ELECTROSPINDLE WITH COOLING CIRCUIT, AND MACHINE WITH SUCH AN ELECTROSPINDLE

Description

FIELD OF INVENTION

[0001] The invention relates to electrospindles, particularly friction stir welding electrospindles. More specifically, the invention relates to an electrospindle and a machine comprising such an electrospindle; see claims 1 and 11. STATE OF THE ART

[0002] Structurally hardened or work-hardened aluminum alloys are used in the automotive, aerospace, and naval industries. However, they are difficult to weld by fusion. Riveting therefore remains in use despite its numerous drawbacks (uneven joints, increased mass, and stress concentration at the bores).

[0003] Another technique involves friction stir welding (or friction stir welding(FSW in English) invented by the British organization The Welding Institute (TWI) in 1991 (see application WO 93 / 10935). The principle is to create a weld bead by mixing the material of the parts to be joined. The bond forms while hot but below the material's melting point, thus avoiding defects related to solidification and reducing residual stresses at the joint. This process can be used on so-called "low-melting-point" alloys (aluminum, copper, magnesium) and allows for the creation of dissimilar welds. The parts to be joined are positioned without play on an anvil and clamped to prevent any relative movement during the welding operation.

[0004] The FSW welding tool is driven by a rotational and forward motion along the joint line. It consists of two parts: a shoulder that heats the material by friction against the surface of the parts to be welded, and a pin that kneads the material to create a joint between the parts.

[0005] FSW (Flexible Welding) overcomes the challenges associated with traditional assembly techniques for aluminum structures and other low-melting-point alloys. While research exists on FSW welding higher-melting-point materials (steel, titanium, etc.), it is almost exclusively used for welding aluminum, at least on an industrial scale. The major drawback of FSW is that it is a contact welding process. The tool-to-workpiece contact generates significant stress, depending on the welding parameters, the thickness, and the composition of the materials being welded.

[0006] Therefore, a machine capable of withstanding the forces involved is necessary. It must be equipped with a suitable electrospindle to provide the required torque and resist the stresses of the process. The electrospindle is the interface between the machine and the tool it carries and rotates. Its precision, power, and torque characteristics make it the key component of the machine.

[0007] Machine typologies

[0008] Today, we mainly distinguish two typologies.

[0009] The first type is the gantry-type FSW machine: these are generally custom-made, very rigid machines capable of welding significant thicknesses. The highest-capacity machines can withstand axial thrust loads of 150 kN. This allows for welding parts up to 50 mm thick. The main drawbacks are the investment cost (exceeding one million euros), the manufacturing lead time (1 to 2 years for large machines), the low headroom limiting the maximum part height, and the limited welding angles restricting 3D welding capabilities.

[0010] A second category consists of industrial robots. High-load robots are suitable for applications involving thin materials and 3D welding. They are primarily intended for applications that do not require extreme precision because, since the robot's rigidity is not that of a Cartesian machine (such as a gantry, which moves by translation as opposed to the rotations of a robot), the tool deviates due to the welding forces. The most capable robots can withstand up to 10 kN of axial thrust and are an economically attractive alternative with a fast manufacturing lead time. The investment ratio is typically 2 to 10 compared to a gantry machine. Fixed shoulder welding (SSFSW)

[0011] It is known to separate the FSW tool into two parts, the pin being rotating and the shoulder fixed. This process is called fixed shoulder welding (or in English Stationary Shoulder FSWor SSFSW). This process has many advantages, including better weld surface quality than conventional FSW and increased robustness, particularly on complex 3D trajectories. Its drawback lies in the complexity of the tools, both in design and manufacturing. It is tending to overtake FSW. Combined with a robot, this process offers interesting possibilities. High-performance robotic electrospindle

[0012] Today, we are looking for electrospindles that can be universally adapted. Indeed, more and more users are required to perform functions additional to FSW / SSFSW welding, such as machining.

[0013] In all three cases, the primary criterion is compactness. Indeed, the size must not reduce the robot's nominal load capacity. An industrial robot has a load curve that defines its load capacity based on the overhang at its wrist. The longer this overhang, the more the robot loses in load capacity. In process terms, this implies a loss of performance, because the lower the load capacity, the thinner and slower the robot welds. This also leads to greater tool deflections and more significant vibrations due to a loss of overall rigidity. Finally, a large volume at the end of the robot arm limits accessibility within tooling (for example, due to the presence of plate clamping elements on either side of the area to be welded).

[0014] In FSW / SSFSW welding, the tool is in contact with the workpieces. On aluminum, the temperature at the weld core is estimated to be around 500°C. It is therefore essential to cool the tool holder and the tool itself to prevent the heat from rising into the electrospindle. Indeed, the spindle bearings are not designed to operate at temperatures exceeding 80°C.

[0015] It is known that for this purpose a nozzle can be placed outside the tool holder to send air onto the work area, but this solution is bulky and inefficient.

[0016] The same applies in machining mode. Dry machining of aluminum is a complex operation. Microlubrication is often necessary to preserve the life of the cutting tools and prevent the aluminum from sticking to them.

[0017] In application EP-3 546 107, at the figure 3The electrospindle includes a cooling fluid conduit rigidly attached to the housing. This conduit extends into a channel in the shaft, leading directly to a cavity in the tool holder. The cooling fluid then rises in the shaft channel, around the housing channel, passes between the shaft and the housing at the upper end of the shaft, is collected in a trough, and then drained away. Thus, the fluid travels the entire length of the shaft before being discharged. However, this arrangement appears practically feasible only if the fluid is air, which is restrictive because it is sometimes desirable to be able to cool with a liquid. Furthermore, it has the disadvantage that the fluid, as it rises, returns some of the heat it acquired in the tool holder to the shaft. This hinders the overall cooling of the shaft.

[0018] One aim of the invention is to improve the cooling of the electrospindle.

[0019] US document 2019 / 126385 describes an electrospindle conforming to the preamble of claim 1. DESCRIPTION OF THE INVENTION

[0020] An electrospindle according to the invention is defined in claim 1, the electrospindle comprising: a housing comprising at least one cooling duct, a shaft mounted to rotate relative to the housing, the shaft having a housing for receiving a tool holder and at least one cooling duct opening into the housing, and a rotating joint comprising a part rigidly attached to the housing and a part rigidly attached to the shaft, the joint connecting the duct of the housing and the duct of the shaft with fluid.

[0021] Thus, the rotary joint allows the use of a gaseous or liquid cooling fluid for efficient cooling of the electrospindle. It enables a simple cooling circuit design, keeping the electrospindle compact.

[0022] Advantageously, the electrospindle includes bearings, all of which are located outside the joint.

[0023] The joint therefore has no bearing. This allows the rotating joint and the electrospindle to be more compact.

[0024] In one embodiment, the housing includes a robot interface in which extend the part of the seal integral with the housing and the cooling duct.

[0025] Thus, housing part of the seal and cooling circuit directly into the robot interface makes the electrospindle even more compact.

[0026] It can be expected that the shaft also includes at least one conduit extending from a face of the shaft located opposite the housing to the shaft cooling conduit.

[0027] Such a conduit allows for the evacuation of any coolant present at the upper interface between the shaft and the housing. For a liquid, this evacuation occurs by gravity (drainage of the rotary joint).

[0028] It can be foreseen that the electrospindle includes a tool holder received in the housing, the tool holder including at least one evacuation conduit in fluid communication with the cooling conduit and in fluid communication with the external environment of the electrospindle without passing through the shaft.

[0029] Thus, the cooling fluid is evacuated without heating the shaft or the electrospindle for better cooling efficiency.

[0030] It can be predicted that the electrospindle will have at least one of the following characteristics: The tool holder includes at least two exhaust ducts; each exhaust duct communicates directly with the cooling duct; the exhaust ducts and the cooling duct intersect at a single point; the exhaust duct, or at least one of the exhaust ducts, extends in the opposite direction from the electrospindle from the cooling duct; the exhaust duct, or at least one of the exhaust ducts, extends towards the electrospindle from the cooling duct; and the electrospindle forms a friction-mixing electrospindle, for example, comprising an external shoulder fixed relative to the housing or lacking an external shoulder fixed relative to the housing.

[0031] The invention also provides for a machine comprising an electrospindle according to the invention, the machine comprising a supply element for the cooling duct of the housing with a cooling fluid.

[0032] The cooling fluid can be air, an air / oil mixture, or a liquid. DESCRIPTION OF THE FIGURES

[0033] We will now present one embodiment of the invention by way of non-limiting example, supported by the drawings in which: [ Fig. 1 ] ] Fig. 2 ] ] Fig. 3 ] THE figures 1 to 3 are perspective, bottom, and axial cross-sectional views of an electrospindle according to an embodiment of the invention; [ Fig. 4 ] there figure 4 is a larger-scale view of part of the electrospindle in the previous figure; [ Fig. 5 ] ] Fig. 6 ] THE Figures 5 And 6 are views analogous to the figure 3showing the electrospindle equipped with a tool in a rotating shoulder configuration; [ Fig. 7 ] there figure 7 is a view of one end of the tool in one embodiment variant; [ Fig. 8 ] there figure 8 is a view analogous to the figure 6 showing the electrospindle equipped with a tool in a fixed shoulder configuration; [ Fig. 9 ] ] Fig. 10 ] THE figures 9 And 10 are axial cross-sectional views of the electrospindle showing details of the cooling circuit; Fig. 11 ] there figure 11 is a view illustrating the operating principle of the electrospindle during a friction-mixing operation; [ Fig. 12 ] ] Fig. 13 ] THE Figures 12 and 13 illustrate a force sensor of the electrospindle; [ Fig. 14 ] ] Fig. 15 ] THE Figures 14 And 15 are views from below and in elevation of the electrospindle showing the position of the markers; [ Fig. 16 ] ] Fig. 17 ] THE figures 16 And17 are views of the machine with the electrospindle from the previous figures and its robot during a calibration operation during the implementation of the process of the invention; [ Fig. 18 ] ] Fig. 19 ] THE Figures 18 and 19 are perspective views illustrating two configurations of the machine's calibration components; and [ Fig. 20 ] there Figure 20 is a perspective view of one of the electrospindle connectors.

[0034] We will describe, with reference to the figures, an embodiment of an electrospindle according to the invention. General presentation

[0035] The toolless electrospindle 2 is shown on the figures 1 to 4 , 9 And 10 The same electrospindle 2, equipped with a tool and in a rotating shoulder configuration, is illustrated in the Figures 5 And 6 The same electrospindle 2, equipped with a tool and in a fixed shoulder configuration, is illustrated in the figures 8 And 14 à 17 . Temperature measurement in rotating shoulder mode

[0036] With particular reference to the figure 3 , the electrospindle comprises a housing 4 and a rotating part 6 mounted to rotate relative to the housing around an axis, most often vertical, ZZ which constitutes a main axis of the electrospindle.

[0037] In this case, an upper part of the housing consists of an interface plate for robot 10, visible at the figure 16 forming a base and intended to support the electrospindle. It is connected to another part 11 of the housing, forming a support for the rotating part, by a ball joint through which the axis ZZ passes.

[0038] The rotating part 6 includes a shaft 12 which has at its lower axial end a housing 14 for receiving a tool holder 16. When the tool holder is received in the housing, it is made rigidly fixed to the shaft by conventional fastening means which will not be detailed here so that it rotates as a single unit with the rotating part.

[0039] The electrospindle includes bearings, in this case three in number, formed by ball bearings 18 mounted in an "O" configuration for the support and guidance of the rotational movement of the rotating part 6 relative to the housing 4. In this example, two of the bearings are located in the lower part of the shaft and the third is located in the upper part of the shaft.

[0040] The electrospindle includes an electric motor 20 capable of rotating the shaft relative to the housing.

[0041] The rotating part 6 comprises, in this case, a lower flange 22 rigidly fixed to a lower axial end of the shaft 12. The flange is annular in shape, such that it has a central recess providing access to this end, in particular to the housing 14, from outside the electrospindle. The flange 22 has a flat lower external face 24 perpendicular to the axis ZZ. The flange is surrounded by an annular cap 23 rigidly attached to the housing.

[0042] The flange 22 includes two connectors 26 arranged on the outer face of the flange and configured for receiving an electrical or electronic signal. They are positioned symmetrically on either side of the axis ZZ. Each connector 26 is equipped with a leaf spring 27, in this case two of them, so that it can receive two cables in two respective slots 29 of the connector. To use it, the spring 27 is actuated, for example by means of a screwdriver, a cable is inserted into the corresponding slot 29, and the spring is released to lock the cable in the connector.

[0043] The electrospindle includes means for wirelessly transmitting the signal from each connector between the rotating part 6 and the housing 4. These means include an annular electronic circuit board 28, rigidly attached to the rotating part. In this example, the board is housed on a shoulder formed on the upper part of the shaft, which extends opposite the housing. The circuit board 28 is arranged so that its electrical current is supplied by induction due to the movement of the rotating part, thus eliminating the need for a battery.

[0044] The transmission means also include an annular electronic card 30 rigidly attached to the housing and coaxial with the card 28. It extends opposite the shoulder so that the card 28 rotates in front of the card 30 of the housing when the shaft rotates.

[0045] We therefore observe that these means are located inside the electrospindle and closer to an upper axial end of the shaft than to its lower axial end which has housing 14.

[0046] The signal is transferred wirelessly from the rotating board 28 to the fixed board 30. The fixed board 30 is powered by a DC or AC voltage from the machine, and the power is transferred to the rotating board 28 by induction, thus eliminating the need for a battery in the rotating part. A distance of between 4 and 8 mm is recommended to ensure proper operation of the induction system. Therefore, there is no maintenance required, and the system can operate continuously.

[0047] The shaft 12 has two straight conduits 32, each extending in a direction parallel to the axis ZZ, at a distance from it. The conduits extend from the respective connectors 26 to the card 28 of the shaft. They house cables (not shown) connecting each connector to the card 28.

[0048] THE Figures 5 And 6 illustrate a tool holder 16 received in the housing 14. The tool holder comprises a support formed here by a chuck 34 which carries a tool 36 at its lower axial end. The sleeve extends outwards from the flange along the axial direction ZZ. Similarly, the tool 36 extends outwards from the chuck 34 along the same direction.

[0049] The tool holder 16 includes a thermocouple 38. The thermocouple includes a main body 40 carrying at an upper axial end a cable 42 and at an opposite lower axial end an elongated sheath 44.

[0050] The cable 42 passes through a lateral conduit 46 of the chuck so that it exits the tool holder, outside the electrospindle opposite the flange 22. The cable 42 at its free end can be connected to one of the connectors 26 of the flange 22.

[0051] The tool 36, for example, is made of tungsten carbide or steel. It also has a generally cylindrical shape. Its upper part features a threaded cavity 50 into which a threaded portion of the thermocouple body 40 is screwed. The thermocouple is thus rigidly fixed to the tool. The threaded hole at the rear of the tool allows adjustment of the thermocouple's position when fitted with a threaded holder, eliminating the need to adjust it based on the length of the sheath or tolerance errors (e.g., the length of the tool 36). The connection between the thermocouple sheath and the cable is made within the threaded holder. A nut 56 at the rear secures the thermocouple in position.

[0052] The thermocouple sheath 44 extends into a first conduit 52 of the tool, following the cavity 50, and then, at the lower axial end of the tool, into a second conduit 54, or terminal conduit, having a diameter smaller than a smaller diameter of the cavity 50. The diameter of this second conduit 54 is just sufficient to allow the passage of the thermocouple sensor. This conduit extends into the tool pin intended to be embedded in the material during welding. In this case, this terminal conduit opens outside the electrospindle. Such a conduit with a very small diameter in front of the tool prevents weakening it while positioning the thermocouple as close as possible to the welded core.

[0053] The thermocouple sensor has a diameter of 3 mm or less, and preferably 0.5 mm or less. The smaller the thermocouple diameter, the greater the responsiveness. Furthermore, using a hot junction connected to ground also improves responsiveness. Here, the thermocouple sheath diameter is 0.5 mm. The table below shows the associated response time. The time is reduced tenfold between a 0.5 mm and a 1.5 mm sheath. Having the hot junction connected to ground provides an additional 33% performance gain. A 0.25 mm sheath would be even better, but is more complex to manufacture mechanically and susceptible to junction fragility. [Table 1]

[0054] Table: Response time according to sheath diameter Diameter (mm) Response time(s) 0,25 0,015 0,5 0,03 1,0 0,15 1,5 0,3 2,0 0,4 [Table 2]

[0055] Table: Response time according to welding method Diameter (mm) Time(s) 63.2% Time(s) 90% Insulated hot solder 0,25 0,015 0,050 0,5 0,030 0,082 Hot soldering to ground 0,25 0,007 0,033 0,5 0,020 0,060

[0056] Alternatively, as illustrated in the figure 7 The terminal conduit 54 of tool 36 may not be open. The thermocouple sheath may therefore reach a mechanical stop. However, this complicates manufacturing because creating a blind conduit of small diameter over a long length is difficult.

[0057] Generally, welding tools are made of tool steel, allowing for easy manufacturing. Often, the sleeve 34 can be a single piece with the tool 36. However, for certain applications (high-yield-strength aluminum, for example), the material of the pin forming the tool 36 can be a grade with better mechanical properties than the tool holder 34. This is the case here, where the welding tool 36 is divided into two mechanical parts: a pin and a pin holder, with the pin inserted into the pin holder. This is called a hybrid tool because only the functional part is made of carbide. The advantage is that obtaining a central hole is completely standardized and achieved by shrink-fitting (drilling a carbide-like material is very difficult because it is a very hard material).Depending on the needs, this hole can be plugged once the thermocouple sheath has been put in place, for example using a thermal sealing cement that can withstand high temperatures (>600°C).

[0058] The two connectors 26 allow the user to connect two thermocouples (for example, one in the welding tool and another in the chuck). Since they are permanently fixed to the shaft, periodic rotational rebalancing is unnecessary. The symmetry of the two connectors 26 naturally improves the balance (which would not be the case with a single connector).

[0059] These elements of the temperature measurement system are fully integrated into the electrospindle and make it extremely compact.

[0060] In this embodiment, it is an ultra-compact, high-performance industrial electrospindle for use in both research environments and high-volume production, compatible with FSW, SSFSW, and machining, with or without temperature measurement. This principle can be extended, in particular, to measure the temperature in machining cutters.

[0061] In this friction-mixing electrospindle, the tool 36 mounted in the chuck 34 illustrated in the figure 6 features an external shoulder mounted to rotate relative to the housing. Fixed shoulder mode

[0062] We illustrated at the figure 8 The same electrospindle, this time in a fixed shoulder configuration. We see the electrospindle 2, the tool holder 16, the chuck 34, and the tool 36 with all their components as in the previous figures.

[0063] This time the electrospindle also carries a cover or shoulder 60 covering the entire lower part of the electrospindle, in particular the rotating flange 22, including the chuck 34 and most of the tool 36. This cover 60 is rigidly fixed to the housing in this case by screws 62. In the present example, it is not fixed directly to the cap 23. It must be put in place after the chuck 34 and the tool 36 have been received on the spindle and the thermocouple cable(s) have been connected to the connectors 26 of the flange. The cover is not in direct contact with the tool holder 16, because an air gap is provided between them, opposite the entire external surface of the tool holder and the rotating flange 22. It is this cover which forms the fixed shoulder at its lower axial end from which the rotating end of the tool 36 emerges.

[0064] It can be observed that the loads received by the fixed shoulder pass directly through the housing or frame of the electrospindle without passing through the flange or the shaft since there is no direct support on the outer ring of the front bearings, which provides great robustness to the assembly.

[0065] We can therefore see that this solution for temperature measurement is universal and can be applied to FSW and SSFSW welding tools, and in different materials. On the rotating joint

[0066] We will now present the cooling circuit of electrospindle 2.

[0067] With reference to figures 9 And 10The housing 4 includes a cooling channel 70, in this case straight and horizontal. The shaft 12 also includes at least one cooling channel 72, in this case straight and vertical, centered on the axis ZZ and opening at its lower axial end into the housing 14 of the tool holder, as can be seen in Figures 5 And 6 .

[0068] The electrospindle 2 includes a rotary joint 74 comprising a fixed upper portion 76 rigidly attached to the housing 4 and a rotating lower portion 78 rigidly attached to the shaft 12. The joint connects the fluid channel 70 of the housing to the channel 72 of the shaft. The portion of the joint 76 attached to the housing and the cooling channel 70 of the housing extend within the robot interface 10. This design allows for a very compact assembly, as the fixed portion of the rotary joint is part of the interface 10, eliminating the need for additional components. The rotary joint is of a universal type. It allows the transfer of a fluid (such as air, lubricant, or a microlubrication fluid, which is an air / oil mixture) from the fixed element to the rotating element. The two parts of the joint rub against each other to ensure a dynamic seal, preventing any leakage or loss.This system guarantees a 100% leak-proof fluid supply to the front section of the electrospindle. All the electrospindle's bearings are located outside the seal, which itself is bearing-free. The rotary seal 74, for example, is the one sold by Deublin under part number 1101-632-343.

[0069] As illustrated in the Figure 10 The shaft 12 further includes drainage channels 80, here six in number, extending from an axial upper face of the shaft located opposite the housing to the cooling channel 72 of the shaft. Two of these channels are visible on the Figure 10Each drainage channel 80 forms a bend and comprises a first straight vertical section opening at its upper end onto the top face of the shaft opposite the housing, and an inclined section whose lower end opens into the cooling channel 72. The drainage channels 80 allow the fluid to flow by gravity through the central passage of the shaft. This eliminates the need for a specific and bulky drainage cable. These channels evacuate fluid that would otherwise be present at the interface between the fixed part 76 of the rotary joint and its rotating part 78 (thus preventing the accumulation of oil-like fluid in the upper area).

[0070] With particular reference to the figure 6The tool holder chuck 34 includes four straight drain channels 46, connected upstream to the cooling channel 72 of the shaft and downstream to the external environment of the electrospindle. The downstream connection is made without passing through the shaft or the housing. The terms "upstream" and "downstream" refer to the direction of the cooling fluid flow. Each drain channel 46 has a downstream end that opens onto an external face of the chuck. One of these channels houses the thermocouple cable 42.

[0071] In this case, the exhaust ducts 46 each communicate directly with the cooling duct 72. The exhaust ducts 46 and the cooling duct intersect at the same point. Two of the exhaust ducts 46 extend downwards, therefore in the opposite direction to the electrospindle, from the cooling duct 72. The other two exhaust ducts 46 extend upwards towards the electrospindle from the cooling duct.

[0072] The machine includes a component not shown for supplying the cooling duct 70 of the crankcase with a cooling fluid.

[0073] The fluid, such as air, enters through the center of the shaft. The tool holder clamping system on the shaft (for example, the one marketed under reference HSK 63 by Semar, which forms a standardized tool holder mounting point) delivers air in a sealed manner to the tool holder. The tool holder's exhaust channels 46 allow the air to circulate and be expelled from the electrospindle. A tool holder 34 could be equipped with a cooling labyrinth to provide convective cooling over a large surface area, which would be even more efficient. The principle remains the same.

[0074] These elements are valid regardless of the operating mode, FSW and SSFSW welding, as can be seen on the figure 8or machining. Depending on the operating mode, the cooling fluid is either air or a liquid such as a lubricant. In the electrospindle, in fact, all modes are possible: dry running, with air, with micro-lubrication, or even with full lubrication. The user doesn't need to worry about long-term operation. In machining mode, this setup allows for liquid delivery to the tool, or even to the cutting edge of the tool (as provided by HAIMER, with a chuck equipped with cooljet technology, for example), for micro-lubrication, resulting in more efficient machining, lubricant savings, and a cleaner work area. Measuring efforts

[0075] As illustrated in Figures 1 , 3 , 5 And 12 à 15The electrospindle 2 includes three force sensors 90. Each sensor 90 is interposed between the base 10, formed by the interface plate attached to the robot, and the support 11 located below it. It is configured to measure only the force exerted by the support 11 on the base 10 in a direction parallel to the ZZ axis. These are beam-type sensors, as illustrated in the diagrams. Figures 12 and 13 The operating principle of such a sensor is very simple. It only measures a force F in one direction (positive and negative) as represented on the figure 13 This sensor has the advantages of being economical (it costs less than 200 euros), providing flexibility in design by freeing up the entire center of the electrospindle, and having a minimalist footprint and extremely good robustness.

[0076] The three sensors 90 extend in the same plane perpendicular to the axis. They are here the image of each other by a rotation around the axis by an angle of 120° and are therefore distributed homogeneously on the periphery of the electrospindle.

[0077] Here, we have planned three sensors taking into account the dimensions of the electrospindle, but we could have planned 4, 5, 6 or even more depending on the mechanical possibilities.

[0078] We consider the frame of reference associated with tool 36. It is in this frame of reference that the forces exerted on the tool during the use of the electrospindle must be calculated. With reference to the figure 14 With the three-sensor system 90, we make the following assumptions: a perfect 120° angle between the three sensors; the measurement centers coincide with the main axis ZZ; and the electrospindle is considered infinitely rigid between the tool and the sensors.

[0079] The axial force is therefore the sum of the three forces measured by the sensors, i.e., Fz = F1 + F2 + F3. For the measurement to be consistent, the system has a degree of freedom of movement via the aforementioned ball joint (centering on a line) between the base 10 and the support 11, which prevents excessive static indeterminacy that would prevent the sensors from deforming. However, in reality, at this stage, complete control is not achieved at the center of the tool 36 because the center of the sensor system does not perfectly coincide with the tool's reference frame (the sensors are fixed by screws in the support 10, requiring a functional mounting clearance). Therefore, we are simply performing a summation of three forces without an exact reference frame, yielding a result with a certain level of accuracy, but not optimal. Furthermore, since the sensors are located on the upper part of the electrospindle, the loss of stiffness of the tool up to the sensors is not taken into account.

[0080] Furthermore, we lack knowledge of the Fx and Fy forces. While knowledge of Fx and Fy forces is not mandatory in FSW / SSFSW or machining, it is beneficial for several reasons. Firstly, these two forces subject the tool in the material to shear stress. Therefore, they must remain within an acceptable range relative to the chosen tool to prevent breakage during use. Secondly, this knowledge allows for the real-time detection of abnormal overloads that could damage the electrospindle or the machine. Moreover, in a variant of the SSFSW process called Corner SSFSW (corner welding), these two forces must be controlled in the same way as the Fz force.

[0081] Therefore, it is useful to calibrate the three-sensor beam 90 arrangement to determine the forces Fx, Fy, and Fz in the tool's frame of reference. If desired, this calibration could be performed for only one or two forces; the system is thus multivariable and fully generalizable (solving an optimization system). The situation is illustrated here on an electrospindle with three sensors, but generalization to "n sensors" is possible.

[0082] The machine thus includes calculation means comprising at least one computer or automated calculation system capable of calculating forces Fx, Fy and Fz exerted on the tool 36 along the Z direction and in two directions X and Y perpendicular to the axis as a function of the forces F1, F2 and F3 measured by the sensors 90 along the Z direction but at a distance from the axis.

[0083] Thus, in the process of controlling the electrospindle during its use for welding or machining: We measure at three different locations, i.e. in the three sensors 90, the respective forces F1, F2 and F3 exerted by the support 11 on the base 10 in a direction parallel to the axis ZZ, then we calculate the forces Fx, Fy and Fz exerted on the tool 36 along the directions X, Y and Z as a function of the measured forces F1, F2 and F3.

[0084] For calibration, with reference to figures 16 to 19The installation comprises a base 92 and a multi-axis sensor 94 rigidly fixed to the base. This sensor is capable of measuring forces in three directions simultaneously, these three directions forming an orthogonal coordinate system. The sensor 94 is securely anchored to the base to allow for the generation of significant forces and obtain useful measurement ranges. This provides a complete 6-axis sensor. This is a major advantage of the proposed calibration system, as it eliminates the need for pulley systems to load the system, making it extremely flexible and precise in an industrial setting.

[0085] To avoid damaging the force sensor, the installation also includes an interface plate 96 rigidly fixed to the multi-axis sensor 94 and featuring a central guide channel opening from an upper side of the interface plate opposite the sensor to insert a calibration tool as on the figure 17 This tool replaces the welding or machining tool in the electrospindle during calibration.

[0086] The base includes 98 calibration points that allow for leveling the entire sensor relative to the electrospindle. This enables alignment of the measurement axes to match the force plate measurements with the robot's tool reference. The key is to ensure a consistent transformation between the measurement references. Vision targets can also be used for increased accuracy.

[0087] Using these elements, the preliminary calibration process is implemented as follows.

[0088] With reference to the figure 18 There are two important points of reference: the 6-axis force sensor reference point (in red) which is fixed; and the robot tool reference point;

[0089] The goal is to achieve a coincidence between the two reference frames. (Alternatively, a known transformation between them can be used). The advantage of this method is that it takes into account the positioning uncertainties of the three beam sensors on the electrospindle and the potential loss of stiffness in the electrospindle, and calibrates the entire assembly relative to the tool reference frame and its center.

[0090] To do this, first of all, the electrospindle 2 mounted on the robot 10 is equipped with the calibration tool. This calibration tool can be mounted either in the tool holder 34 in place of the welding tool 36, as referenced in the figure 6, either through the shoulder carrier 60 with reference to the figure 8 In the first case, the charge will pass through the bearings of the electrospindle; in the second, it will pass directly through the body 4 of the electrospindle. This is a tool 36 similar to the welding tool but of lower quality and a simplified shape with reduced clearance, which is inserted into a guide channel of the interface plate 96, thus allowing the precise alignment of the two reference marks.

[0091] Using the robot, the tool is brought successively close to each of the three calibration points 98 in order to align, in the calculation methods, the X, Y, Z coordinate system of the electrospindle, originating at the free end of the tool 36, with the coordinate system of the multi-axis sensor 94, originating in the central conduit. The two coordinate systems can thus be superimposed in the calculation methods as illustrated in the figure 18 .

[0092] Next, with tool 36 housed in the conduit of sensor 94, the tool is moved along the XYZ directions to generate contact between tool 36 and sensor 94, producing reaction forces from the sensor against the tool. Indeed, once the tool is inserted into the center of the interface, small movements along the directions of the tool coordinate system (or the sensor coordinate system) can be used to apply stress to the system. The data collection procedure is therefore extremely simplified; it does not require a pulley system or the use of dynamometers to apply stress and measure forces. The procedure should allow for various observations and the scanning of a wide range of forces to obtain optimal conditioning of the optimization problem and find a relevant solution. The result is a calibration matrix linking the two force measurement systems.

[0093] During an observation, the forces Fx, Fy, and Fz exerted on the tool by sensor 94 are simultaneously measured in three directions. The forces F1, F2, and F3 seen by sensors 90 on the electrospindle are also measured along the ZZ direction. A certain number of observations (denoted m hereafter) are necessary to obtain optimal accuracy of the solution.

[0094] The detailed implementation, for example, proceeds as follows in the case of n beam-type sensors. Suppose we have a dataset of m force observations F i = F x i F y i F z i T ∈ ℝ for m conditions x i = F 1 i , F 2 i , . . , Fn i ∈ ℝ , i = 1 , . . , m We would like to approximate the vector F i as a function of x i = [F1 i ,F2 i ,..,Fn i ] (overdetermined system in the case where m>n, a unique solution does not exist, so we approximate the solution): F ∼ f x c where the function f depends linearly on p=3xn coefficients c = c 1 , c 2 , . . , c p T ∈ ℝ The regression problem consists of determining the p coefficients [c1, ..., cp] that will minimize (in a direction to be specified) the distance between F and the prediction f ( x From an algebraic point of view, the regression problem can be rewritten as follows; we would like to find a solution to an overdetermined system: Wc = F

[0095] Or W is a matrix of dimensions (3m x 3n) and m > n that depends on the data conditions x i = F 1 i , F 2 i , . . , Fn i ∈ ℝ , i = 1 , . . , m , c is the vector of unknown coefficients of dimensions (3n x 1) and F is the data vector of observations with dimensions (3m x 1).

[0096] The least squares problem boils down to finding the vector of coefficients c which minimizes the (Euclidean) distance that separates WC of F, that is to say, to find c which minimizes the Euclidean norm F - Wc.We detail here the solution for three beam-type sensors, i.e., n=3. From a practical point of view, we seek to solve the following problem: F OUTIL ∼ MF CAPTEURS ⇔ F x F y F z ∼ c 1 c 2 c 3 c 4 c 5 c 6 c 7 c 8 c 9 F 1 F 2 F 3 Or : F TOOL the vector of forces in the tool frame (X, Y, Z); M the calibration matrix containing the values ​​of the vector c ; And F SENSORS the vector of forces from the beam sensors.

[0097] The goal is to determine "the best" calibration matrix M by solving an optimization problem by making m observations.

[0098] For m=4>3 observations, the regression problem given by equation (3.2) can be written as: Wc = F ⇔ F 1 1 F 2 1 F 3 1 0 0 0 0 0 0 0 0 0 F 1 1 F 2 1 F 3 1 0 0 0 0 0 0 0 0 0 F 1 1 F 2 1 F 3 1 F 1 2 F 2 2 F 3 2 0 0 0 0 0 0 0 0 0 F 1 2 F 2 2 F 3 2 0 0 0 0 0 0 0 0 0 F 1 2 F 2 2 F 3 2 F 1 3 F 2 3 F 3 3 0 0 0 0 0 0 0 0 0 F 1 3 F 2 3 F 3 3 0 0 0 0 0 0 0 0 0 F 1 3 F 2 3 F 3 3 F 1 4 F 2 4 F 3 4 0 0 0 0 0 0 0 0 0 F 1 4 F 2 4 F 3 4 0 0 0 0 0 0 0 0 0 F 1 4 F 2 4 F 3 4 c 1 c 2 c 3 c 4 c 5 c 6 c 7 c 8 c 9 = F x 1 F y 1 F z 1 F x 2 F y 2 F z 2 F x 3 F y 3 F z 3 F x 4 F y 4 F z 4

[0099] Problem (3.4) is overdetermined (12 equations for 9 unknowns). The matrix Wis not square, and has dimensions (12 x 9). A unique solution does not exist, but there is a least-squares solution. Consider the following squared error: ε = F − Wc 2 = c T W T Wc − 2 c T W T F + F T F

[0100] The vector of unknowns c is obtained by minimizing equation (3.5). Since the function is convex, the solution is obtained by solving the following partial differentiation: ∂ ε ∂ c = 2 W T Wc − 2 W T F = 0

[0101] The vector of solutions c is given by: c = W + F Or W +< = (WT< W) -1< WT< is the pseudo-inverse or generalized inverse of W. It corresponds to an approximate solution of the linear system (3.4) minimizing the squared error ε = ( F - WC ) 2< . The calibration matrix M can then be formatted to solve equation (3.3).

[0102] In order to achieve convergence of the solution vector and obtain efficient and robust calibration, it is necessary to perform as many observations m as possible. It is trivial to say that the greater the number of observations m, the more equation (3.3) tends towards accuracy rather than approximation. Similarly, it is trivial to say that the greater the number of beam sensors, the greater the accuracy. The arrangement of the possible number of sensors is strongly related to the maximum permitted space. To obtain a coherent mathematical problem condition (matrix condition), W), It is necessary to cover a fairly wide range of efforts during the different observations in order to have maximum sensitivity on the measurements.

[0103] In analysis, conditioning measures the dependence of the solution to a numerical problem on the data, in order to verify the validity of the solution. Indeed, the data generally depend on experimental measurements and are therefore subject to errors. Dependence is most often a numerical quantity. More generally, conditioning is a measure of the numerical difficulty of the problem. A problem with low conditioning is said to be well-conditioned, and a problem with high conditioning is said to be poorly-conditioned.

[0104] Example (calculations performed using MATLAB software): This demonstrates the relevance of the proposed method by considering the subsystem containing only Fz (the Fx and Fy values ​​can be difficult to interpret without actual testing, but the principle remains the same; the example is intended to be illustrative). The number of beam sensors is three.

[0105] Indeed, due to the arrangement of the sensors on the electrospindle, the result should be very close to a unit summation for Fz, i.e., Fz = F1 + F2 + F3. We will now demonstrate the advantages and robustness of the measurement noise injection method, and show that the assumption is not accurate.

[0106] Considering only the 3rd line, equation (3.3) becomes: F z = c 7 c 8 c 9 F 1 F 2 F 3

[0107] For m=5>3 observations, the regression problem given by equation (3.2) can therefore be written: Wc = F ⇔ F 1 1 F 2 1 F 3 1 F 1 2 F 2 2 F 3 2 F 1 3 F 2 3 F 3 3 F 1 4 F 2 4 F 3 4 F 1 5 F 2 5 F 3 5 c 7 c 8 c 9 = F z 1 F z 2 F z 3 F z 4 F z 5

[0108] Case 1: perfect system (sensors at 120°, measurement center on the electrosh axis, perfect and unity summation). The following tests are therefore noted: Wc = F ⇔ − 2500 − 1500 − 1000 − 3500 − 2000 − 2000 1000 − 6000 − 7000 1500 − 8500 1000 1000 − 2000 − 2000 c 7 c 8 c 9 = − 5000 − 7500 − 12000 − 6000 − 3000

[0109] We observe that this system is overdetermined, with 5 equations for 3 unknowns. Nevertheless, with respect to the hypotheses, we observe that the solution c7=c8=c9=1 respects the system.

[0110] By performing the solution as observations are made, we obtain the following solution vector: Observation m=1 → c7=1.316 / c8=0.7895 / c9=0.5263 Observation m=2 → c7=1.226 / c8=0.6604 / c9=0.9434 Observation m=3 / m=4 / m=5 → c7=1 / c8=1 / c9=1

[0111] When m < 3, the system has fewer equations than unknowns, and several solutions are possible. For m = 3, there is a unique solution, a system of 3 equations with 3 unknowns. For m > 3, there are more equations than unknowns, and the system is overdetermined. Nevertheless, we observe the convergence of the solution on this perfect system, which corresponds to the unique and perfect solution. The matrix conditioning W is from 2.69 at the end of the optimization, which is very good.

[0112] Case 2: Imperfect system (sensors not perfectly at 120°, noise in the measurements leading to a non-unit sum). Therefore, the following tests are noted: Wc = F ⇔ − 2505 − 1498 − 995 − 3493 − 2008 − 2012 1004 − 6010 − 7009 1511 − 8504 1001 995 − 2010 − 2005 c 7 c 8 c 9 = − 5003 − 7499 − 12011 − 5999 − 2998

[0113] By performing the solution as observations are made, we obtain the following solution vector: Observation m=1 → c7=1.318 / c8=0.7881 / c9=0.5235 Observation m=2 → c7=1.226 / c8=0.6763 / c9=0.9245 Observation m=3 → c7=0.9942 / c8=1.033 / c9=0.9706 Observation m=4 → c7=0.9984 / c8=1 / c9=0.9987 Observation m=5 → c7=0.995 / c8=1 / c9=0.998

[0114] The final conditioning of W is 2.682. We observe that the solution converged from approximately m=4. We find a solution very close to 1, but not exactly, because measurement noise has been taken into account, demonstrating the robustness of the approach. The result is therefore more precise.

[0115] In practice, the result is therefore more robust when calculating Fz (and by extension Fx and Fy) via optimization because we take into account all the disturbances and inaccuracies related to the industrial environment (measurement noise, mechanical tolerances, rigidity defects etc...).

[0116] It is worthwhile to verify the validity of the calibration obtained through welding or machining tests, for example.

[0117] For this reason, the installation includes, with reference to the figure 19 rigid fixing elements 100 for a plate 102 on the upper part of the sensor 94, the mounting plate 96 having been removed beforehand if necessary. The plate 102 is a metal plate that can undergo an operation similar to those provided for in production with the electrospindle.

[0118] During the verification, with the plate thus fixed to the sensor, a friction-mixing line is produced on the plate using the electrospindle (or alternatively, machining).

[0119] During the line construction stage, the forces Fx, Fy, and Fz on the friction-mixing tool are measured simultaneously using the multi-axis sensor 94 and the forces F1, F2, and F3 seen by the three sensors 90 of the electrospindle.

[0120] Then we calculate the forces Fx, Fy, and Fz exerted on the tool along the X, Y, and Z directions as a function of the forces F1, F2, and F3 of the sensors.

[0121] Finally, we compare the forces F1, F2 and F3 measured on the tool and the forces F1, F2 and F3 calculated, which allows us to determine if the system has been calibrated correctly.

[0122] Thus, by correlating the forces measured in the tool's reference frame (the result of the optimization procedure) with the forces measured by the 6-axis sensor, the entire method can be verified. Tests can also be performed when the electrospindle is disassembled, or to check for any drift over time.

[0123] This method also has the advantage of taking into account any potential rigidity problems in the electrospindle by direct measurement at the tool level during calibration.

[0124] The process can be used in an automated manner when using the electrospindle for production and also during calibration.

[0125] Many modifications can be made to the invention without going outside its scope.

[0126] It may be possible, independently of the implementation of the connector(s), to provide that the tool holder includes the thermocouple 38 and to provide a support carrying the thermocouple 38 and screwed into a recess 50 of the tool.

[0127] In one embodiment, beforehand: Over several observations, forces Fx, Fy and Fz are measured on the tool in three directions and forces F1, F2 and F3 on the base; and a vector of unknowns is determined. c optimal solution for the following problem: F OUTIL ∼ MF CAPTEURS ⇔ F x F y F z ∼ c 1 c 2 c 3 c 4 c 5 c 6 c 7 c 8 c 9 F 1 F 2 F 3 Or : F OUTIL is a vector of forces Fx, Fy and Fz; Mis a calibration matrix containing the values ​​of the vector of unknowns c ; And F SENSORS is a vector of forces F1, F2 and F3.

[0128] We can replace the consideration of the forces Fx, Fy and Fz by considering two forces from among the forces Fx, Fy and Fz, or by considering only one force from among the forces Fx, Fy and Fz.

[0129] We can predict that we will determine the vector c using the equation Wc = F

[0130] Or W is a matrix comprising all measured F1, F2 and F3 forces.

[0131] Advantageously, a conditioning of the matrix W is determined after each observation.

[0132] We can predict that after each observation we will determine whether we observe a convergence of the vector of unknowns. c.

[0133] It can be anticipated that, beforehand, a reference point on the electrospindle is aligned with a reference point on a sensor measuring the forces on the tool.

[0134] We can predict that: A friction-mixing line is created on at least one plate; during the line creation step, the forces on the tool and the forces on the base are measured simultaneously; the forces on the tool are calculated as a function of the forces on the base; and the forces measured on the tool and the calculated forces are compared. DESCRIPTION OF THE FIGURES

[0135] We will now present one embodiment of the invention by way of non-limiting example, supported by the drawings in which: THE figures 1 to 3 are perspective, bottom, and axial cross-sectional views of an electrospindle according to an embodiment of the invention; the figure 4is a larger-scale view of part of the electrospindle from the previous figure; the Figures 5 And 6 are views analogous to the figure 3 showing the electrospindle equipped with a tool in a rotating shoulder configuration; the figure 7 is a view of one end of the tool in one embodiment variant; the figure 8 is a view analogous to the figure 6 showing the electrospindle equipped with a tool in a fixed shoulder configuration; the figures 9 And 10 These are axial cross-sectional views of the electrospindle showing details of the cooling circuit; the figure 11 is a view illustrating the operating principle of the electrospindle during a friction-mixing operation; the Figures 12 and 13 illustrate a force sensor of the electrospindle; the Figures 14 And 15 These are views from below and in elevation of the electrospindle showing the position of the reference marks; the figures 16 And 17are views of the machine with the electrospindle from the previous figures and its robot during a calibration operation during the implementation of the process of the invention; the Figures 18 and 19 are perspective views illustrating two configurations of the machine's calibration components; and the Figure 20 is a perspective view of one of the electrospindle connectors.

[0136] We will describe, with reference to the figures, an embodiment of an electrospindle according to the invention. General presentation

[0137] The toolless electrospindle 2 is shown on the figures 1 to 4 , 9 And 10 The same electrospindle 2, equipped with a tool and in a rotating shoulder configuration, is illustrated in the Figures 5 And 6 The same electrospindle 2, equipped with a tool and in a fixed shoulder configuration, is illustrated in the figures 8 And 14 à 17 . Temperature measurement in rotating shoulder mode

[0138] With particular reference to the figure 3 , the electrospindle comprises a housing 4 and a rotating part 6 mounted to rotate relative to the housing around an axis, most often vertical, ZZ which constitutes a main axis of the electrospindle.

[0139] In this case, an upper part of the housing consists of an interface plate for robot 10, visible at the figure 16 forming a base and intended to support the electrospindle. It is connected to another part 11 of the housing, forming a support for the rotating part, by a ball joint through which the axis ZZ passes.

[0140] The rotating part 6 includes a shaft 12 which has at its lower axial end a housing 14 for receiving a tool holder 16. When the tool holder is received in the housing, it is made rigidly fixed to the shaft by conventional fastening means which will not be detailed here so that it rotates as a single unit with the rotating part.

[0141] The electrospindle includes bearings, in this case three in number, formed by ball bearings 18 mounted in an "O" configuration for the support and guidance of the rotational movement of the rotating part 6 relative to the housing 4. In this example, two of the bearings are located in the lower part of the shaft and the third is located in the upper part of the shaft.

[0142] The electrospindle includes an electric motor 20 capable of rotating the shaft relative to the housing.

[0143] The rotating part 6 comprises, in this case, a lower flange 22 rigidly fixed to a lower axial end of the shaft 12. The flange is annular in shape, such that it has a central recess providing access to this end, in particular to the housing 14, from outside the electrospindle. The flange 22 has a flat lower external face 24 perpendicular to the axis ZZ. The flange is surrounded by an annular cap 23 rigidly attached to the housing.

[0144] The flange 22 includes two connectors 26 arranged on the outer face of the flange and configured for receiving an electrical or electronic signal. They are positioned symmetrically on either side of the axis ZZ. Each connector 26 is equipped with a leaf spring 27, in this case two of them, so that it can receive two cables in two respective slots 29 of the connector. To use it, the spring 27 is actuated, for example by means of a screwdriver, a cable is inserted into the corresponding slot 29, and the spring is released to lock the cable in the connector.

[0145] The electrospindle includes means for wirelessly transmitting the signal from each connector between the rotating part 6 and the housing 4. These means include an annular electronic circuit board 28, rigidly attached to the rotating part. In this example, the board is housed on a shoulder formed on the upper part of the shaft, which extends opposite the housing. The circuit board 28 is arranged so that its electrical current is supplied by induction due to the movement of the rotating part, thus eliminating the need for a battery.

[0146] The transmission means also include an annular electronic card 30 rigidly attached to the housing and coaxial with the card 28. It extends opposite the shoulder so that the card 28 rotates in front of the card 30 of the housing when the shaft rotates.

[0147] We therefore observe that these means are located inside the electrospindle and closer to an upper axial end of the shaft than to its lower axial end which has housing 14.

[0148] The signal is transferred wirelessly from the rotating board 28 to the fixed board 30. The fixed board 30 is powered by a DC or AC voltage from the machine, and the power is transferred to the rotating board 28 by induction, thus eliminating the need for a battery in the rotating part. A distance of between 4 and 8 mm is recommended to ensure proper operation of the induction system. Therefore, there is no maintenance required, and the system can operate continuously.

[0149] The shaft 12 has two straight conduits 32, each extending in a direction parallel to the axis ZZ, at a distance from it. The conduits extend from the respective connectors 26 to the card 28 of the shaft. They house cables (not shown) connecting each connector to the card 28.

[0150] THE Figures 5 And 6 illustrate a tool holder 16 received in the housing 14. The tool holder comprises a support formed here by a chuck 34 which carries a tool 36 at its lower axial end. The sleeve extends outwards from the flange along the axial direction ZZ. Similarly, the tool 36 extends outwards from the chuck 34 along the same direction.

[0151] The tool holder 16 includes a thermocouple 38. The thermocouple includes a main body 40 carrying at an upper axial end a cable 42 and at an opposite lower axial end an elongated sheath 44.

[0152] The cable 42 passes through a lateral conduit 46 of the chuck so that it exits the tool holder, outside the electrospindle opposite the flange 22. The cable 42 at its free end can be connected to one of the connectors 26 of the flange 22.

[0153] The tool 36, for example, is made of tungsten carbide or steel. It also has a generally cylindrical shape. Its upper part features a threaded cavity 50 into which a threaded portion of the thermocouple body 40 is screwed. The thermocouple is thus rigidly fixed to the tool. The threaded hole at the rear of the tool allows adjustment of the thermocouple's position when fitted with a threaded holder, eliminating the need to adjust it based on the length of the sheath or tolerance errors (e.g., the length of the tool 36). The connection between the thermocouple sheath and the cable is made within the threaded holder. A nut 56 at the rear secures the thermocouple in position.

[0154] The thermocouple sheath 44 extends into a first conduit 52 of the tool, following the cavity 50, and then, at the lower axial end of the tool, into a second conduit 54, or terminal conduit, having a diameter smaller than a smaller diameter of the cavity 50. The diameter of this second conduit 54 is just sufficient to allow the passage of the thermocouple sensor. This conduit extends into the tool pin intended to be embedded in the material during welding. In this case, this terminal conduit opens outside the electrospindle. Such a conduit with a very small diameter in front of the tool prevents weakening it while positioning the thermocouple as close as possible to the welded core.

[0155] The thermocouple sensor has a diameter of 3 mm or less, and preferably 0.5 mm or less. The smaller the thermocouple diameter, the greater the responsiveness. Furthermore, using a hot junction connected to ground also improves responsiveness. Here, the thermocouple sheath diameter is 0.5 mm. The table below shows the associated response time. The time is reduced tenfold between a 0.5 mm and a 1.5 mm sheath. Having the hot junction connected to ground provides an additional 33% performance gain. A 0.25 mm sheath would be even better, but is more complex to manufacture mechanically and susceptible to junction fragility. Table: Response time according to sheath diameter Diameter (mm) Response time(s) 0,25 0,015 0,5 0,03 1,0 0,15 1,5 0,3 2,0 0,4 Table: Response time according to welding method Diameter (mm) Time(s) 63.2% Time(s) 90% Insulated hot solder 0,25 0,015 0,050 0,5 0,030 0,082 Hot soldering to ground 0,25 0,007 0,033 0,5 0,020 0,060

[0156] Alternatively, as illustrated in the figure 7The terminal conduit 54 of tool 36 may not be open. The thermocouple sheath may therefore reach a mechanical stop. However, this complicates manufacturing because creating a blind conduit of small diameter over a long length is difficult.

[0157] Generally, welding tools are made of tool steel, allowing for easy manufacturing. Often, the sleeve 34 can be a single piece with the tool 36. However, for certain applications (high-yield-strength aluminum, for example), the material of the pin forming the tool 36 can be a grade with better mechanical properties than the tool holder 34. This is the case here, where the welding tool 36 is divided into two mechanical parts: a pin and a pin holder, with the pin inserted into the pin holder. This is called a hybrid tool because only the functional part is made of carbide. The advantage is that obtaining a central hole is completely standardized and achieved by shrink-fitting (drilling a carbide-like material is very difficult because it is a very hard material).Depending on the needs, this hole can be plugged once the thermocouple sheath has been put in place, for example using a thermal sealing cement that can withstand high temperatures (>600°C).

[0158] The two connectors 26 allow the user to connect two thermocouples (for example, one in the welding tool and another in the chuck). Since they are permanently fixed to the shaft, periodic rotational rebalancing is unnecessary. The symmetry of the two connectors 26 naturally improves the balance (which would not be the case with a single connector).

[0159] These elements of the temperature measurement system are fully integrated into the electrospindle and make it extremely compact.

[0160] In this embodiment, it is an ultra-compact, high-performance industrial electrospindle for use in both research environments and high-volume production, compatible with FSW, SSFSW, and machining, with or without temperature measurement. This principle can be extended, in particular, to measure the temperature in machining cutters.

[0161] In this friction-mixing electrospindle, the tool 36 mounted in the chuck 34 illustrated in the figure 6 features an external shoulder mounted to rotate relative to the housing. Fixed shoulder mode

[0162] We illustrated at the figure 8 The same electrospindle, this time in a fixed shoulder configuration. We see the electrospindle 2, the tool holder 16, the chuck 34, and the tool 36 with all their components as in the previous figures.

[0163] This time the electrospindle also carries a cover or shoulder 60 covering the entire lower part of the electrospindle, in particular the rotating flange 22, including the chuck 34 and most of the tool 36. This cover 60 is rigidly fixed to the housing in this case by screws 62. In the present example, it is not fixed directly to the cap 23. It must be put in place after the chuck 34 and the tool 36 have been received on the spindle and the thermocouple cable(s) have been connected to the connectors 26 of the flange. The cover is not in direct contact with the tool holder 16, because an air gap is provided between them, opposite the entire external surface of the tool holder and the rotating flange 22. It is this cover which forms the fixed shoulder at its lower axial end from which the rotating end of the tool 36 emerges.

[0164] It can be observed that the loads received by the fixed shoulder pass directly through the housing or frame of the electrospindle without passing through the flange or the shaft since there is no direct support on the outer ring of the front bearings, which provides great robustness to the assembly.

[0165] We can therefore see that this solution for temperature measurement is universal and can be applied to FSW and SSFSW welding tools, and in different materials. On the rotating joint

[0166] We will now present the cooling circuit of electrospindle 2.

[0167] With reference to figures 9 And 10The housing 4 includes a cooling channel 70, in this case straight and horizontal. The shaft 12 also includes at least one cooling channel 72, in this case straight and vertical, centered on the axis ZZ and opening at its lower axial end into the housing 14 of the tool holder, as can be seen in Figures 5 And 6 .

[0168] The electrospindle 2 includes a rotary joint 74 comprising a fixed upper portion 76 rigidly attached to the housing 4 and a rotating lower portion 78 rigidly attached to the shaft 12. The joint connects the fluid channel 70 of the housing to the channel 72 of the shaft. The portion of the joint 76 attached to the housing and the cooling channel 70 of the housing extend within the robot interface 10. This design allows for a very compact assembly, as the fixed portion of the rotary joint is part of the interface 10, eliminating the need for additional components. The rotary joint is of a universal type. It allows the transfer of a fluid (such as air, lubricant, or a microlubrication fluid, which is an air / oil mixture) from the fixed element to the rotating element. The two parts of the joint rub against each other to ensure a dynamic seal, preventing any leakage or loss.This system guarantees a 100% leak-proof fluid supply to the front section of the electrospindle. All the electrospindle's bearings are located outside the seal, which itself is bearing-free. The rotary seal 74, for example, is the one sold by Deublin under part number 1101-632-343.

[0169] As illustrated in the Figure 10 The shaft 12 further includes drainage channels 80, here six in number, extending from an axial upper face of the shaft located opposite the housing to the cooling channel 72 of the shaft. Two of these channels are visible on the Figure 10Each drainage channel 80 forms a bend and comprises a first straight vertical section opening at its upper end onto the top face of the shaft opposite the housing, and an inclined section whose lower end opens into the cooling channel 72. The drainage channels 80 allow the fluid to flow by gravity through the central passage of the shaft. This eliminates the need for a specific and bulky drainage cable. These channels evacuate fluid that would otherwise be present at the interface between the fixed part 76 of the rotary joint and its rotating part 78 (thus preventing the accumulation of oil-like fluid in the upper area).

[0170] With particular reference to the figure 6The tool holder chuck 34 includes four straight drain channels 46, connected upstream to the cooling channel 72 of the shaft and downstream to the external environment of the electrospindle. The downstream connection is made without passing through the shaft or the housing. The terms "upstream" and "downstream" refer to the direction of the cooling fluid flow. Each drain channel 46 has a downstream end that opens onto an external face of the chuck. One of these channels houses the thermocouple cable 42.

[0171] In this case, the exhaust ducts 46 each communicate directly with the cooling duct 72. The exhaust ducts 46 and the cooling duct intersect at the same point. Two of the exhaust ducts 46 extend downwards, therefore in the opposite direction to the electrospindle, from the cooling duct 72. The other two exhaust ducts 46 extend upwards towards the electrospindle from the cooling duct.

[0172] The machine includes a component not shown for supplying the cooling duct 70 of the crankcase with a cooling fluid.

[0173] The fluid, such as air, enters through the center of the shaft. The tool holder clamping system on the shaft (for example, the one marketed under reference HSK 63 by Semar, which forms a standardized tool holder mounting point) delivers air in a sealed manner to the tool holder. The tool holder's exhaust channels 46 allow the air to circulate and be expelled from the electrospindle. A tool holder 34 could be equipped with a cooling labyrinth to provide convective cooling over a large surface area, which would be even more efficient. The principle remains the same.

[0174] These elements are valid regardless of the operating mode, FSW and SSFSW welding, as can be seen on the figure 8or machining. Depending on the operating mode, the cooling fluid is either air or a liquid such as a lubricant. In the electrospindle, in fact, all modes are possible: dry running, with air, with micro-lubrication, or even with full lubrication. The user doesn't need to worry about long-term operation. In machining mode, this setup allows for liquid delivery to the tool, or even to the cutting edge of the tool (as provided by HAIMER, with a chuck equipped with cooljet technology, for example), for micro-lubrication, resulting in more efficient machining, lubricant savings, and a cleaner work area. Measuring efforts

[0175] As illustrated in Figures 1 , 3 , 5 And 12 à 15The electrospindle 2 includes three force sensors 90. Each sensor 90 is interposed between the base 10, formed by the interface plate attached to the robot, and the support 11 located below it. It is configured to measure only the force exerted by the support 11 on the base 10 in a direction parallel to the ZZ axis. These are beam-type sensors, as illustrated in the diagrams. Figures 12 and 13 The operating principle of such a sensor is very simple. It only measures a force F in one direction (positive and negative) as represented on the figure 13 This sensor has the advantages of being economical (it costs less than 200 euros), providing flexibility in design by freeing up the entire center of the electrospindle, and having a minimalist footprint and extremely good robustness.

[0176] The three sensors 90 extend in the same plane perpendicular to the axis. They are here the image of each other by a rotation around the axis by an angle of 120° and are therefore distributed homogeneously on the periphery of the electrospindle.

[0177] Here, we have planned three sensors taking into account the dimensions of the electrospindle, but we could have planned 4, 5, 6 or even more depending on the mechanical possibilities.

[0178] We consider the frame of reference associated with tool 36. It is in this frame of reference that the forces exerted on the tool during the use of the electrospindle must be calculated. With reference to the figure 14 With the three-sensor system 90, we make the following assumptions: a perfect 120° angle between the three sensors; the measurement centers coincide with the main axis ZZ; and the electrospindle is considered infinitely rigid between the tool and the sensors.

[0179] The axial force is therefore the sum of the three forces measured by the sensors, i.e., Fz = F1 + F2 + F3. For the measurement to be consistent, the system has a degree of freedom of movement via the aforementioned ball joint (centering on a line) between the base 10 and the support 11, which prevents excessive static indeterminacy that would prevent the sensors from deforming. However, in reality, at this stage, complete control is not achieved at the center of the tool 36 because the center of the sensor system does not perfectly coincide with the tool's reference frame (the sensors are fixed by screws in the support 10, requiring a functional mounting clearance). Therefore, we are simply performing a summation of three forces without an exact reference frame, yielding a result with a certain level of accuracy, but not optimal. Furthermore, since the sensors are located on the upper part of the electrospindle, the loss of stiffness of the tool up to the sensors is not taken into account.

[0180] Furthermore, we lack knowledge of the Fx and Fy forces. While knowledge of Fx and Fy forces is not mandatory in FSW / SSFSW or machining, it is beneficial for several reasons. Firstly, these two forces subject the tool in the material to shear stress. Therefore, they must remain within an acceptable range relative to the chosen tool to prevent breakage during use. Secondly, this knowledge allows for the real-time detection of abnormal overloads that could damage the electrospindle or the machine. Moreover, in a variant of the SSFSW process called Corner SSFSW (corner welding), these two forces must be controlled in the same way as the Fz force.

[0181] Therefore, it is useful to calibrate the three-sensor beam 90 arrangement to determine the forces Fx, Fy, and Fz in the tool's frame of reference. If desired, this calibration could be performed for only one or two forces; the system is thus multivariable and fully generalizable (solving an optimization system). The situation is illustrated here on an electrospindle with three sensors, but generalization to "n sensors" is possible.

[0182] The machine thus includes calculation means comprising at least one computer or automated calculation system capable of calculating forces Fx, Fy and Fz exerted on the tool 36 along the Z direction and in two directions X and Y perpendicular to the axis as a function of the forces F1, F2 and F3 measured by the sensors 90 along the Z direction but at a distance from the axis.

[0183] Thus, in the process of controlling the electrospindle during its use for welding or machining: We measure at three different locations, i.e. in the three sensors 90, the respective forces F1, F2 and F3 exerted by the support 11 on the base 10 in a direction parallel to the axis ZZ, then we calculate the forces Fx, Fy and Fz exerted on the tool 36 along the directions X, Y and Z as a function of the measured forces F1, F2 and F3.

[0184] For calibration, with reference to figures 16 to 19The installation comprises a base 92 and a multi-axis sensor 94 rigidly fixed to the base. This sensor is capable of measuring forces in three directions simultaneously, these three directions forming an orthogonal coordinate system. The sensor 94 is securely anchored to the base to allow for the generation of significant forces and obtain useful measurement ranges. This provides a complete 6-axis sensor. This is a major advantage of the proposed calibration system, as it eliminates the need for pulley systems to load the system, making it extremely flexible and precise in an industrial setting.

[0185] To avoid damaging the force sensor, the installation also includes an interface plate 96 rigidly fixed to the multi-axis sensor 94 and featuring a central guide channel opening from an upper side of the interface plate opposite the sensor to insert a calibration tool as on the figure 17 This tool replaces the welding or machining tool in the electrospindle during calibration.

[0186] The base includes 98 calibration points that allow for leveling the entire sensor relative to the electrospindle. This enables alignment of the measurement axes to match the force plate measurements with the robot's tool reference. The key is to ensure a consistent transformation between the measurement references. Vision targets can also be used for increased accuracy.

[0187] Using these elements, the preliminary calibration process is implemented as follows.

[0188] With reference to the figure 18 There are two important points of reference: the 6-axis force sensor reference point (in red) which is fixed; and the robot tool reference point;

[0189] The goal is to achieve a coincidence between the two reference frames. (Alternatively, a known transformation between them can be used). The advantage of this method is that it takes into account the positioning uncertainties of the three beam sensors on the electrospindle and the potential loss of stiffness in the electrospindle, and calibrates the entire assembly relative to the tool reference frame and its center.

[0190] To do this, first of all, the electrospindle 2 mounted on the robot 10 is equipped with the calibration tool. This calibration tool can be mounted either in the tool holder 34 in place of the welding tool 36, as referenced in the figure 6, either through the shoulder carrier 60 with reference to the figure 8 In the first case, the charge will pass through the bearings of the electrospindle; in the second, it will pass directly through the body 4 of the electrospindle. This is a tool 36 similar to the welding tool but of lower quality and a simplified shape with reduced clearance, which is inserted into a guide channel of the interface plate 96, thus allowing the precise alignment of the two reference marks.

[0191] Using the robot, the tool is brought successively close to each of the three calibration points 98 in order to align, in the calculation methods, the X, Y, Z coordinate system of the electrospindle, originating at the free end of the tool 36, with the coordinate system of the multi-axis sensor 94, originating in the central conduit. The two coordinate systems can thus be superimposed in the calculation methods as illustrated in the figure 18 .

[0192] Next, with tool 36 housed in the conduit of sensor 94, the tool is moved along the XYZ directions to generate contact between tool 36 and sensor 94, producing reaction forces from the sensor against the tool. Indeed, once the tool is inserted into the center of the interface, small movements along the directions of the tool coordinate system (or the sensor coordinate system) can be used to apply stress to the system. The data collection procedure is therefore extremely simplified; it does not require a pulley system or the use of dynamometers to apply stress and measure forces. The procedure should allow for various observations and the scanning of a wide range of forces to obtain optimal conditioning of the optimization problem and find a relevant solution. The result is a calibration matrix linking the two force measurement systems.

[0193] During an observation, the forces Fx, Fy, and Fz exerted on the tool by sensor 94 are simultaneously measured in three directions. The forces F1, F2, and F3 seen by sensors 90 on the electrospindle are also measured along the ZZ direction. A certain number of observations (denoted m hereafter) are necessary to obtain optimal accuracy of the solution.

[0194] The detailed implementation, for example, proceeds as follows in the case of n beam-type sensors. Suppose we have a dataset of m force observations F i = F x i , F y i , F z i T ∈ ℝ for m conditions x i = F 1 i , F 2 i , . . , Fn i ∈ ℝ , i = 1 , .. , m We would like to approximate the vector F i as a function of x i = [F1 i ,F2 i ,..,Fn i ] (overdetermined system in the case where m>n, a unique solution does not exist, so we approximate the solution): F ∼ f x c where the function f depends linearly on p=3xn coefficients c = c 1 , c 2 , . . , c p T ∈ ℝ The regression problem consists of determining the p coefficients [c1, ..., cp] that will minimize (in a direction to be specified) the distance between F and the prediction f ( x From an algebraic point of view, the regression problem can be rewritten as follows; we would like to find a solution to an overdetermined system: Wc = F

[0195] Or : W is a matrix of dimensions (3m x 3n) and m > n that depends on the data conditions x i = F 1 i , F 2 i , . . , Fn i ∈ ℝ , i = 1 , .. , m , c is the vector of unknown coefficients of dimensions (3n x 1) and F is the data vector of observations with dimensions (3m x 1).

[0196] The least squares problem boils down to finding the vector of coefficients c which minimizes the (Euclidean) distance that separates WC of F, that is to say, to find c which minimizes the Euclidean norm F - Wc.We detail here the solution for three beam-type sensors, i.e., n=3. From a practical point of view, we seek to solve the following problem: F OUTIL ∼ MF CAPTEURS ⇔ F x F y F z ∼ c 1 c 2 c 3 c 4 c 5 c 6 c 7 c 8 c 9 F 1 F 2 F 3 Or : F TOOL the vector of forces in the tool frame (X, Y, Z); M the calibration matrix containing the values ​​of the vector c ; And F SENSORS the vector of forces from the beam sensors.

[0197] The goal is to determine "the best" calibration matrix M by solving an optimization problem by making m observations.

[0198] For m=4>3 observations, the regression problem given by equation (3.2) can be written as: Wc = F ⇔ F 1 1 F 2 1 F 3 1 0 0 0 0 0 0 0 0 0 F 1 1 F 2 1 F 3 1 0 0 0 0 0 0 0 0 0 F 1 1 F 2 1 F 3 1 F 1 2 F 2 2 F 3 2 0 0 0 0 0 0 0 0 0 F 1 2 F 2 2 F 3 2 0 0 0 0 0 0 0 0 0 F 1 2 F 2 2 F 3 2 F 1 3 F 2 3 F 3 3 0 0 0 0 0 0 0 0 0 F 1 3 F 2 3 F 3 3 0 0 0 0 0 0 0 0 0 F 1 3 F 2 3 F 3 3 F 1 4 F 2 4 F 3 4 0 0 0 0 0 0 0 0 0 F 1 4 F 2 4 F 3 4 0 0 0 0 0 0 0 0 0 F 1 4 F 2 4 F 3 4 c 1 c 2 c 3 c 4 c 5 c 6 c 7 c 8 c 9 = F x 1 F y 1 F z 1 F x 2 F y 2 F z 2 F x 3 F y 3 F z 3 F x 4 F y 4 F z 4

[0199] Problem (3.4) is overdetermined (12 equations for 9 unknowns). The matrix Wis not square, and has dimensions (12 x 9). A unique solution does not exist, but there is a least-squares solution. Consider the following squared error: ε = F − Wc 2 = c T W T Wc − 2 c T W T F + F T F

[0200] The vector of unknowns c is obtained by minimizing equation (3.5). Since the function is convex, the solution is obtained by solving the following partial differentiation: ∂ ε ∂ c = 2 W T Wc − 2 W T F = 0

[0201] The vector of solutions c is given by: c = W + F Or W +< = eW T< W) -1< WT< is the pseudo-inverse or generalized inverse of W. It corresponds to an approximate solution of the linear system (3.4) minimizing the squared error ε = (F - WC ) 2< . The calibration matrix M can then be formatted to solve equation (3.3).

[0202] In order to achieve convergence of the solution vector and obtain efficient and robust calibration, it is necessary to perform as many observations m as possible. It is trivial to say that the greater the number of observations m, the more equation (3.3) tends towards accuracy rather than approximation. Similarly, it is trivial to say that the greater the number of beam sensors, the greater the accuracy. The arrangement of the possible number of sensors is strongly related to the maximum permitted space. To obtain a coherent mathematical problem condition (matrix condition), W), It is necessary to cover a fairly wide range of efforts during the different observations in order to have maximum sensitivity on the measurements.

[0203] In analysis, conditioning measures the dependence of the solution to a numerical problem on the data, in order to verify the validity of the solution. Indeed, the data generally depend on experimental measurements and are therefore subject to errors. Dependence is most often a numerical quantity. More generally, conditioning is a measure of the numerical difficulty of the problem. A problem with low conditioning is said to be well-conditioned, and a problem with high conditioning is said to be poorly-conditioned.

[0204] Example (calculations performed using MATLAB software): This demonstrates the relevance of the proposed method by considering the subsystem containing only Fz (the Fx and Fy values ​​can be difficult to interpret without actual testing, but the principle remains the same; the example is intended to be illustrative). The number of beam sensors is three.

[0205] Indeed, due to the arrangement of the sensors on the electrospindle, the result should be very close to a unit summation for Fz, i.e., Fz = F1 + F2 + F3. We will now demonstrate the advantages and robustness of the measurement noise injection method, and show that the assumption is not accurate.

[0206] Considering only the 3rd line, equation (3.3) becomes: F z = c 7 c 8 c 9 F 1 F 2 F 3

[0207] For m=5>3 observations, the regression problem given by equation (3.2) can therefore be written: Wc = F ⇔ F 1 1 F 2 1 F 3 1 F 1 2 F 2 2 F 3 2 F 1 3 F 2 3 F 3 3 F 1 4 F 2 4 F 3 4 F 1 5 F 2 5 F 3 5 c 7 c 8 c 9 = F z 1 F z 2 F z 3 F z 4 F z 5

[0208] Case 1: perfect system (sensors at 120°, measurement center on the electrosh axis, perfect and unity summation). The following tests are therefore noted: Wc = F ⇔ − 2500 − 1500 − 1000 − 3500 − 2000 − 2000 1000 − 6000 − 7000 1500 − 8500 1000 1000 − 2000 − 2000 c 7 c 8 c 9 = − 5000 − 7500 − 12000 − 6000 − 3000

[0209] We observe that this system is overdetermined, with 5 equations for 3 unknowns. Nevertheless, with respect to the hypotheses, we observe that the solution c7=c8=c9=1 respects the system.

[0210] By performing the solution as observations are made, we obtain the following solution vector: Observation m=1 → c7=1.316 / c8=0.7895 / c9=0.5263 Observation m=2 → c7=1.226 / c8=0.6604 / c9=0.9434 Observation m=3 / m=4 / m=5 → c7=1 / c8=1 / c9=1

[0211] When m < 3, the system has fewer equations than unknowns, and several solutions are possible. For m = 3, there is a unique solution, a system of 3 equations with 3 unknowns. For m > 3, there are more equations than unknowns, and the system is overdetermined. Nevertheless, we observe the convergence of the solution on this perfect system, which corresponds to the unique and perfect solution. The matrix conditioning W is from 2.69 at the end of the optimization, which is very good.

[0212] Case 2: Imperfect system (sensors not perfectly at 120°, noise in the measurements leading to a non-unit sum). Therefore, the following tests are noted: Wc = F ⇔ − 2505 − 1498 − 995 − 3493 − 2008 − 2012 1004 − 6010 − 7009 1511 − 8504 1001 995 − 2010 − 2005 c 7 c 8 c 9 = − 5003 − 7499 − 12011 − 5999 − 2998

[0213] By performing the solution as observations are made, we obtain the following solution vector: Observation m=1 → c7=1.318 / c8=0.7881 / c9=0.5235 Observation m=2 → c7=1.226 / c8=0.6763 / c9=0.9245 Observation m=3 → c7=0.9942 / c8=1.033 / c9=0.9706 Observation m=4 → c7=0.9984 / c8=1 / c9=0.9987 Observation m=5 → c7=0.995 / c8=1 / c9=0.998

[0214] The final conditioning of W is 2.682. We observe that the solution converged from approximately m=4. We find a solution very close to 1, but not exactly, because measurement noise has been taken into account, demonstrating the robustness of the approach. The result is therefore more precise.

[0215] In practice, the result is therefore more robust when calculating Fz (and by extension Fx and Fy) via optimization because we take into account all the disturbances and inaccuracies related to the industrial environment (measurement noise, mechanical tolerances, rigidity defects etc...).

[0216] It is worthwhile to verify the validity of the calibration obtained through welding or machining tests, for example.

[0217] For this reason, the installation includes, with reference to the figure 19 rigid fixing elements 100 for a plate 102 on the upper part of the sensor 94, the mounting plate 96 having been removed beforehand if necessary. The plate 102 is a metal plate that can undergo an operation similar to those provided for in production with the electrospindle.

[0218] During the verification, with the plate thus fixed to the sensor, a friction-mixing line is produced on the plate using the electrospindle (or alternatively, machining).

[0219] During the line construction stage, the forces Fx, Fy, and Fz on the friction-mixing tool are measured simultaneously using the multi-axis sensor 94 and the forces F1, F2, and F3 seen by the three sensors 90 of the electrospindle.

[0220] Then we calculate the forces Fx, Fy, and Fz exerted on the tool along the X, Y, and Z directions as a function of the forces F1, F2, and F3 of the sensors.

[0221] Finally, the forces F1, F2, and F3 measured on the tool are compared with the calculated forces F1, F2, and F3, which allows us to determine if the system has been calibrated correctly. Thus, by comparing the forces measured in the tool's reference frame (the result of the optimization procedure) with the forces measured by the 6-axis sensor, we can verify the entire method. We can also perform tests when the electrospindle is disassembled or check for drift over time.

[0222] This method, not included in the present invention, also has the advantage of taking into account possible rigidity problems in the electrospindle by direct measurement at the tool level during calibration.

[0223] The process can be used in an automated manner when using the electrospindle for production and also during calibration.

[0224] The invention can be modified in many ways without departing from its scope as defined in the claims.

[0225] It may be possible, independently of the implementation of the connector(s), to provide that the tool holder includes the thermocouple 38 and to provide a support carrying the thermocouple 38 and screwed into a recess 50 of the tool.

Claims

1. Electric spindle (2) comprising: - a housing (4) comprising at least one cooling duct (70), and - a shaft (12) mounted so as to rotate relative to the housing, the shaft having a receptacle (14) for receiving a tool holder (16) and at least one cooling duct (72) opening into the receptacle, characterised in that: the housing comprises an interface (10) for attaching the electric spindle to a robot, into which the cooling duct (70) extends, the electric spindle comprises a rotary joint (74) having an upper part (76) rigidly attached to the housing and a part (78) rigidly attached to the shaft, the upper and lower parts of the joint rubbing against each other, the rotary joint (74) fluidically connecting the cooling duct (70) of the housing and the cooling duct (72) of the shaft, the fixed part (76) of the rotary joint (74) being arranged in the interface (10) for attachment to the robot and being connected directly to the cooling duct (70).

2. Electric spindle according to the preceding claim, which comprises bearings (18), all the bearings being located outside the joint (74).

3. Electric spindle according to one of the preceding claims, wherein the shaft (12) further comprises at least one conduit (80) extending from a face of the shaft facing the housing (4) to the cooling duct (72) of the shaft.

4. Electric spindle according to one of the preceding claims, which comprises a tool holder (16) received in the housing, the tool holder comprising at least one discharge conduit (46) in fluidic communication with the cooling duct (72) and in fluidic communication with the environment outside the electric spindle without passing through the shaft.

5. Electric spindle according to claim 4, wherein the tool holder comprises at least two discharge conduits (46).

6. Electric spindle according to the preceding claim, wherein the discharge conduits (46) each communicate directly with the cooling duct.

7. Electric spindle according to one of claims 5 and 6, wherein the discharge conduits (46) and the cooling duct (72) intersect at a same point.

8. Electric spindle according to one of claims 4 to 7, wherein the discharge conduit or at least one of the discharge conduits (46) extends in the opposite direction to the electric spindle from the cooling duct.

9. Electric spindle according to any one of claims 4 to 8, wherein the discharge conduit or at least one of the discharge conduits (46) extends towards the electric spindle from the cooling conduit.

10. Electric spindle according to one of the preceding claims forming a friction stir welding electric spindle, for example comprising an external shoulder mounted fixedly relative to the housing (4) or lacking an external shoulder mounted fixedly relative to the housing.

11. Machine comprising an electric spindle according to one of the preceding claims, wherein the machine comprises a means for supplying the cooling duct of the housing (4) with a cooling fluid, the cooling fluid being air or a liquid.