CN117705329A - Method for facilitating application of a desired clamping force by tightening a tool - Google Patents

Abstract

The present invention relates to a method that facilitates the application of a desired clamping force by a tightening tool. Method for controlling a device (10, 30) for applying a clamping force to a fastener (20) by a tightening tool (10), determining the applied clamping force based on a measured time-of-flight of an ultrasonic signal, said ultrasonic signal Emitted via the proximal end of fastener (20) and reflected from the distal portion of fastener (20). The method is based on machine learning methods to select the appropriate frequency for the ultrasonic signal being emitted.

Description

A convenient way to apply the desired clamping force by tightening the tool

Technical field

The invention relates to a method of facilitating the application of a desired clamping force to a fastener by means of a tightening tool for tightening a joint and to a device for performing the method, the applied clamping force being based on measured ultrasonic signals. Determined by time of flight, the ultrasonic signal is emitted via the proximal end of the fastener and reflected from the distal portion of the fastener as the joint is tightened.

Furthermore, a computer program is provided comprising computer-executable instructions for causing the device to perform the steps of the method when executed on a processing unit included in the device.

Furthermore, a computer program product is provided that includes a computer-readable medium embodying thereon an implemented computer program.

Background technique

In industrial applications, various types of tools are used to facilitate and assist work. For example, automated tightening tools are utilized to tighten structural joints using fasteners such as bolts and nuts or screws. In such an environment, these tools are absolutely necessary to provide sufficiently high quality and tightening force during the tightening process.

When tightening joints, it is crucial to have control over the tightening action provided by the tool. This can be done in a variety of ways, such as torque control, angle control, or by measuring the elongation of the bolt during the tightening process, also known as clamping force control.

The problem that arises during the tightening process is that for the bolt or screw that is being tightened, some undesirable tightening results may occur, for example, the bolt is not tightened correctly, resulting in a slightly larger or smaller clamping force at the joint. This problem occurs both when operators manually operate tightening tools, and in applications where the tightening process is fully automated.

Contents of the invention

It is an object of embodiments to solve or at least alleviate this problem in the prior art and thus provide an improved method for applying a desired clamping force by a tightening tool.

In a first aspect, the object is achieved by a method of controlling a device for applying a clamping force to a fastener by a tightening tool for tightening a joint, the applied clamping force being determined based on a measured time of flight of an ultrasonic signal , the ultrasonic signal is emitted via the proximal end of the fastener and reflected from the distal end of the fastener when the joint is tightened. The method includes recording a reflected ultrasonic signal for each different frequency of an ultrasonic signal transmitted at a plurality of different frequencies; and assigning to each recorded reflected ultrasonic signal a representation of each reflected ultrasonic signal and a corresponding transmitted ultrasonic wave. a quality measure of how similar the signals are; training the machine learning model to learn the quality of each recorded reflected ultrasound signal by providing the machine learning model with each recorded reflected ultrasound signal and the assigned quality measure; and providing the training The trained machine learning model provides at least one further recorded reflected ultrasonic signal, wherein the trained machine learning model determines whether the provided at least one further recorded reflected ultrasonic signal meets the quality standard, and if so, indicates that the recorded The selected frequency of the transmitted ultrasonic signal of at least one further reflected ultrasonic signal may be used to determine the clamping force exerted by the tightening tool when applying the clamping force to the fastener.

In a second aspect, the object is achieved by a device configured to control the application of a clamping force by a tightening tool to the fastener to tighten the joint, the applied clamping force being determined based on a measured time of flight of the ultrasonic signal , the ultrasonic signal is emitted via the proximal end of the fastener and reflected from the distal end of the fastener when the joint is tightened, the apparatus including a processing unit and a memory, the memory including instructions executable by the processing unit, The apparatus thereby operates to record a reflected ultrasonic signal for each different frequency of an ultrasonic signal emitted at a plurality of different frequencies; to assign to each recorded reflected ultrasonic signal a corresponding number representing each reflected ultrasonic signal. a quality measure of how similar the transmitted ultrasonic signals are; training the machine learning model to learn the quality of each recorded reflected ultrasonic signal by providing the machine learning model with each recorded reflected ultrasonic signal and the assigned quality measure; and providing at least one further recorded reflected ultrasonic signal to the trained machine learning model, wherein the trained machine learning model determines whether the provided at least one further recorded reflected ultrasonic signal meets the quality standard, and if so, then The selected frequency of the transmitted ultrasonic signal representing the recording of at least one further reflected ultrasonic signal may be used to determine the clamping force exerted by the tightening tool when applying the clamping force to the fastener.

As mentioned before, when using a tightening tool, it is important that the clamping force generated (i.e., the force that holds the bolt joint together) is accurate.

There are generally four different ways to control the tightening of threaded joints to obtain the required clamping force: torque control, angle control, gradient control and clamping force or elongation control.

Of the four methods, clamping force control using ultrasonic technology offers a more accurate and cheaper solution because it only requires access to one bolt end and is independent of friction. The ultrasonic method is based on using signal echo technology to measure bolt elongation.

However, the quality of the measurement is based in part on the quality of the echo signal, which in turn depends on the frequency of the ultrasonic signal emitted through the fastener.

Advantageously, in one embodiment, the determination of the quality of the recorded echo signals only needs to be performed in the training phase of the machine learning model; once in the execution phase, the trained machine learning model will be utilized before performing the tightening operation. Determining whether the recorded echo signals meet quality standards is a much less demanding process than determining the quality from scratch each time a joint is tightened.

In one embodiment, the trained machine learning model is provided with quality standards to meet.

In one embodiment, a machine learning model to be trained is provided with quality standards to be met.

In one embodiment, a cross-correlation is performed to find the correlation between the transmitted ultrasonic signal and the recorded reflected ultrasonic signal for each frequency.

In one embodiment, a quality criterion is considered met if the quality metric exceeds a quality threshold.

In one embodiment, machine learning is based on one or more of neural networks, random forest-based classification, and regression analysis.

In one embodiment, an alert is provided as to whether the provided at least one further recorded reflected ultrasound signal meets quality criteria.

In one embodiment, the alert indicates a recommended frequency.

In one embodiment, the alert is provided to the operator of the tightening tool, the tightening tool itself, a supervisory control room, or a remote cloud function.

In one embodiment, the tightening tool is controlled to provide an audible and/or visual alarm to the operator of the tool.

In a third aspect, a computer program comprising computer-executable instructions is provided. When the computer-executable instructions are executed on a processing unit included in the device, the computer-executable instructions are used to cause the device to perform the method of the first aspect. the steps described.

In a fourth aspect, a computer program product comprising a computer-readable medium having a computer program implemented according to the third aspect is provided.

In general, all terms used in the claims should be interpreted according to their ordinary meanings in the technical field, unless expressly defined otherwise herein. Unless expressly stated otherwise, all references to an "element, means, component, means, step, etc." should be construed publicly as referring to at least one example of that element, means, component, means, step, etc. Unless expressly stated, the steps of any method disclosed herein need not be performed in the exact order disclosed.

Description of the drawings

Various aspects and embodiments are now described, by way of example, with reference to the accompanying drawings, in which:

Figure 1 shows an industrial tool in the form of a tightening tool in which a tool embodiment can be implemented;

Figure 2 shows a joint using ultrasonic clamping force controlled tightening using fasteners in the form of bolts and nuts;

Figures 3a to 3c show the quality of the reflected signal relative to the transmitted signal;

4 shows a flowchart of a method that facilitates applying a desired clamping force to a fastener with a tightening tool to tighten a joint, according to an embodiment;

Figure 5 illustrates training of a machine learning model according to an embodiment;

Figure 6 illustrates utilizing a trained machine learning model in an embodiment; and

Figure 7 shows an apparatus in which a method according to an embodiment can be implemented.

Detailed ways

Various aspects of the invention will be described more fully hereinafter with reference to the accompanying drawings, in which specific embodiments of the invention are shown.

These aspects may, however, be embodied in many different forms and should not be construed as limiting; rather, these embodiments are provided by way of example so that this disclosure will be thorough and complete and will be sufficient to illustrate the invention. The scope of all aspects is fully conveyed to those skilled in the art. Throughout this specification, the same reference numbers refer to the same elements.

Figure 1 illustrates an industrial tool in the form of a tightening tool 10 that may implement a tool embodiment configured to apply torque to a fastener (eg, bolt 20) for tightening a joint.

The tightening tool 10 may be cordless or powered by a wire and has a body 11 and a tool head 12 . The tool head 12 has an output shaft 13 with a socket (not shown) configured to be rotatably driven by an electric motor arranged inside the body 11 to apply torque to the bolt 20 .

The tightening tool 10 may be arranged with a display 14 by means of which information relevant to the operation of the tool 10 can be presented to an operator of the tool 10 and an interface 15 through which the operator can enter data into the tool 10 .

The tightening tool 10 may further be arranged with communication capabilities in the form of a radio transmitter/receiver 16 for wirelessly transmitting operating data (eg, applied torque) to a remote controller (eg, cloud server 30). Alternatively, communication between tool 10 and controller 30 may occur via a wired connection.

Thus, the tool 10 may, for example, transmit measured operating data to the controller 30 for further evaluation, and the controller 30 may, for example, send operating settings imposed by the tool 10 or instructions displayed to the operator via the display 14 .

As mentioned previously, when using the tightening tool 10, it is important that the clamping force generated (i.e., the force that holds the bolt joint together) is accurate.

There are generally four different ways to control the tightening of threaded joints to obtain the required clamping force: torque control, angle control, gradient control, and clamping force or elongation control.

The method used in the following is clamping force/elongation control (both terms will be used in the following). Clamping force control utilizing ultrasonic technology offers a more accurate and cheaper solution because it only requires access to one bolt end and is independent of friction.

Figure 2 shows the joint 22 tightened using fasteners in the form of bolts 20 and nuts 21. Ultrasonic clamping force control requires the application of ultrasonic sensors 23 to the bolt heads. The ultrasonic method is based on using signal echo technology to measure bolt elongation.

The sensor 23 sends an ultrasonic signal of a specific frequency to the proximal end of the bolt 20, which passes through the bolt 20 and reflects from the distal end 24 of the bolt 20, and is subsequently received by the sensor 23 (called an echo). Time of flight (TOF), that is, the time elapsed between the sensor 23 transmitting and receiving the ultrasonic signal, was measured for the bolt 20 in the untightened state and for the tightened bolt. The difference between the two TOF measurements is called delta TOF (ΔTOF).

The difference in bolt length before and after tightening is called elongation. In other words, the bolt 20 will stretch as it is tightened. For example, assuming that the bolt length is 100mm in the untightened state and 102mm when tightened to provide the desired clamping force in the joint 22, the elongation required to provide the desired clamping force in the joint 22 is 2mm. It will be appreciated that as the bolt 20 elongates during tightening, the TOF increases and vice versa.

The clamping force exerted by bolt 20 to joint 22 can be estimated as:

F _cl =C ₁ ΔTOF (1)

where C1 is a constant that depends on the material properties of the measuring chain and bolt. Approximately one third of the TOF is determined by the elongation of the bolt 20 and two thirds is determined by the stress acting on the bolt 20 . This is called the acoustoelastic effect.

Now, in order to accurately measure the TOF, the signal reflected from the far end 24 of the bolt 20 needs to be of sufficiently high quality. If not, the measured TOF will be inaccurate and therefore the applied clamping force F _cl estimated using equation (1) will also be inaccurate.

Therefore, in a practical situation, the tightening tool 10 can exert a clamping force according to equation (1) that appears to be correct based on the measured TOF but is not actually the case because the estimate relies on the result as a low-quality echo signal Poorly measured TOF. In the worst case, this could lead to dangerous consequences (or at least improperly tightened bolts).

The quality of the reflected ultrasonic signal used to measure the TOF is highly dependent on the frequency of the ultrasonic signal sent by the sensor 23 .

When measuring TOF, one of two common methods is usually used; the first method uses the time difference between the transmitted signal and its first echo, while the second method uses the time difference between the first echo and the second echo. . The embodiments described herein are equally applicable to both methods.

Ultrasonic signals emitted at different frequencies will cause changes in the quality of the reflected signal, so there is a "best" frequency to emit an ultrasonic signal to obtain the highest quality reflected signal. The optimal firing frequency depends on the material, size and temperature of the bolt, the size and shape of the far end of the bolt, stresses in the bolt, and other factors.

Figures 3a to 3c show examples of three different transmitted and reflected ultrasonic signals in the setup of Figure 2. When measuring clamping force using the clamping force/elongation control method, the emitted ultrasonic signal commonly used is the so-called barker signal or chirp signal. Figures 3a to 3c illustrate Barker encoded signals.

If eg a chirp signal is utilized instead of a buck signal, additional signal parameters can be monitored, eg burst length. In such scenarios, the embodiments described herein may be implemented utilizing burst length as a parameter in addition to frequency or as an alternative to frequency.

In Figure 3a, the reflected signal (bottom) is of poor quality and therefore not similar to the transmitted signal (top). Therefore, the accuracy of the measured TOF may be poor.

In Figure 3b, the quality is slightly higher (i.e., medium quality), so the reflected signal has some similarities to the transmitted signal. Therefore, the accuracy of the measured TOF will be higher than Figure 3a, but may not necessarily be acceptable in high-precision scenarios.

In Figure 3c, the reflected signal is of higher quality and is similar to the transmitted signal to a large extent. Therefore, the accuracy of the measured TOF is likely to be high, and the applied clamping force estimated based on equation (1) is most likely to be accurate.

From the illustrations of Figures 3a to 3c it can be concluded that it is important that the quality of the reflected ultrasonic signal is high enough, and given that the quality is highly dependent on the frequency of the transmitted ultrasonic signal, the frequency must be chosen carefully.

Therefore, for each specific bolt tightening condition affected by the factors discussed above, there is an ultrasonic signal emitted by the sensor 23 that produces the highest quality reflected signal and the most accurate measurement of the TOF and thus applied clamping force. one or more optimal frequencies so that the tightening tool 10 can accurately apply the desired clamping force to a specific joint.

Given that the optimal frequency of the emitted ultrasonic signal will vary with different tightening conditions, the optimal frequency must be found for each tightening – or at least the frequency that produces a reflected signal of sufficient quality. It is computationally intensive and time consuming.

As will be described below, the tightening model is trained by applying machine learning (ML) to provide a large number of recorded echo signals, that is, ultrasonic signals of different frequencies are emitted by the sensor 23 through the bolt 20 and from the bolt 20 Far-end 24 reflections, while assigning a quality metric to each reflection signal, it will then be possible to determine whether the reflection signal provided to the trained ML model is of sufficiently high quality, and can therefore be relied upon to estimate using equation (1) The clamping force applied.

When the measurement data is collected during the training phase, torque is applied to the bolt 20 in order to tighten the joint 22 . Torque can be applied by hand and the bolt 20 is usually only lightly tightened during the training phase.

4 shows a flowchart of a method that facilitates applying a desired clamping force to a fastener with a tightening tool to tighten a joint, according to an embodiment.

As described in detail above with reference to FIG. 2 , a clamping force control method (also called elongation control) will be utilized, in which the applied clamping force when tightening the joint 22 is determined based on the measured TOF of the ultrasonic signal. The ultrasonic signal is transmitted via the proximal end of bolt 20 and reflected from the distal end 24 of bolt 20 .

Therefore, in the first step S101, the sensor 23 is controlled to emit a plurality of ultrasonic wave signals of different frequencies, and for each different frequency of the emitted ultrasonic wave signal, a reflected ultrasonic wave signal is recorded.

As discussed previously, it is important that for each selected frequency of the transmitted ultrasonic signal, the quality of the reflected signal recorded at step S101 is high enough such that the TOF can be accurately measured for subsequent estimation according to equation (1) The clamping force applied.

Therefore, in step S102, each recorded reflected ultrasonic signal is assigned a quality metric that represents the degree of similarity of each reflected ultrasonic signal to the corresponding transmitted ultrasonic signal. In other words, the better the similarity, the higher the quality.

In one example, data processing in the form of cross-correlation is performed to find correlations between the transmitted signal and the recorded reflections of that signal. In one example, the cross-correlation can be 1 if the two are completely similar, and 0 if they are not similar.

Referring to Table 1 below, for a particular frequency f of the transmitted ultrasonic signal, a quality metric q will be assigned to the recorded reflection signal based on the similarity (eg, cross-correlation) between the transmitted signal and its recorded reflection.

Table 1. The quality metric q specifies the quality of the reflected signal for each selected frequency f of the transmitted signal.

frequency of emitted signal The quality of the reflected signal f1 q1 f2 q2 f3 q3 fn qn

In Table 1, each reflected signal of frequencies f1 to fn is assigned an individual quality measure q1 to qn.

As will be appreciated, during the training phase dozens or even hundreds of frequencies can be emitted through the bolt, a quality metric can be derived and associated with each frequency. It can then be concluded that the best quality metric among all quality metrics represents the best frequency to be selected for the ultrasonic signal.

In practice, it is usually not a matter of finding the best frequency, but of finding one of the potentially many transmit frequencies that produces an echo signal of sufficiently high quality.

As will be appreciated, during normal day-to-day operation of the tightening tool 10, it is not practical to perform such data processing every time a bolt is to be tightened, since performing cross-correlation for each tightening operation is processing heavy and time consuming. . This problem is solved by training ML models.

In step S103, the ML model is trained by providing each recorded reflected ultrasound signal and the assigned quality metric to the ML model. The ML model is therefore trained to learn which frequencies will produce echo signals of sufficiently high quality.

Step S103 marks the end of the training phase, after which the execution phase begins, in which the trained ML model is utilized to determine whether the selected frequency of the transmitted ultrasonic signal produces a reflection signal of sufficient quality.

Therefore, at step S104, during normal operation of the tightening tool 10, the trained ML model is provided with at least one further recorded reflected ultrasonic signal recorded before tightening the joint.

The trained ML model determines whether the provided further recorded reflected ultrasound signal meets predetermined quality criteria.

If so, the output of the ML model indicates that the selected frequency of the emitted ultrasonic signal that records the reflected signal can be used to determine the clamping force to be applied by the tightening tool 10 when applying the desired clamping force to the bolt.

Figure 5 illustrates training of an ML model according to an embodiment, while Figure 6 illustrates utilization of a trained ML model in an embodiment.

Therefore, during the training phase, the reflected signals of the ultrasonic signals transmitted through the bolt 20 emitted by the sensor 23 at different frequencies are recorded, and a quality measure is assigned to each reflected signal, as previously described in steps S101 and S102 of FIG. 4 of.

For example, assume that ten echo signals are recorded (actually, more signals will be recorded during the training phase) and provided to the ML model as a training set, as shown in Table 2.

Table 2. Quality measure q is associated with each frequency f of the recorded echo.

frequency of emitted signal The quality of the reflected signal f1, f2, f3, f4 q1 f5, f6, f7, f8 q2 f9, f10 q3

In this particular example, the frequencies are divided into categories, where the assigned quality measure is used as the label for each category.

Therefore, for frequency classes including frequencies f1 to f4, q1 means that the cross-correlation between the transmitted signal and the corresponding echo is in the range 0.7 to 0.79, and for frequencies f5 to f8, q2 means that the cross-correlation is in the range 0.6 to 0.69 ; and for frequencies f9 and f10, q3 indicates that the cross-correlation is in the range of 0.5 to 0.59.

This data is used to train the ML model, so that the ML model can then decide in the execution phase whether the provided echo signal is of sufficiently high quality.

Depending on the implementation, any appropriate machine learning method can be used to train the ML model, for example, various types of neural networks (deep neural network (DNN), recurrent neural network (RNN), and convolutional Neural network (convolutional neural network, CNN)) or random forest-based classification, regression analysis, etc.

As shown in Figure 6, after the training set is provided to the ML model for sufficient training, the execution phase can begin.

In the execution phase, when the tightening tool 10 is to tighten the bolt 20 , the frequency of the ultrasonic signal transmitted by the sensor 23 through the bolt 20 is selected during the tightening operation.

Therefore, the TOF can be accurately measured in order to estimate the desired clamping force to be applied to the joint 22 based on equation (1), assuming that the selected frequency produces an echo signal of sufficient quality. That is, the bolt 20 will be tightened by the tool 10 until a ΔTOF representing the desired clamping force when utilizing the selected frequency is reached.

Now, referring to Table 2, assuming that the predetermined quality standard requires q≥0.7 as the quality standard to be met, and the tightening tool 10 selects the frequency f5 for the emitted ultrasonic signal, then the trained ML model will conclude that as the frequency f5 The resulting reflected signal of the transmitted signal will not be of sufficiently high quality since the assigned quality metric does not meet the predetermined quality criterion requiring q ≥ 0.7 (for frequency classes including frequencies f5 to f8, the quality metric is preferably 0.69 ).

Therefore, the trained ML model will accordingly output an indication that it does not meet the quality criteria for frequency f5 (in practice it usually outputs "1" or "0").

As shown in Figure 6, during the execution phase, predetermined quality standards can optionally be provided to the trained ML model together with the recorded reflection signals to indicate the quality requirements of the reflection signals to the trained ML model, because for different Application requirements may vary.

Alternatively, the quality requirements can be provided to the ML model during the training phase to indicate to the ML model the quality that the echo signal must exhibit in order to comply with the quality requirements.

Since the quality standards are not met, the tightening tool 10 will not use the frequency f5 to measure the TOF (and therefore indirectly measure the clamping force), but will select a new frequency (say, f4) for the emitted ultrasonic signal and record it accordingly. The echo signal is provided to the trained ML model.

Again, the trained ML model determines whether the new echo signal meets the quality standard q≥0.7. Turning to Table 2, any signal in the category including frequencies f1 to f4 is assigned a quality metric q1 of at least 0.7, and the ML model output meets the indication of quality standards.

Therefore, the tightening tool 10 will emit an ultrasonic signal at frequency f4 and measure the TOF to apply the desired clamping force according to equation (1) when tightening the joint 22.

In another example, assuming that the bolt 20 is to be tightened with lower quality requirements (say, q ≥ 0.6), the selected transmit frequency f5 does result in compliance with the quality standard (all frequencies f5 to f8 are indicated as having at least 0.6 The quality measure of the recorded echo signal q2), and the trained ML model will output the indication accordingly.

Advantageously, the determination of the quality of the recorded echo signal only needs to be performed in the training phase; once in the execution phase, the trained ML model will be used to determine whether the recorded echo signal meets the quality criteria before performing the tightening operation. The processing volume is much smaller.

In one embodiment, an alert provided on the output of the trained ML model in the execution phase indicates whether the quality of the echo signal meets the quality criteria.

In another embodiment, the determined optimal frequency (or frequencies) is presented. Therefore, where many frequencies are tested, it may be preferable to provide alerts only for the frequencies considered best.

Alerts may be provided to the operator of the tightening tool 10, the tightening tool 10 itself, the supervisory control room, or the remote controller 30.

In the event that the tightening tool 10 alerts the operator of the tool 10 , the alert may be provided audibly and/or visually, such as through the tool display 14 .

Advantageously, if an alert indicates that a selected frequency frequently appears to produce high quality echo signals, the sensor 23 may be initialized to utilize that frequency.

Conversely, if the selected frequency is often unsuccessful, the frequency can be ignored.

Even though the training phase may be performed locally in the tightening tool 10, the training of the ML model will typically be performed on a device with high-volume processing capabilities (eg, controller 30). Therefore, the controller 30 will communicate with the sensor 23 (or a device that controls the sensor 23) for controlling the sensor 23 to emit ultrasonic signals at multiple frequencies, and then provide the recorded echo signals to the controller 30.

The controller 30 , aware of the presence of ultrasonic signals emitted by the sensor 23 , will then determine and assign a quality metric to each recorded echo signal (eg, perform cross-correlation) and train the ML model.

Subsequently, during the execution phase, the trained ML model is fed further recorded echo signals, which are evaluated for quality. This can again be performed at the controller 30 , but it can alternatively be envisaged that the trained ML model is provided to the tightening tool and stored in the tightening tool, so that the calculations during the execution phase are performed locally on the tightening tool 10 , as in the training phase Such calculations are much less computationally demanding than calculations performed during

Thus, referring to FIG. 7 , the steps of the method described herein may be performed on the tightening tool 10 or the controller 30 , or may be performed cooperatively between the tool 10 and the controller 30 . Either way, the tightening tool 10 and the controller 30 may utilize a processing unit 41 implemented in the form of one or more microprocessors (CPUs) to perform one or more of these steps. A plurality of microprocessors are arranged to execute a computer program 42 (SW) downloaded to a storage medium 43 (MEM) associated with the microprocessors (eg, random access memory (RAM), flash memory, or hard drive). The processing unit 41 is arranged to cause the tool/controller to perform a method according to an embodiment when a suitable computer program 42 comprising computer executable instructions is downloaded to the storage medium 43 and executed by the processing unit 41 . The storage medium 43 may also be a computer program product including the computer program 42. Alternatively, the computer program 42 may be transferred to the storage medium 43 via a suitable computer program product, such as a digital versatile disc (DVD) or a memory stick. As a further alternative, the computer program 42 may be downloaded to the storage medium 43 via a network. The processing unit 41 may alternatively be implemented in the form of a digital signal processor (DSP), an application specific integrated circuit (ASIC), a field programmable gate array (FPGA), a complex programmable logic device (CPLD), or the like. Additionally, a communications interface (INT) 44 may be provided to allow wired or wireless communication of data.

Aspects of the invention have been described above primarily with reference to some embodiments and examples thereof. However, as a person skilled in the art will readily appreciate, other embodiments than those disclosed above are equally possible within the scope of the invention as defined by the appended patent claims.

Therefore, although various aspects and embodiments have been disclosed herein, other aspects and embodiments will be apparent to those skilled in the art. The various aspects and embodiments disclosed herein are for illustrative purposes and not intended to be limiting, with the true scope and spirit being indicated by the appended claims.

Claims (13)

1. A method for controlling a device (10, 30) for applying a clamping force to a fastener (20) by a tightening tool (10) for tightening a joint (22), based on the measured time of flight of an ultrasonic signal To determine the applied clamping force, the ultrasonic signal is emitted via the proximal end of the fastener (20) and reflected from the distal end (24) of the fastener (20) as the joint (22) is tightened. Methods include: recording (S101) a reflected ultrasonic signal for each different frequency of the ultrasonic signal emitted at a plurality of different frequencies; Assigning (S102) a quality metric to each recorded reflected ultrasonic signal, the quality metric representing how similar each reflected ultrasonic signal is to a corresponding transmitted ultrasonic signal; Train (S103) the machine learning model by providing each recorded reflected ultrasonic signal and the assigned quality metric to the machine learning model to learn the quality of each recorded reflected ultrasonic signal; and Providing (S104) at least one further recorded reflected ultrasonic signal to the trained machine learning model, wherein the trained machine learning model determines whether the provided at least one further recorded reflected ultrasonic signal meets the quality standard, and if it does , then indicating that the selected frequency of the transmitted ultrasonic signal recording at least one further reflected ultrasonic signal can be used to determine the clamping force exerted by the tightening tool (10) when applying the clamping force to the fastener (20) force. 2. The method of claim 1, providing (S104) at least one further recorded reflected ultrasonic signal to the trained machine learning model further comprising: Provides the trained machine learning model with quality standards to meet. 3. The method of claim 1, training (S103) the machine learning model further includes: Provide the machine learning model with quality standards to meet. 4. The method of any one of the preceding claims, assigning (S102) a quality metric to each recorded reflected ultrasound signal further comprising: A cross-correlation is performed to find the correlation between the transmitted ultrasonic signal and the recorded reflected ultrasonic signal for each frequency. 5. A method according to any one of the preceding claims, wherein a quality criterion is deemed to be met if the quality metric exceeds a quality threshold. 6. A method according to any one of the preceding claims, wherein machine learning is based on one or more of neural networks, random forest based classification and regression analysis. 7. The method of any one of the preceding claims, further comprising: An alert is provided as to whether the provided at least one further recorded reflected ultrasonic signal meets quality standards. 8. The method of claim 7, the alert indicating a recommended frequency. 9. Method according to claim 7 or 8, wherein the alarm is provided to an operator of the tightening tool (10), the tightening tool (10) itself, a supervisory control room and/or a remote cloud function (30). 10. Method according to claim 9, wherein the tightening tool (10) is controlled to provide an audible and/or visual warning to an operator of the tool (10). 11. A computer program (42) comprising computer-executable instructions which, when executed on a processing unit (41) comprised in the device (10, 30), cause the device (10, 30) to perform the claims Steps described in any one of 1 to 10. 12. Computer program product comprising a computer-readable medium (43) having the computer program (42) according to claim 11 implemented thereon. 13. A device (10, 30) configured to control the application of a clamping force by a tightening tool (10) to a fastener (20) to tighten a joint (22), determined based on a measured time of flight of an ultrasonic signal the applied clamping force, the ultrasonic signal emitted via the proximal end of the fastener (20) and reflected from the distal end (24) of the fastener (20) when tightening the joint (22), the device ( 10, 30) includes a processing unit (41) and a memory (43), said memory including instructions (42) executable by said processing unit (41), whereby said device (10, 30) operates as: recording the reflected ultrasonic signal for each different frequency of the ultrasonic signal transmitted at a plurality of different frequencies; assigning a quality metric to each recorded reflected ultrasonic signal, the quality metric representing how similar each reflected ultrasonic signal is to a corresponding transmitted ultrasonic signal; train the machine learning model to learn the quality of each recorded reflected ultrasound signal by providing the machine learning model with each recorded reflected ultrasound signal and the assigned quality metric; and Providing at least one further recorded reflected ultrasonic signal to the trained machine learning model, wherein the trained machine learning model determines whether the provided at least one further recorded reflected ultrasonic signal meets the quality standard, and if so, indicates The selected frequency of the transmitted ultrasonic signal recording at least one further reflected ultrasonic signal can be used to determine the clamping force applied by the tightening tool (10) when applying the clamping force to the fastener (20).