 is configured to calculate the kurtosis of each of the arrays of inertial data, by use of the following equation:

(29) $\begin{array}{cc}\mathrm{Kurtosis}=\frac{Q1-Q3}{P_{95}-P_5}& \left(11\right)\end{array}$
Where: Q1=first quartile; Q3=third quartile; P.sub.95=95.sup.th percentile; and P.sub.5=5.sup.th percentile.

(30) A fifth exemplary quantity comprises the rates of change of each array of inertial data. Signal processing unit 303 is configured to calculate the first and second derivatives of each of the arrays of inertial data, by use of the following equations:

(31) $\begin{array}{cc}\mathrm{DevX}\left[j\right]=\frac{X\left[j\right]-X\left[j-1\right]}{\mathrm{dt}}& \left(12\right)\\ {\mathrm{Dev}}_{2}X\left[j\right]=\frac{\mathrm{DevX}\left[j\right]-\mathrm{DevX}\left[j-1\right]}{{\mathrm{dt}}^2}& \left(13\right)\end{array}$
Where: X= DevX=first derivative of X; and Dev2X=second derivative of X;

(32) A sixth exemplary quantity comprises the Shannon-Wiener diversity index. Signal processing unit 303 is configured to calculate a Shannon-Wiener diversity index for each of the arrays of inertial data by use of the following equations:

(33) $\begin{array}{cc}H=-\underset{i=1}{\overset{S}{.Math.}}{p}_i*\ln \left(p_i\right)& \left(14\right)\\ p_i=\frac{S_i}{.Math.S_i}& \left(15\right)\end{array}$
Where: p.sub.i=probability of the next entry belonging to state i; s.sub.i=state i;

(34) In embodiments of the invention, the Shannon-Wiener diversity index may be calculated based on, for example based only on, inertial data relating to the pitch of clamp release tool 200.

(35) Further exemplary quantities include the first, second, and third quartiles; the maximum value; the minimum value; the mean value; and the modal value of sorted arrays of the inertial data. The array of inertial data for a sensor can be considered as two sub-arrays, wherein the pre-trigger inertial data comprises a first sub-array and the post-trigger inertial data comprises a second sub-array. The logged inertial data therefore comprises twelve sub-arrays. Signal processing unit 303 is configured to sort the inertial data in each sub-array by magnitude and then determine the first, second, and third quartiles (where the second quartile is the median value); the maximum value; the minimum value; the mean value; and the modal value of each of the sorted sub-arrays. In this exemplary embodiment, inertial measurement unit 301 outputs 12-bit data, and therefore the likelihood of having repeated values in a sub-array is low. Therefore, in calculating the modal value, the inertial data is grouped into 100, preferably equally sized, bins and the modal bin is determined.

(36) Each of the above quantities comprises a factor that may be evaluated in determining whether a detected movement corresponds to a successful release of a POP clamp.

(37) In order to determine whether a detected movement corresponds to a successful release of a POPP clamp, signal processing unit 303 is configured evaluate one or more rules. The one or more rules define the characteristics of a movement deemed to be one corresponding to the clamp moving to the closed position. The one or more rules operate on the basis of a plurality of parameters, which may include one or more of the change in orientation of the clamp release tool, the frequency content of the logged inertial data, and one or more of the above calculated quantities. In embodiments of the invention, signal processing unit 303 is therefore configured to, on the basis of the determined frequency content, determine whether the clamp has moved to the closed position. In alternative embodiments, signal processing unit 303 is configured to determine, on the basis of a count of the number of the one or more rules satisfied, determine a percentage certainty of whether the detected movement corresponds to the clamp moving to the closed position.

(38) The one or more rules are determined by use of a machine learning algorithm. The machine learning algorithm comprises a classification tree configured to, on the basis of the plurality of parameters, classify a detected movement as either being a successful release of a POPP clamp (a clip) or as not being a successful release of a POPP clamp (a hit).

(39) Examples of movements that should be classified as a hit include unsuccessfully released clamps, an operator dropping the clamp release tool, and an operator knocking the clamp release tool against another object.

(40) The classification tree is configured to operate on the basis of a plurality of weights, which determining the significance of each of the plurality of parameters in classifying a detected movement.

(41) The machine learning algorithm is configured to train the classification tree on the basis of a corpus of training data. The corpus of training data comprises a plurality of entries, wherein each entry corresponds to a historic movement corresponding to either a clip (success) or a hit (fail). Training the classification tree comprises evaluating the performance of a given plurality of weights in classifying entries from the corpus of training data and, on the basis of that evaluation, generating one or more new candidate pluralities of weights. By repeating the training process, the machine learning algorithm iterates towards a plurality of weights that are optimised for classifying movements as clips or hits.

(42) Single classification trees are very sensitive to the contents of the corpus of training data and are susceptible to generating over-fitted solutions. Therefore, even a small change to the corpus of training data can result in a very different classification tree. The machine learning algorithm is therefore configured to generate a forest of classification trees, wherein each classification tree of the forest is trained using a randomly selected subset of the training data. In this exemplary embodiment, the machine learning algorithm is configured to generate a forest comprising 500 classification trees. The classification decision is then based on simple majority voting of the individual classifications of the trees of the forest.

(43) The machine learning algorithm is further configured to generate the one or more rules based on the forest of trees. The one or more rules are generated by a rule-based learner, for example by use of the simplified Tree Ensemble Learner, which produces one or more rules that approximate the output of the forest of trees. The resulting rules often do not perfectly replicate the output of the forest of tress, and often result in more classification errors, but are significantly less complex and easier to implement.

(44) It has been found that the most significant of the parameters to input to the machine learning algorithm (in that it provides the greatest improvement in classification accuracy) is the DFT sliding window data. The second most significant is the maximum peak in the frequency content, the third is the Shannon-Wiener diversity index, and the fourth is the number of significant oscillations and average change.

(45) Thus, signal processing unit 303 is configured to receive a signal corresponding to a detected movement of the tool head 201; record inertial data that defines the detected movement of the tool head; and, on the basis of the inertial data, determining whether the clamp has moved to the closed position. Signal processing unit 303 operates principally by detecting the characteristic vibrations, which result from a successful clamp release, the presence of the characteristic vibrations being indicative of a successful clamp release.

(46) In the event that signal processing unit 303 determines that a detected movement corresponds to the successful release of a POPP clamp, signal processing unit 303 is configured to light an LED 305, to signal to an operator of the clamp release tool that a successful release has occurred. Signal processing unit 303 is further configured to transmit a signal to a wireless communication module 307 indicating that a clip has been detected.

(47) Wireless communication module 307 is configured to transmit to a receiver a signal indicating that a clip has been detected. In an automotive production line, this enables the number of clips detected by a given tool at a given location in the production line to be recorded, thus providing traceability data (for example the number of successfully operated clamp closures per vehicle, as performed at given times by a given operator with a particular tool).

(48) In this exemplary embodiment, wireless communication module 307 is configured to transmit a message wirelessly. Alternative embodiments of the invention may instead incorporate a wired communication link. Wireless communication module 307 is further configured to transmit a periodic heartbeat signal to the receiver. The heartbeat signal includes a low battery flag, which indicates a status of a battery of clamp release tool 200. Channel selection 309 comprises one or more switches, the position of which indicates a selected one of a plurality of frequency bands (or channel) over which wireless communication module 307 may communicate with the receiver.

(49) The receiver is configured to receive the signal from wireless communication module 307 indicating that a clip has been detected. The receiver comprises a closed clamp signal received LED, and is configured to light the closed clamp signal received LED in response to receipt of the signal from wireless communication module 307 indicating that a clip has been detected. The receiver is configured to communicate with a quality control system, such that the quality control system is updated as clamps are successfully released. The receiver is further configured to receive the heartbeat signal from wireless communication module 307. The receiver comprises a low battery LED, and is configured to light the low battery LED in response to receipt of a heartbeat signal in which the low battery flag is set. The receiver comprises a channel selection switch, the position of which indicates a selected one of a plurality of frequency bands (or channel) over which the receiver may communicate with wireless communication module 307. The receiver further comprises a tool present LED, which is configured to light for as long as a heartbeat signal is being regularly received; a data received LED; a receiver OK LED; and a power on LED.

(50) In operation, inertial measurement unit 301 detects movement of tool head 201. When tool head 201 is used to release a POPP clamp, vibrations are induced in tool head 201, which are detected by inertial measurement 301. The vibrations induced by the release of a POPP clamp are of a sufficient magnitude to satisfy the trigger event condition. Signal processing unit 303 therefore records inertial data that defining the movement of tool head 201. Signal processing unit 303 then calculates a plurality of parameters for use in determining whether the inertial data corresponds to a successful clamp release, the plurality of parameters including the change in orientation of clamp release tool 201, the frequency content of the inertial data, and a number of statistical quantities. The plurality of parameters comprise inputs to one or more rules applied by signal processing unit 303, the output of which determine whether the detected movement corresponds to the clamp having moved to the closed position.

(51) FIG. 4 shows a flow chart illustrating the steps of a method 400 of operating a pre-opened pre-positioned clamp, the clamp being operable between an open position and a closed position, according to a second embodiment of the invention.

(52) A first step of the method, represented by box 401, comprises releasing the clamp from the open position using a tool head. The tool head is comprised in a clamp release tool, and the releasing is performed by an operator of the clamp release tool. The operator is typically a human but, in embodiments of the invention, may be a machine.

(53) A second step of the method, represented by box 403, comprises detecting a movement of the tool head resulting from the releasing. The release of the clamp from the open position using the tool head causes the tool head to vibrate. Thus, release of the clamp from the open position is associated with a movement of the tool head, which comprises both the operator's movements of the tool head in order to release the clamp and vibration of the tool head induced by the release of the clamp. The detecting comprises detecting movement of the tool head in respect of all three planes of motion, and rotation of the tool head in respect of its pitch, roll, and yaw.

(54) An optional third step of the method, represented by box 405, comprises, on the basis of the detected movement, determining that a trigger event has occurred. The trigger event comprises detection of four consecutive significant oscillations, wherein a significant oscillation comprises an acceleration of more than 980 ms.sup.?2.

(55) A fourth step of the method, represented by box 407, recording inertial data that defines the detected movement of the tool head. The recording is performed in response to the detection of the trigger event and comprises recording inertial data from immediately before and immediately after the trigger event. Thus, the recorded inertial data comprises data defining movement preceding the trigger event and data defining movement following the trigger event.

(56) A fifth step of the method, represented by box 409, comprises determining a frequency content of the inertial data. Determining the frequency content comprises a performing a Fourier transform on the recorded inertial data.

(57) A sixth step of the method, represented by box 411, comprises, on the basis of the determined frequency content, determining whether the clamp has moved to the closed position. If the clamp was successfully released, and therefore is now clamping the hose as expected, the frequency content will comprise one or more frequencies that are characteristic of the clamp having successfully moved to the closed position. Determining whether the clamp has moved to the closed position comprises evaluating one or more parameters. The one or more parameters may comprise a change in orientation of the clamp release tool. Determining whether the clamp has moved to the closed position may therefore comprise determining a change in orientation of the clamp release tool.

(58) The one or more parameters may comprise a Pearson's correlation coefficient, a Shannon-Wiener diversity index, a kurtosis, or a skewness of the recorded inertial data. Determining whether the clamp has moved to a closed position comprises applying one or more rules. The one or more rules define the characteristics of a movement deemed to be one that corresponds to the clamp moving to the closed position. The one or more rules are based on the output of a classification tree. More specifically, the one or more rules are based on the output of a majority vote by a plurality of classification trees. The classification trees operate to determine whether the inertial data defines a detected movement that corresponds to a successful release of the clamp. The classification trees perform the classification on the basis of a plurality of weights, the plurality of weights having been previously determined by a machine learning algorithm.

(59) FIG. 5 shows a flow chart illustrating the steps of a method 500 of manufacturing signal processing electronics for a clamp attachment tool according to a third embodiment of the invention.

(60) The tool comprises a tool head arranged to engage with the clamp and inertial measurement electronics arranged to take measurements of a movement of the tool head. The clamp is operable between an open position and a closed position.

(61) A first step of method 500, represented by box 501 comprises providing signal processing electronics, arranged to receive a signal corresponding to the measurements.

(62) An optional second step of method 500, represented by box 503, comprises determining one or more rules for classifying a detected movement as either corresponding to the clamp having moved to the closed position or as not. The one or more rules define the characteristics of a movement deemed to be one corresponding to the clamp moving to the closed position.

(63) Determining the one or more rules comprises operating a machine learning algorithm. Operating a machine learning algorithm comprises train the classification tree on the basis of a corpus of training data. The corpus of training data comprises a plurality of entries, wherein each entry corresponds to a historic movement corresponding to either a clip or a hit. Training the classification tree comprises evaluating the performance of a given plurality of weights in classifying entries from the corpus of training data and, on the basis of that evaluation, generating one or more new candidate pluralities of weights. By repeating the training process, the machine learning algorithm iterates towards a plurality of weights that are optimised for classifying movements as clips or hits.

(64) Single classification trees are very sensitive to the contents of the corpus of training data and are susceptible to generating over-fitted solutions. Therefore, even a small change to the corpus of training data can result in a very different classification tree. Operating a machine learning algorithm therefore comprises generating a forest of classification trees, wherein each classification tree of the forest is trained using a randomly selected subset of the training data. In this exemplary embodiment, operating a machine learning algorithm comprises generating a forest comprising 500 classification trees. The classification decision is then based on simple majority voting of the individual classifications of the trees of the forest.

(65) Operating a machine learning algorithm comprises further comprises generating the one or more rules based on the forest of trees. The one or more rules are generated by a rule-based learner, for example by use of the simplified Tree Ensemble Learner, which produces one or more rules that approximate the output of the forest of trees. The resulting rules often do not perfectly replicate the output of the forest of tress, and often result in more classification errors, but are significantly less complex and easier to implement.

(66) A third step of method 500, represented by box 505, comprises programming the signal processing electronics to record data relating to the measurements, determine a frequency content of the data and, on the basis of the determined frequency content, determine whether the clamp has moved to the closed position. Determining whether the clamp has moved to the closed position comprises applying the one or more rules.

(67) Whilst the present invention has been described and illustrated with reference to particular embodiments, it will be appreciated by those of ordinary skill in the art that the invention lends itself to many different variations not specifically illustrated herein. By way of example only, certain possible variations will now be described.

(68) In alternative embodiments of the invention, some or all of the functionality provided by signal processing unit 303 in the first embodiment is instead provided by the receiver. In such embodiments, wireless communication module 307 is configured to transmit inertial data from inertial measurement unit 301 to the receiver, which subsequently determines whether a detected movement constitutes a clip or a hit.

(69) In alternative embodiments of the invention, signal processing electronics 303 is implemented one or more of a processor, a field programmable gate array, a programmable logic device, and discrete electronic components. In alternative embodiments of the invention, signal processing electronics 303 is implemented wholly or in part as a software module.

(70) Where in the foregoing description, integers or elements are mentioned which have known, obvious or foreseeable equivalents, then such equivalents are herein incorporated as if individually set forth. Reference should be made to the claims for determining the true scope of the present invention, which should be construed so as to encompass any such equivalents. It will also be appreciated by the reader that integers or features of the invention that are described as preferable, advantageous, convenient or the like are optional and do not limit the scope of the independent claims. Moreover, it is to be understood that such optional integers or features, whilst of possible benefit in some embodiments of the invention, may not be desirable, and may therefore be absent, in other embodiments.