METHOD FOR DETECTING FLAWS IN THE PROCESS FOR THE CONTINUOUS LASER WELDING OF METALLIC PORTIONS

• US 20140149075A1
• Filed: 08/29/2011
• Published: 05/29/2014
• Est. Priority Date: 08/31/2010
• Status: Active Grant

• Current Assignee: Consiglio Nazionale Delle Ricerche (Repubblica Italiana)
• Original Assignee: Consiglio Nazionale Delle Ricerche (Repubblica Italiana)
• Inventors: Toci, Guido, Pini, Roberto
• Granted Patent: US 9,505,086 B2
• Time in Patent Office: Days
• US Class Current: 702/182
• CPC Class Codes: B23K 26/032 using optical means
• CPC Class Codes: B23K 31/125 Weld quality monitoring

Abstract

In a method for detecting flaws in continuous laser welding processes, including a sensor measures plasma emission spectrum in a range of wavelengths between λ_max and λ_min in N channels, each measuring radiation strength emitted in the band centred at wavelength λ_i (1≦i≦N) and width Δλ_i. Deviation detection is based upon calculating intensity relationships of signals measured at specific wavelengths and in comparison with pre-determined threshold values. A calibration phase ensures a series of K weldings is performed, under optimized conditions, for each a number J of spectra O_J,K, is acquired, each one corresponding to a vector of N intensity values at wavelengths λ_i, 1≦i≦N, and a series of weldings K′ is then performed, (K′≠K), a single welding parameter A is altered by a known quantity, for each welding being acquired a sequence of J′ spectra A_J′,K′, (J′≠J), which are stored.

Claims

1. A method for detecting flaws in the process for the continuous laser welding of metallic portions, wherein a spectroscopic sensor (7) measures the plasma emission spectrum (2) in a range of wavelengths comprised between λ

_max and λ

_min in N channels so that each channel measures the intensity of the radiation emitted in the band centred at wavelength λ

_i (1≦

i≦

N) and with width Δ

λ

_i, the spectrum acquisition taking place at regular time intervals,

the detection of the deviation of the process physical parameters from optimum conditions being based upon the calculation, for each acquired spectrum, of intensity relationships of the signal measured at some specific wavelengths and in the comparison of such relationships with pre-determined threshold values,

said method comprising a calibration phase wherein;

a series of K weldings is performed, under optimized conditions with respect to the process parameters, for each one thereof a number J of spectra O_J,K, is acquired, each one corresponding to a vector of N intensity values at the wavelengths λ

_i, 1≦

i≦

N, which are stored in a processor;

a series of weldings K′

is performed, wherein K′

is not necessarily equal to K, wherein a single welding parameter A is altered by a known quantity, for each one of said weldings being acquired a sequence of J′

spectra A<sub>J′

,K′</sub>

, with J′

not necessarily equal to J, which are stored.

2. The method according to claim 1, wherein said parameter A is chosen from the group of parameters consisting of:

a laser emission power, with an alteration towards values higher or lower than an optimum value;

a piece translation speed, with an alteration towards values higher or lower than the optimum value;

a focus distance of the laser beam with respect to the piece surface with an alteration towards values higher or lower than the optimum value;

a separation or gap between two sheets to be welded; and

an assistance gas flow.

3. The method according to claim 1, wherein for each one of the spectra acquired respectively under optimized conditions O_J,K, and not optimized conditions A_J′

,K′

, the matrixes of ratios RO_J,K(q,m) and RA_J,K(q,m) are calculated according to the relation

RO_J,K(q,m)=O_J,K(q)/O_J,K(m) 

 

(2)

RA_J,K(q,m)=A_J,K(q)/A_J,K(m) 

 

(2a)

which correspond to a ratio of intensities at wavelengths λ

_q and λ

_m,

to detect the pairs of indexes q and m wherein intensity ratios of the spectra between the weldings under optimized conditions and the weldings under anomalous conditions for each pair of indexes mainly differentiate and the welding under anomalous conditions for each pair of indexes q, m, where 1<

q<

N and 1<

m<

N, takes place if for all spectra of the series O_J,K and A<sub>J′

,K′</sub>

the relation

RA<sub>J′

,K′</sub>

(q,m)>

RO_J,K(q,m) 

 

(3)

is verified

or if the relation

RA<sub>J′

,K′</sub>

(q,m)<

RO_J,K(q,m) 

 

(4)

is verified.

4. The method according to claim 3, wherein if the relation (3) or (4) is not verified for any pair of indexes, a weaker discrimination is detected by the following statistical criterion:

a search for pairs of indexes (q,m) which for all spectra and for all sequences J, K and J′

, K′

verify the relation

P<sub>100−

x</sub>(RO_J,K(q,m))≦

P_x(RA<sub>J′

,K′</sub>

(q,m)) 

 

(5)

or the relation

P<sub>100−

x</sub>(RA<sub>J′

,K′</sub>

(q,m))≦

P_x(RO_J,K(q,m)) 

 

(6)

wherein P_x is a percentile at a probability level x, P<sub>100−

x</sub> is a percentile at a probability level 100−

x;

for a given pair of indexes q, m the relation (5) being verified if a distribution of the values of RO_J,K(q,m) falls for a fraction 100−

x below a certain reference value, and a distribution of the values of RA<sub>J′

,K′</sub>

(q,m) falls for a fraction not higher than x below such reference value;

the relation (6) being verified if for a given pair of indexes q, m the distribution of the values of RA<sub>J′

,K′</sub>

(q,m) falls for a fraction 100−

x below a given reference value, and the distribution of the values of RO_J,K(q,m) falls for a fraction not higher than x below such reference value;

the value of x therefor the condition (5) or (6) being verified with the strict equality condition constituting a figure of merit for the pair of wavelengths λ

_q and λ

_m, as it designates a probability of discriminating between optimum or anomalous conditions by using the intensity ratio at the wavelengths λ

_q and λ

_m.

5. The method according to claim 4, wherein, for at least some of the anomalous conditions A for performing the welding:

a series of pair of wavelengths (λ

_q, λ

_m)_A,n verifying the condition (5) or the condition (6), with x smaller or equal to a given level x_min; and

a series of associated figures of merit (probability values) x(A,n) therefor the condition (5) or (6) is verified with the equality sign,

are detected,

for these pairs of wavelengths a series of higher threshold values S(A,n) is determined so that;

S(A,n)=P<sub>100−

x(A,n)</sub>(RO_J,K(q,m)_(A,n))=P_x(A,n)(RA<sub>J′

,K′</sub>

)q,m)_(A,n)) 

 

(7)

and a series of lower threshold values I(A,n) is determined so that;

I(A,n)=P<sub>100−

x(A,n)</sub>(RA<sub>J′

,K′</sub>

(q,m)_(A,n))=P_x(A,n)(RO_J,K(q,m)_(A,n)) 

 

(8)

in case for the pair of indexes (q,m)_(A,n) the conditions (3) or (4) are verified, thresholds S(A,n) and I(A,n) can be defined by using respectively the relations

S(A,n)=[min(RA<sub>J′

,K′</sub>

(q,m))+max(RO_J,K(q,m))]/2 

 

(9)

I(A,n)=[max(RA<sub>J′

,K′</sub>

(q,m))+min(RO_J,K(q,m))]/2 

 

(10)

being max and min the maximum and minimum values of respective series of ratios, in these two cases the value zero being assigned to the associated figure of merit (probability value) x(A,n).

6. The method according to claim 1, further comprising a diagnostic phase wherein the spectroscopic sensor acquires at regular time intervals the plasma emission spectrum during welding which is analysed to detect the anomaly A.

7. The method according to claim 6, wherein for the pairs of wavelengths (λ

_q, λ

_m)_A,n detected in the calibration phase, the corresponding ratios R(q,m)_n are calculated, each one of the so-calculated ratios is compared with the threshold I(A,n) or with the threshold S(A,n);

to verify condition

R(q,m)_n>

S(A,n) 

 

(11)

or the condition

R(q,m)_n<

I(A,n) 

 

(12)

taking place of the condition (11) or the condition (12) showing that the level of R(q,m)_n has entered the range of values corresponding to a condition of anomaly A.

8. The method according to claim 7, wherein the diagnostic phase is performed for at least one of the pairs of channels (q,m)_A,n, based upon how many pairs and which pairs (q,m)_A,n, (q′

,m′

)_A,n′

, (q″

,m″

)_A,n″

verify a condition for passing the threshold (11) or (12) being possible deciding if an anomaly has really occurred, which type among the experimented cases in the learning phase, and in case using this piece of information for corrective actions.

9. The method according to claim 8, wherein the diagnostic phase is repeated for each anomaly proved in a preceding calibration phase.

10. The method according to claim 1, wherein the plasma emission spectrum is acquired by a sensor configured to measure simultaneously an instantaneous or integrated intensity over a time range Δ

t, of an optical radiation in N spectral bands in a region 200-2000 nm, centred at different wavelengths λ

_i, (i=1 . . . N) with bandwidths Δ

λ

_i.

11. The method according to claim 10, wherein said sensor is constituted by a diffraction grating type spectrometer with array detector by a dispersive prism type spectrometer with array detector;

by a series of band-pass filters centred at different wavelengths with a radiation power measurer associated to each filter;

by a sensor constituted by an array detector of CCD or CMOS type in front thereof a band pass filter is placed with transmission spectrum depending from the position;

by a Fourier transform type spectrometer with beam division.

12. The method according to claim 1, wherein the spectrum to be analysed coming from the sensor is previously subjected to pre-processing procedures to calculate again or define again the number of bands, to calculate again or define again the width of the single bands, to modify the signal level in each band, to improve the signal-noise ratio, by at least one of the following:

subtraction of a background spectrum;

transformation of the signal intensity level of each band by means of a defined mathematical relation;

interpolation at different wavelengths;

re-sampling at different wavelengths;

average or sum between channels with contiguous wavelengths;

convolution with a spectral filter function;

average, sum or convolution with a time filter function of two or more spectra acquired at different periods of time.