Method of controlling the quality of industrial processes and system therefor

Abstract

A method for controlling the quality of industrial processes, of the type comprising the steps of: making available one or more relating to the industrial process acquiring one or more real signals, indicative of the quality of said industrial process, comparing said one or more reference signal to said one or more real signals to identify defects in said industrial process. According to the invention the method further comprises the operations of: obtaining a transformed signal from said reference signal; obtaining a transformed signal from said real signal; calculating energies of said transformed signals, respectively reference and real, said comparing operation comprising: comparing said energies of said transformed signals respectively reference and real to extract corresponding time frequency distributions for selected frequency values; calculating energies of said time frequency distributions; comparing the energies of said time frequency distributions with threshold values to identify energy values associated to defects.

Description

The present invention relates to methods for controlling the quality of an industrial process, comprising the steps of:

making available one or more reference signals relating to the industrial process

acquiring one or more real signals indicating the quality of said industrial process,

comparing said one or more reference signals to said one or more real signals to identify defects of said industrial process

Monitoring defects in industrial processes is assuming a growing economic importance due to its impact in the analysis of the quality of industrial products. The ability to obtain an assessment of the quality of the industrial process in line and automatically has many advantages, both in economic terms and in terms of process velocity. Therefore, the desirable characteristics of the system are:

on line and real time processing;

ability to recognise the main production defects with accuracy.

Currently, the problem of recognising the quality of an industrial process, and thus of identifying any defects, takes place through an off-line inspection by experts, or with automatic methods which, through sensors, identify only some of the aforementioned defects, in a manner that is not satisfactory and that is also sensitive to the different settings of the machine.

Methods and systems for controlling the quality of industrial processes are known, for instance applied to the on-line monitoring of the laser welding process, in particular in the case of metal sheet welding. The controlling system is able to assess the presence of porosities in the welded area or, in the case of butt-weeded thin metal sheets, the presence of defects due to the superposition or to the disjunction of the metal sheets.

Said used systems base quality control on a comparison between the signals obtained during the process and one or more predetermined reference signals, indicative of a high quality weld. Said reference signals, usually in a variable number between two and ten, are predetermined starting from multiple samples of high quality welds. Obviously, this way of proceeding implies the presence of an experienced operator able to certify the quality of the weld at the moment of the creation of the reference signals, entails time wastage and at times also material wastage (which is used to obtain the samples needed to obtain the reference signals). In some cases, reference signals indicating a defective weld are also arranged, and this entails additional problems and difficulties.

The European patent application EP-A-1275464 in the name of the present Applicant teaches to divide into blocks the signal acquired by means of a photodiode which collects the radiation emitted by a weld point, calculating the mean of the signal in each sampled block and taking in account the blocks whose value is lower than or equal to the offset of the photodiode to be indicative of the presence of a defect. Said method eliminates the need for the reference, but it allows for a very approximate detection of defects.

The object of the present invention is to overcome all the aforesaid drawbacks.

In view of achieving said object, the invention relates to a method for controlling the quality of industrial processes having the characteristics set out at the beginning and further characterised by the fact that it further comprises the operations of:

obtaining a transformed signal from said reference signal;

obtaining a transformed signal from said real signal;

calculating energies of said transformed signals, respectively reference and real signal;

said comparison operation comprising:

comparing said energies of said transformed signals, respectively reference and real, to each other to extract corresponding time frequency distributions for selected frequency values;

calculating energies of said time frequency distributions;

comparing the energies of said time frequency distributions with threshold values to identify energy values associated to defects.

In the preferred embodiment, said steps of obtaining a transformed signal from said reference signal and of obtaining a transformed signal from said real signal comprise a filtering operation by the application of a DWT (Discrete Wavelet Transform), whilst said operation of comparing said energies of said transformed signals, respectively reference and real, to obtain corresponding time frequency distributions comprises operating a calculation of the conjugate of the Fourier transform of the envelope of the real signal and of the envelope of the normalised signal, obtaining conjugate transformed signals, respectively real and reference, and comparing the energies of the reference signal and of the real signal, extracting the frequency values for which the energy of the real signal is greater than the reference signal.

Naturally, the invention also relates to the system for controlling the quality of industrial which implements the method described above, as well as the corresponding computer product directly loadable into the memory of a digital computer such as a processor and comprising software code portions to perform the method according to the invention when the product is executed on a computer.

Additional characteristics and advantages of the present invention shall become readily apparent from the description that follows with reference to the accompanying drawings, provided purely by way of non limiting example, in which:

FIG. 1 is a block diagram showing a system that implements the method according to the invention;

FIG. 2 shows a detail of the system of FIG. 1;

FIGS. 3, 4 and 5 are flowcharts representing operations of the method according to the invention;

FIG. 6 is a diagram of quantities computed by the method according to the invention.

The method according to the invention shall now be exemplified with reference to a laser welding method. Said laser welding method, however, constitutes only a non limiting example of industrial process to which the method for controlling the quality of industrial processes according to the invention can be applied.

With reference to FIG. 1, the number 10 designates as a whole a system for controlling the quality of a laser welding process. The example refers to the case of two metal plates 2, 3 which are welded by means of a laser beam. The number 4 designates the focusing head as a whole, including a lens 5 whereat arrives the laser beam originated by a laser generator (not shown) and reflected by a semi-reflecting mirror 6, after the passage through a lens L. The radiation E emitted by the weld area passes through the reflecting mirror 6 and is sensed by a sensor 7 constituted by a photodiode able to sent its output signal to an electronic control and processing unit 8 associated to a personal computer 9.

In an actual embodiment, the semi-reflecting mirror 6 used is a ZnSe mirror, with a diameter of 2 inches, thickness 5 mm. The sensor 7 is a photodiode with spectral response between 190 and 1100 nm, an active area of 1.1×1.1 mm and a quartz mirror.

FIG. 2 shows in greater detail the control and processing electronic unit 8 associated to the personal computer 9. Said processing unit 8 comprises an antialiasing filter 11 which operates on the signal sent by the sensor 7, hence an acquisition board 12 is provided, equipped with an analog-digital converter, which samples the filtered signal and converts it into digital form. Said acquisition board 12 is preferably directly associated to the personal computer 9.

Also in the case of an actual embodiment, the acquisition card 12 is a PC card NI 6110E data acquisition card, with maximum acquisition frequency of 5 Ms/sec.

The antialiasing filter 11 filters the signal by means of a low pass filter (e.g. a Butterworth IIR filter).

In the personal computer 9 according to the invention is implemented a method for controlling quality, based on a comparison between a real signal x_real acquired by means of the photodiode 7 and a reference signal x_ref, representing a defective weld, stored in said personal computer 9.

The reference signal, designated as x_ref(t) is acquired at an acquisition frequency f_s, and hence, according to Nyquist's theorem, has associated a frequency band of the signal with value f_s/2, whilst the number of samples acquired for the reference signal x_ref(t) is N.

FIG. 3 shows a flow chart which represents the operations conducted on the reference signal x_ref(t).

In a first step 100 is executed a filtering operation of the reference signal x_ref(t) by the application of a DWT (Discrete Wavelet Transform). At the output of the step 100, therefore, one obtains a signal x_ref—DWT having N/2 samples in the band 0:f_s/4.

Subsequently, in a step 101 to the x_ref—DWT signal is applied a Hilbert transform operation, obtaining a complex analytical signal x_ref—HIL, having N/2 samples and with null negative frequencies.

To said analytical signal x_ref—HIL is applied, in a step 102, a normalisation operation, which outputs a normalised signal x_ref—norm.

On said normalised signal x_ref—norm, in a step 103, an operation of calculating an envelope of the normalised signal, designated as x_{ref—inv—norm}, is performed, whilst in a step 104 to said envelope of the normalised signal x_{ref—inv—norm} is applied a Fourier transform operation (FFT), obtaining a transformed envelope X_{ref—inv—norm}.

Lastly, in a step 105, an operation of calculating the energy of the reference signal, designated E_ref, is conducted, applying the following relationship: ∫|x_{ref—inv—norm}(t) |²dt=∫|X_{ref—inv—norm}(f) |² df   (1)

In regard to the real signal x_real (t), it is also acquired at an acquisition frequency f_s, and hence, according to Nyquist's theorem, has associated a frequency band of the signal with value f_s/2, whilst the number of samples acquired for the real signal x_real (t) is N.

FIG. 4 shows a f low chart which represents the operations conducted on the real signal x_real(t).

In particular, FIG. 4 shows a first step 200 in which a filtering operation of the real signal x_real(t) is executed by the application of a DWT transform. At the output of the step 200, therefore, one obtains a signal x_real—DWT having N/2 samples in the band 0:f_s/4.

On said signal x_real—DWT, in a step 211, is performed a Fourier transform operation, obtaining a transformed signal FFT_—real, which, subsequently, in a step 212, is normalised, obtaining a transformed normalised signal FFT_{—real—norm}.

In a step 250, on the transformed normalised signal FFT_{—real—norm} an operation of calculating a mean frequency f₀ is conducted, according to the relationship: f₀=∫f*FFT_{—real—norm}(f)*FFT_{—real—norm} (f) df   (2)

In a step 251, an operation of calculating a standard deviation B is conducted, according to the relationship: B=(∫f ²* FFT_{—real—norm}*FFT_{—real—norm} df−f₀²)^1/2   (3)

In a step 252 are then calculated a lower band F_Sn=(f₀−B/2) and an upper band F_Dx=(f₀+B/2).

In parallel, in a step 201, to the x_real—DWT signal is applied a Hilbert transform operation, obtaining a complex analytical signal x_real—HIL, having N/2 samples and with null negative frequencies.

To said analytical signal x_real—HIL is applied, in a step 202, a normalisation operation, which outputs a normalised signal x_real—norm.

On said normalised signal x_real—norm, in a step 203, an operation of calculating the envelope, designated as x_{real—inv—norm}, is conducted, whilst in a step 204 to said envelope of the normalised signal x_{real—inv—norm} is applied a Fourier transform operation (FFT), obtaining a transformed envelope X_{real—inv—norm}.

Lastly, in a step 205, an operation of calculating the energy of the real signal E_real is performed, applying the following relationship: ∫|x_{real—inv—norm}(t) |² dt=∫|X_{real—inv—norm}(f) |² df   (4)

The operations of calculating the energies E_real and E_ref are conducted in a band delimited between the lower band F_Sn and the upper band F_Dx calculated at the step 252. More in detail, the calculation is performed on the band so delimited, considering frequency steps, for example of one Hertz, i.e.:

In this way the operation of calculating the energies E_ref and E_real produces two respective vectors, respectively a vector of energies of the reference ∫_F—Sn^step|X_{real—inv—norm}(f) |² df ∫_step^FDX |X_{real—inv—norm}(f) |² df ∫_F—Sn^step|X_{ref—inv—norm}(f) |² df ∫_step^FDX |X_{ref—inv—norm}(f) |² df

Energy_Ref_step (1, . . . k) and a vector of energies of the real signal Energy_Real_step (1, . . . k), both comprising k values in frequency.

Subsequently, a procedure of calculating the time-frequency quadratic distributions is performed, shown in the flowchart of FIG. 5, and comprising the following operations:

in a step designated as 300, calculating the conjugate of the Fourier transform (FFT) of the envelope of the real signal X_{real—inv—norm}(f) and of the envelope of the reference signal X_{ref—inv—norm}(f), obtaining conjugate transformed signals, respectively real X*_{real—inv—norm}(f) and reference X*_{ref—inv—norm}(f);

in a step 301, taking in account the energies of the reference signal E_ref and of the real signal E_real, represented by the respective energy vector of the reference Energy_Ref_step (1, . . . k) and the energy vector of the real signal Energy_Real_step (1, . . . k), and for each element k of said two vectors, evaluating whether the following criterion is met: Energy_Real_step (1, . . . k)>Energy_Ref_step (1 . . . k)   (5)

This operation can also be appreciated with reference to the chart of FIG. 6, which shows the amplitudes of the energies of the reference signal E_ref and of the real signal E_real (shown with thicker lines) as a function of frequency.

if the criterion (5) is met, then in a step 302 an operation of extracting the frequency value for which said criterion (5) is met is performed, said value being indicated as f_e. Depending on the number of times the condition is met, up to k values of frequency f_e are obtained. FIG. 6 shows the regions corresponding to the values of frequency f_e for which the criterion (5) is met;

in a step 303 a matrix M is constructed whose rows are constituted by extracted frequency values f_e, whilst the columns are constituted by N/2 time values t₁ . . . t_N/2 of the output signal from the DWT transform operation 200;

in a step 304, for each row of the matrix M is calculated a time-frequency quadratic distribution both for the reference signal, designated as Tfd_ref, and for the real signal, designated Tfd_real, using the Margenau_Hill relationship, i.e.: Tfd_real=Real (x_real—DWT(t)•X_{real—inv—norm}*(f)•e^−j2πf)   (6) Tfd_ref=Real (x_ref—DWT(t)•X_{ref—inv—norm}* (f)•e^−j2πf)   (7)

in a step 305 for both reference and real signals are then calculated energies associated to the distributions for each time instant, respectively designated Et_ref and Et_real;

in a step 306 is then calculated a maximum value of energy max_Tfd_ref for the time frequency distribution of the reference Tfd_ref.

To obtain an estimate of the defects S, lastly in a step 307 each time value of the energy Et_real of the time-frequency quadratic distribution of the real signal Tfd_real is compared with the maximum value of energy max_Tfd_ref. If said value of the time-frequency quadratic distribution of the real signal Tfd_real exceeds the maximum value of the energy max_Tfd_ref then a defect is present at that time coordinate.

It is thereby possible to temporally locate defects.

To evaluate the defects, with reference to FIG. 7, the quantities taken into consideration are the energy of the real signal E_real, originated at the step 205 of FIG. 4, as well as the lower band F_Sn=(f₀−B/2) and the upper band F_Dx=(f₀+B/2) of the defect calculated at the step 252. Lastly, the extension and location of the defect in the frequency band is considered, as evaluated at the step 307 of FIG. 5.

Said parameters, i.e. the energy of the real signal E_real, the lower band F_Sn and the upper band F_Dx, the extension and location of the defect, according to an aspect of the invention, are sent to a defect classifier 400 which, receiving at its input the identified characteristics (or a subset thereof) evaluates the quality of the weld as: “correct”/“not-correct”/“insufficient-penetration”/“discontinuous-laser-power”/“incorrect-mounting”/“porosity”.

In this way advantageously, the outputs of the steps 205, 252 and 307, relating to the time/frequency analysis of defects are used to instruct the defect classifier 400 automatically, thereby avoiding steps of instructing the classifier 400 by an operator. Lastly, in a block 401 it is possible to cross check the results of the outputs of the steps 205, 252 and 307 and of the block 401 for a final evaluation of the defect.

Naturally, without altering the principle of the invention, the construction details and the embodiments may vary widely from what is described and illustrated purely by way of example herein, without thereby departing from the scope of the present invention.

Claims

1. A method for controlling the quality of an industrial processes, of the type comprising the steps of: making available one or more reference signals relating to the industrial process acquiring one or more real signals, indicative of the quality of said industrial process, comparing said one or more reference signal to said one or more real signals to identified defects in said industrial process, wherein it further comprises the operations of: obtaining a transformed signal from said reference signal; obtaining a transformed signal from said real signal; calculating energies of said transformed signals, respectively reference and real signals: said comparison operation comprising: comparing said energies of said transformed signals, respectively reference and real signals to each other to extract corresponding time frequency distributions for selected frequency values; calculating energies of said time frequency distributions; comparing the energies of said time frequency distributions with threshold values to identify energy values associated to defects; providing said energy values associated to the defects to a classifier.

2. The method of claim 1, wherein it comprises providing said classifier also with the energy of said real transformed signal.

3. The method of claim 1, wherein said steps of obtaining a transformed signal from said reference signal and of obtaining a transformed signal from said real signal comprise a filtering operation by the application of a DWT (Discrete Wavelet Transform).

4. The method of claim 3, wherein said steps of obtaining a transformed signal from said reference signal and of obtaining a transformed signal from said real signal further comprise the operations, applied both to the reference signal and to the real signal, of: applying a Hilbert transform to the signal obtained from the filtering operation; normalizing the signal obtained from the Hilbert transform operation; calculating an envelope of the normalized signal; applying an FFT transform to said envelope of the normalizing signal to obtain said transformed signals, respectively reference and real.

5. The method of claim 3, wherein it further comprises executing a Fourier transform operation on the real signal obtained from the filtering operation by applying a DWT transform obtaining a second transformed signal and normalizing said second transformed signal, obtaining a second transformed normalizing signal.

6. The method of claim 5, wherein it comprises processing said second transformed normalizing signal to obtain a set of values representing the spectrum of the real signal and provide values selected in said set of values to said classifier.

7. The method of claim 6, wherein it comprises using at least part, of said set of values representing the spectrum of the real signal to calculate said energies of said transformed signals, respectively reference and real.

8. The method of claims 1, wherein said operation of comparing said energies of said transformed signals, respectively reference and real to obtain corresponding time frequency distributions comprises the operations of: calculating a conjugate of the Fourier transform of said transformed signals, respectively reference and real obtaining conjugate transformed signals, respectively real and reference; comparing the energies of the reference signal and of the real signal extracting the frequency values for which the energy of the real signal is greater than the energy of the reference signal; building a matrix whose rows are constituted by said extracted frequency values and whose columns are time values of the signal obtained from the filtering operation by means of DWT transform; calculating for each row of said matrix a time-frequency quadratic distribution for the reference signal and for the real signal.

9. The method of any of claim 3, wherein said operation of calculating for each row of said matrix a time-frequency quadratic distribution for the reference signal and for the real signal is conducted applying the Margenau-Hill relationship.

10. The method of claim 3, wherein said operation of calculating energies of said time frequency distributions comprises the operations of: calculating said energies for each time instant and also calculating a maximum value of energy; using said maximum value, of energy as a threshold value; and comparing said maximum value of energy with each time value of energy of the time-frequency quadratic distribution of the real signal to identify energy values associated with defects.

11. Method as claimed in claim 1, wherein it comprises a step of crossing the results of said classifier with one or more values selected among said energy values associated to the defects, said energy of said real transformed signal and said set of values representative of the spectrum of the real signal.

12. A system for controlling the quality of an industrial process, comprising: sensor means for measuring one or more process parameters, and an electronic control and processing unit for processing the signals emitted by said sensor means, wherein: said electronic control and processing unit to process the signals emitted by said sensor means implements the method for controlling the quality of an industrial process as claimed in claims 1 through 11.

13. System as claimed in claim 12, wherein said industrial process is a laser weld process.

14. A computer product loadable into the memory of an electronic computer and comprising software code portions to perform the method as claimed in claim 1 when the product is executed on a computer.