Monitoring of Arc Welding by Analyzing Modulation of Radiation From Carrier Signals Superimposed on Weld Current

This invention will now be described by way of example with reference to the drawings, in which:

FIG. 1A illustrates an overpenetrated weld;

FIG. 1B illustrates an underpenetrated weld;

FIG. 1C illustrates a correctly penetrated weld;

FIG. 2A illustrates the mode of vibration of a penetrated weld;

FIG. 2B illustrates the mode of vibration of an underpenetrated weld; and

FIG. 3 illustrates an apparatus for use in carrying out the method of the invention.

Referring now to the drawings, FIG. 1A shows an overpenetrated weld which is sagging and possibly about to fall through, while FIG. 1B shows an underpenetrated weld which has a crevice behind the weld, which can be a position of weakness or corrosion. FIG. 1C shows a correctly penetrated weld which extends through the workpiece but which is not sagging.

For the method of the present invention, the weldpool is caused to vibrate, thus causing capillary waves on the weldpool. A weldpool produced in arc welding can be made to vibrate in its natural mode of oscillation by a short pulse in the welding current or a shockwave generated in the welding gas. FIGS. 2A and 2B show the fundamental modes of vibration of a penetrated and an unpenetrated weldpool. The frequency of vibration is determined by the weldpool's surface tension for practical weldpool sizes, and thus capillary waves are formed. The frequency of the fundamental vibration, .nu., can be calculated for a weldpool of mass M, frontface surface area A, surface tension .sigma. and density .rho.. ##EQU1## where, .chi. is a shape or form factor which is approximately unity for practical weldpools.

For unpenetrated weldpools: ##EQU2## where .chi. is the appropriate form factor for unpenetrated weldpools.

Thus, for the same material, the fundamental frequency depends on weldpool mass M for penetrated pools, and on the frontface pool area A for unpenetrated pools.

The distinction between penetrated and unpenetrated weldpools is made by examining their amplitude of vibration. Penetrated pools have a much larger amplitude of vibration than unpenetrated pools. This is due to the way they are excited into vibration. The spatial distribution of the excitation impulse favors larger amplitude vibrations in penetrated pools because the exciting force has a relatively flat profile over most practical weldpool sizes. The amplitude of vibration can be predicted for penetrated weldpools using a simple spring-dashpot model.

For oscillations excited by pulsing the welding current the amplitude is proportional to:

(3)

where I.sub.p is the pulse current (about 200 A in practice) which lasts for t.sub.p seconds (about 4 ms in practice), and I is the welding current.

For oscillations excited by a shock generated in the welding gas, the amplitude is proportional to:

where .DELTA.P is the pressure of the shock which lasts t.sub.p seconds and A is the frontface area of the pool.

The above expressions enable pool amplituded to be normalized, so that the normalized amplitudes are independent of the exciting impulse. The advantage of using short impulses is that all the frequencies of interest have the same normalization. Exciting impulses are chosen to be strong enough to create pool amplitudes of about 1 mm (peak to peak) in penetrated pools, thereby swamping any "noise".

The working frequency range for penetrated weldpools is from about 20 Hz to 80 Hz. The weldpool frequency at which the weldpool begins to sag depends on the quarter power of the ratio of the weldpool density to surface tension Thus, the "sagging" frequency is relatively insensitive to material properties and is close to 30 Hz for most materials. This is an important result for all positional welding.

The arc light emitted during welding is proportional to arc length, except at very low currents. Thus, as the weldpool vibrates the arc light is modulated in proportion to the arc length; if the amplitude of weldpool vibration is 1 mm when the arc length is 3 mm, the arc light is modulated by 33%.

There are two methods of normalization. Firstly, the detected intensity is normalized by dividing by the mean intensity. The signal is then equal to the amplitude of weldpool vibration divided by the arc length. Secondly, the signal is normalized by the strength of the excitation impulse. This is done by dividing by either expression (3) or expression (4) above, depending on whether the excitation is by pulsing the welding current or by a shock generated in the welding gas. The normalization means that calibration is not necessary.

The resolution of any frequency measurement depends on the time taken to make the measurement. If T is the time taken, a resolution .DELTA..nu., in frequency is obtained. Where:

While welding, a compromise is required between resolution in the weldpool size measurement and how frequently the measurement is made. For control of the welding process, a good compromise is a resolution of 2 Hz, which enables the weldpool mass to be measured to 10% for a 40 Hz pool frequency and the measurement updated every 1/2 second.

A weld operation is usually defined in terms of arc length or arc voltage, travel speed and welding current. The method of the present invention allows a weld operation to be defined in terms of penetration and weldpool size. Penetration is determined by comparing the amplitude of the normalized weldpool spectra with a threshold value below which the weldpool does not penetrate, and weldpool size is calculated from the fundamental frequency. The method of the invention can be carried out using an IBM PC XT computer to control the following steps:

1. Excite weldpool vibration.

2. Superimpose a carrier signal of at least 100 Hz on the arc current.

3. Detect radiation emitted by the arc as sound or light.

4. Normalize the detected radiation and calculate the spectrum.

5. Determine penetration and weldpool size from the spectrum.

6. Adjust welding current and repeat the cycle.

The cycle time should be slightly longer than the measurement time to allow the computer to perform the calculations needed in the last three steps and also to excite the weldpool Monitoring of the weldpool's fundamental frequency and amplitude of vibration is carried out and can be stored in a file for later use for quality assurance purposes, even if the welding operation employs constant current. Quality control is effected if the welding operation is specified in terms of a target weldpool frequency.

The method of the invention can be used with a variety of materials and geometries, to perform close butt welds in a single pass up to 6 mm thick. The method is tolerant of variations in thickness, heat sinks and fit-up, and can orbitally weld tubes.

FIG. 3 shows an apparatus for use in carrying out the method of the invention. As shown, an arc intensity detector D, comprising, for example, an incoherent glass fiber bundle shrouded in a copper tube, is mounted on the welding torch T and is pointed at the arc provided by the torch. The alignment is not critical; all that is required is that some of the emitted arc light be collected. The collected light is passed by way of an optical fiber cable to processing apparatus (not shown) for processing as discussed above.

If acoustic radiation is to be detected, then the detector D is a microphone rather than a glass fiber bundle detector.

It will be appreciated that both light and acoustic radiation emitted by the arc can be simultaneously detected using an optical fiber cable and a microphone, both types of detected radiation being analyzed as necessary.

The method of this invention is well suited to robotic and automatic welding operations and gives the advantages that the apparatus required for carrying out the method is small, inexpensive, robust, easy to set-up and align at the front face (torch side) of the weld, and requires no calibration Further, weldpool size and penetration can be calculated from measurement of the natural modes of weldpool vibration, so that welding engineers are able to specify their weld procedures in terms of weldpool size and penetration.

The ability to monitor weldpool size and penetration has important consequences on quality control and quality assurance in welding, since weldpool parameters can be measured on-line together with process parameters. This will improve weld quality and reduce the need for post weld inspection since a record of weldpool size along the length of the weld can be kept while welding.