CN101201339A - Apparatus and method for monitoring resistance spot welding quality - Google Patents

Abstract

The invention belongs to a resistance spot welding quality monitoring device and a monitoring method in the field of metallic material resistance spot welding. The invention is characterized in that an independent displacement transducer, an independent electric pressure transducer, an independent current sensing device, an independent pressure transducer are arranged at different positions of a spot-welding equipment; the sensors are connected with a data acquisition analytic device. As the device adopts the method of diagnostic information extraction and the information integration during the course of spot welding, the quality of resistance spot welding can be monitored and determined on line; and the diagnostic information during the course of spot welding can be analyzed effectively. In addition, the diagnostic information abstracted has high signal to noise ratio, and distinct characteristics; and is stable. The resistance spot welding quality monitoring specimens abstracted have the advantages of expandability and so on, and can be widely used. The device is properly designed, and can be installed and maintained conveniently. Accordingly, the invention is highly practical; the implementation cost is low; and the invention can be promoted and applied easily.

Description

Resistance spot welding quality monitoring device and monitoring method

Technical field

The invention belongs to metal material resistance spot welding field, especially a kind of resistance spot welding quality monitoring device and monitoring method that constitutes by multisensor and data collector.

Background technology

Resistance spot welding process is an interactional process of highly non-linear and a large amount of uncertain factors.Because the resistance spot welding forming core is in closed state and can't directly observes, and compole is short during forming core, therefore, the fluctuation in short-term of welding condition will cause defectives such as more serious splash, incomplete penetration or incomplete fusion, and is all the more so to the resistance spot welding of aluminium and aluminium alloy.In actual production, owing to be subjected to all multifactor influences such as disturbance of variation, wear to electrodes, fit-up gap and the physical construction of workpieces surface condition, even produce according to same weld technology tissue, still quality fluctuation might appear, in case quality problems occur, often adopt a large amount of anatomy experiments or increase the solder joint number to guarantee architecture quality, will certainly increase production cost like this.At present, the quality to resistance spot welding still lacks reliable, practical nondestructive test technology and method.

Resistance spot welding process relates to multi-field knowledge such as mechanics, calorifics, electricity, metallurgy, its resistance spot welding process relates to the variation of the physical-chemical parameters such as mechanics, calorifics, electricity, metallurgy, the variation of these parameters can adopt advanced sensing technology, computer data to handle and control technology realizes monitoring and controlling resistance pinpoint welding procedure.Many researchers have been noted that the resistance spot welding process more complicated, and the information that single-sensor obtains is limited, need to adopt multisensor to obtain more information.Reject redundancy, complement each other, the core technology of monitored resistance spot welding is that final the extraction has obvious, stable key feature information.

The information that resistance spot welding process can be monitored is a lot, as: electrode current, electrode voltage, spot welding energy, dynamic resistance, welding foroe, electrode displacement, acoustic emission, ultrasound wave, infrared radiation etc., it is the key that realizes on-line monitoring and control that searching and resistance spot welding quality have the characteristic information of obvious corresponding relation, mainly contains following certain methods:

(1) 1996 year, Hao M adopts multisensor to study the aluminium resistance spot welding process, has extracted 274 characteristic quantities, and this method is by regretional analysis, seek the relation between electrical quantity and the spot size, find the splash phenomenon that electrode voltage and electrode displacement can the reflected resistance pinpoint welding procedures.But the document do not obtain can the reflected resistance point quality the universality characteristic information.

(2) permanent heat or permanent energy control method, this method is measuring resistance spot-wedling electrode voltage and electrode current simultaneously, because the big electric current of resistance spot welding causes electromagnetic noise sensor very big, realize noise-reduction method though can dwindle the area that twisted-pair feeder covers by employing, but, because the restriction of the on-the-spot service condition of commercial Application, it is very little that the area of twisted-pair feeder envelope can't reduce to, thereby cause the serious distortion of signal.

(3) dynamic resistance monitoring method is a kind of monitoring method of comparative maturity.For example: Broomhead claims as far back as nineteen ninety, adopt the dynamic resistance feedback to realize the closed-loop control of truck parts or battery electrode resistance spot welding process, regrettably, the Changing Pattern of aluminium and aluminium alloy resistance spot welding process dynamic resistance is different with mild carbon steel, inapplicable aluminium of this method and aluminium alloy.

In sum, the domestic and international research majority also rests on laboratory stage, just can monitor out the variation of some parameter of resistance spot welding or make some judgement by means of machine with artificial the combination.But, The more the better at the sensor that experimental stage is used, verify mutually by many information, replenish mutually, reject redundancy, finally extract evident characteristic, the wide key feature information of the scope of application, but the sensor that uses in the industry practical stage is few more good more, this just requires, and relation between characteristic information and the quality is simple clear and definite, signal to noise ratio (S/N ratio) is high, algorithm is succinct, processing is quick, finally, utilizes tangible key feature to be convenient to realize on-line monitoring and grade estimation.To how using less sensor obtains obvious corresponding relation to point quality the reaching of characteristic information to quality monitoring of resistance spot welding, prior art does not also have practical way preferably, and stability and the accommodation for the resistance spot welding quality monitoring also can't satisfy practical requirement simultaneously.

Summary of the invention

The objective of the invention is to overcome the deficiencies in the prior art, provide a kind of multisensor and data collection and analysis treating apparatus can carry out the resistance spot welding quality monitoring device and the monitoring method of on-line monitoring and prediction point quality.

The present invention solves its technical matters and takes following technical scheme to realize:

This resistance spot welding quality monitoring device, constitute by resistance spot welding equipment body, last horn, following horn, top electrode, bottom electrode, it is characterized in that: a displacement transducer is installed in the rigid support top that top electrode installs, the lower surface of this displacement transducer is parallel with the metal plate that last horn outside level is stretched out, a pressure transducer is installed in bottom at following horn, and displacement transducer and pressure transducer all are connected with equipment data acquisition analyzing by shielded cable.

And, a voltage sensor also is installed on top electrode and bottom electrode, in the bottom electrode outside one current sensor is installed also, the two all is connected with equipment data acquisition analyzing by shielded cable.

And described displacement transducer is an eddy displacement sensor; Described pressure transducer is highly sensitive quartzy electrostrictive strain extreme pressure force transducer, and the two all is connected with equipment data acquisition analyzing by shielded cable.

And described voltage sensor is a LEM Hall effect electrode voltage sensor; Described current sensor is that a skeleton adopts flexible material, the air core coil of the number of turn between 500 to 50000 circles.

And, described equipment data acquisition analyzing comprises that data acquisition process unit, Data Transmission Controlling unit and data analysis processing unit constitute, wherein, the data acquisition process unit is made of displacement transducer unit, voltage sensor unit, current sensor unit, pressure sensor unit, these four sensor units are connected with the Data Transmission Controlling unit by serial line interface, and the Data Transmission Controlling unit is connected with data processing centre (DPC) by parallel interface.

And described serial line interface can be the combination of RS232 interface, RS485 interface or RS232 and RS485 interface.

The monitoring method of this resistance spot welding quality, this method realizes by following step:

(1) the data acquisition process unit gather in real time displacement, voltage, electric current, four kinds of data of pressure or wherein the starting point of the pinpoint welding procedure that arrives of certain several data or wherein a kind of data based sensor monitors and terminating point as the reference point of data cutout, intercepting and normalization sensing data;

(2) synchro control of sensor data acquisition is realized data acquisition process unit is carried out in the Data Transmission Controlling unit, and normalized sensing data is transferred to data processing centre (DPC);

(3) data processing centre (DPC) extracts the point quality characteristic information according to the position and the scope of every kind of sensor characteristics information time of origin of sensing data analysis and frequency;

(4) data processing centre (DPC) carries out statistical study to every kind of sensor special time period and frequency band signal, further improves the signal to noise ratio (S/N ratio) of characteristic information;

(5) data processing centre (DPC) uses statistical analysis technique or through differentiating every kind of sensing data of analytical, obtains the statistical information of statistics extreme difference, average, variance, standard deviation, mean absolute deviation;

(6) data processing centre (DPC) realizes the online quality judging of direct labor or judges that by instrument each critical point is implemented in line mass and judges according to the characteristic information and the statistical information that obtain.

(7) the data analysis processing unit adopts principal component analysis (PCA), neural network method or fuzzy mathematics to realize the fusion of characteristic information in conjunction with neural net method, estimate the internal relation between each information characteristics of resistance spot welding, carry out the on-line prediction and the classification of resistance spot welding quality.

And the mode of described collection voltage also can directly be got voltage between spot welding top electrode and the bottom electrode as voltage signal, and secondary voltage, can be connected by direct and follow-up modulate circuit behind electric resistance partial pressure less than 10V usually.

And the method for described extraction point quality characteristic information can use time-frequency domain analytical approach based on small echo, based on hardware or software or the hard soft method that combines with software, based on LABVIEW multisensor resistance spot welding feature extracting method.

Description of drawings

Fig. 1 is a resistance spot welding quality monitoring device synoptic diagram;

Fig. 2 is the equipment data acquisition analyzing theory diagram;

Fig. 3 is each sensor signal pre-service, standardization and data compression exemplary plot;

Fig. 4 is a spot welding displacement signal wavelet analysis result schematic diagram;

Fig. 5 is the graph of a relation of small echo details component extreme difference and spot size;

Fig. 6 is the graph of a relation of displacement signal standard deviation and spot size;

Fig. 7 is spot welding displacement signal intermediate frequency filtering figure as a result;

Fig. 8 is based on LABVIEW multisensor resistance spot welding feature extracting method synoptic diagram;

Fig. 9 judges and the classification results synoptic diagram for the principal component analysis (PCA) quality of welding spot;

Figure 10 is an oscillograph monitoring result synoptic diagram.

Embodiment

Below in conjunction with accompanying drawing the embodiment of the invention is further described:

Be the example explanation with DJ-1000 dc point welding machine as the main body of resistance electric welding equipment below, this resistance spot welding quality monitoring device, constitute by resistance spot welding equipment body 7, last horn 3, following horn 11, top electrode 4, bottom electrode 9, installation position displacement sensor 1 above the rigid support 5 that top electrode installs, the lower surface of this displacement transducer is parallel with the metal plate 2 that last horn outside level is stretched out, this displacement transducer realization is synchronized with the movement with top electrode, and generation reflects the variation of distance between the two with the relative position variation of metal plate; Two electrodes of voltage sensor 6 are installed in respectively on top electrode and the bottom electrode by screw, current sensor 10 is installed in the bottom electrode outside, pressure transducer 8 is installed in down the bottom of horn by screw, and the sensor is connected to equipment data acquisition analyzing by the private mask cable link.This equipment data acquisition analyzing comprises that data acquisition process unit, Data Transmission Controlling unit and data analysis processing unit constitute.The data acquisition control unit is connected with the Data Transmission Controlling unit by serial line interface, this serial line interface can be the combination of RS232 interface, RS485 interface or RS232 and RS485 interface, in the present embodiment, the Data Transmission Controlling unit is connected with data processing centre (DPC) by parallel interface.Above-mentioned data acquisition control unit 12 is made of four displacement transducer unit, voltage sensor unit, current sensor unit, pressure sensor unit of independently being made up of little processing controls chip and peripheral circuit thereof.The above-mentioned four kinds of separate sensor units of Data Transmission Controlling unit controls, the assurance sensor data acquisition is synchronous, and control data is according to certain agreement self-checking, accurately upload the data to the data analysis processing unit, and the data analysis processing unit is the computer installation of spot welding quality analysis software of packing into.

In the resistance spot welding quality monitoring device, information sensing is the basis of resistance spot welding quality monitoring system, and purpose is the resistance spot welding quality information that picked up signal is true and reliable, feature is obvious, signal to noise ratio (S/N ratio) is high, thereby provides foundation for quality judging.The resistance spot welding process electrode current is big, and (2000A～10000A), weld interval are short, and (40ms～500ms), the non-sinusoidal wave characteristics that frequency is low belong to moment, nonlinear complicated dynamic process.The quality of sensor performance is the key problem in technology and the basis of this research.In the selection of sensor and customization, to consider the commercial Application environment, promptly dustproof, moistureproof, grease proofing, insulate isolation, Chinese People's Anti-Japanese Military and Political College's electrode current interference of electromagnetic field etc.According to mentioned above principle, the selected various sensors of the present invention are:

Displacement transducer is an eddy displacement sensor, as with welding resistance spot welding defective closely-related electrode displacement information, it is the comprehensive embodiment of each parameter influence of resistance spot welding, at present, the electrode displacement sensor that can select has laser electrode displacement sensor, LVDT (Linear Variable Differential Trans-former) electrode displacement sensor, grating electrode displacement transducer, eddy current electrode displacement sensor etc.Wherein, grating electrode displacement transducer and laser electrode displacement sensor belong to photoelectric sensor, and service condition is higher relatively, and monitoring production scene equipment is not first-selected; The LVDT electrode displacement sensor belongs to feeler, is subject to pressure process and impacts, and the automatic reaction frequency is lower, so the non-contact turbulent flow electrode displacement sensor is selected in this recommendation.

Voltage sensor is a LEM Hall effect electrode voltage sensor, realizes that the electricity of monitor signal and resistance spot welding secondary circuit is isolated.In addition, also can directly get voltage between spot welding top electrode (4) and the bottom electrode (9) as voltage signal, secondary voltage, can directly be connected with data transfer control circuit behind electric resistance partial pressure less than 10V usually, then, and by photoelectricity realization isolation.According to the Faraday law of electromagnetic induction, adopting twisted-pair feeder is to reduce induction noise the best way.

Current sensor is a great current sensor, is a kind of air core coil of special construction, and the rate of change of electromotive force that is produced in the cell winding or corresponding induced voltage signal and resistance spot welding electric current is linear.The size of this signal is relevant with the pickup wire coil structures, and the coil mean radius is more little, and cross-sectional area is big more, and the number of turn is many more, and the signal that is obtained is strong more.Sensor to the kind of electrode current, size, frequency etc. without limits.By integration, there are linear relationship in this induced voltage signal and electrode current to be measured.For improve the current sensor signal signal to noise ratio (S/N ratio), reducing cost can be according to electrode diameter specialized designs great current sensor customized.Skeleton adopts flexible material, and the number of turn is between 500～50000 circles, with 1000～5000 circle the bests.

Pressure transducer is highly sensitive quartzy electrostrictive strain extreme pressure force transducer, its power supply 18～30VDC, no-load current＜20mA; Output voltage ± 10V, output current＜1mA, resistance 10 Ω, noise＜± 25mV, frequency response 10kHz; In 5 ℃～50 ℃ temperature ranges, then temperature drift be no more than 0.07%/℃.Measuring range adjustable, easy for installation.

In the present embodiment, the resistance spot welding welding method of resistance spot welding quality monitoring comprises shock wave spot welding or impulsed spot welding, and the welding material that is fit to is aluminium and aluminium alloy.In the present embodiment, the aluminium alloy of welding material is LF6 and LD10CS, and wherein LF6 is 1mm～2.5mm, and LD10CS is 4mm～5.5mm.Two kinds of aluminum alloy plate materials can be combined into different lap joints, and this sheet material is strict chemical corrosion of process and mechanical grinding before welding.The spot welding great current sensor adopts the TOROID coil, designs and produces voluntarily, and its internal diameter 110mm, the number of turn 1400～5000, the transversal section is the bandlet of rectangle.

Eddy displacement sensor probe coil temperature influence, in-20 ℃～120 ℃ temperature ranges, temperature drift is no more than 0.025%/℃; Frequency response 0～10KHz.Power supply-24VDC, under the load 10K Ω condition, the range of linearity 3～13mm, voltage signal output 0～10V.Maximum drive signal cable length 300m.This sensor is fit to industrial environment and uses.

With the method for said apparatus explanation resistance spot welding quality monitoring, the method performing step of its quality monitoring is as follows below:

The 1st step: in resistance spot welding operation, be installed in the variation that displacement transducer, voltage sensor, current sensor and pressure transducer on the resistance spot welding equipment can monitor correlation parameter.Four data acquisition process unit are gathered displacement, voltage, electric current, four kinds of data of pressure in real time, the starting point of the resistance spot welding process that arrives according to each sensor monitors and terminating point are as the reference point of data cutout, the intercepting and the various sensing datas of standardizing, it is corresponding mutually in time to reach each sensor signal, reduces the purpose of data total amount simultaneously.

The 2nd step: the synchro control of sensor data acquisition is realized four data acquisition process unit are carried out in the Data Transmission Controlling unit, receives the standardizing number certificate of data acquisition process unit and is transferred to data processing centre (DPC); Adopt the serial communication mode that combines based on RS232 and RS485 to realize the long distance of digital signal between Data Transmission Controlling unit and the data acquisition process unit, transmission data acquiring frequency 0.1kHz～100MHz, its optimum range 2kHz～15kHz, data rate 1～5s adopts the parallel communications mode to transmit data between Data Transmission Controlling unit and the data processing centre (DPC).Electric current, voltage, electrode displacement and four each synchronous acquisition 5s of sensor of welding foroe, sample frequency 10kHz, 12 precision, full scale (amplitude) 4096.About 2 seconds, will be equivalent to 400K 8 bit data and be transferred to data processing centre (DPC), meet on-the-spot actual needs.

The 3rd step: data processing centre (DPC) extracts the point quality characteristic information according to the position and the scope of every kind of sensor characteristics information time of origin of sensing data analysis and frequency, and the method for its extraction has following three kinds:

The 1st kind of method: based on the time-frequency domain analytical approach of small echo, its principle is as follows:

F (t) is linear to be decomposed

The unique condition of result is, when k ≠ l

Promptly

Be orthogonal basis.

Said process is reciprocal.

There is similar results in two-parameter system

The essence of in fact Here it is wavelet transformation.

The wavelet transformation expression formula of normal use is

A ≠ 0, and a ∈ R, b ∈ R

In the formula (4)

Be called wavelet function (wavelet function), this function is the function of time of being determined by time scale a and two parameters of time displacement b.In case a and b determine, just are equivalent to determine a STFT conversion, the compiling of the corresponding one group of Short Time Fourier Transform of a series of a and b.Wavelet transformation can be analyzed time domain and frequency domain information as required simultaneously.

Multiresolution analysis (MRA) not only provides a kind of simple method for Orthogonal Wavelets structure, and provides theoretical foundation for the fast algorithm of orthogonal wavelet transformation.The information extraction that wavelet decomposition is actually different frequency in the test signal comes out, and throws on the time shaft of unified reference system and carry out comparative analysis.Select the small echo kind and decompose the number of plies according to the characteristics of signal, can carry out the analysis of time domain and frequency domain simultaneously signal.

Several variation features amounts commonly used are as follows:

Extreme difference (Range) is the poor of observed reading maximum and minimum value, promptly

range＝max{X _i}-min{X _i}(5)

Variance is the mean square deviation of observed reading apart from average, promptly

$\mathrm{var}={\mathrm{<figure-callout\; id="2"\; label="S"\; filenames="A20061013012600191.png,A20061013012600192.png"\; state="\{\{state\}\}">S</figure-callout>}}^2=\frac{1}{n-1}\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}{(X_i-\stackrel{\&OverBar;}{X})}^2---\left(6\right)$

Standard deviation: be the arithmetic square root of variance, promptly

$\mathrm{std}=\sqrt{\mathrm{var}}=\sqrt{S^2}---\left(7\right)$

Three kinds of representative point welding position shifting signals adopt DB3 wavelet decomposition result as shown in Figure 4.A7 is the approximation signal of displacement D, and d7, d6, d5 and d4 represent each layer detail signal respectively, and the detail signal below four layers mainly is a high-frequency information, does not have obvious characteristics (omission).Find by contrast: the extreme difference of the d7 details component of splash displacement signal reaches about 0.5V in Fig. 4 (b) spot welding, obviously greater than other two classes solder joint; And the extreme difference of Fig. 4 (c) raw bits shifting signal D is significantly less than other two classes solder joint, differs to reach about 1.0V, and the point quality characteristic information is obvious.Utilize the spot-wedling electrode displacement signal can evaluate quality of welding spot.

Quality of welding spot divides three classes, promptly qualified solder joint, splash solder joint and incomplete fusion or incomplete penetration solder joint.By the whether splash of anatomic observation solder joint; Determine according to GB QB2205-95 whether spot size is qualified.

Fig. 5 is the statistical distribution of DB3 small echo 7 level detail component extreme differences and spot size corresponding relation.Statistical sample is totally 197 solder joints, 3 kinds of spot-welding technology standards, corresponding B01, B04 and B05 respectively.Wherein, 117 in B01 solder joint sample, 33 in B04 solder joint sample, the splash tendency is bigger; 47 in B05 solder joint sample.More intuitively having confirmed from Fig. 5 small echo details component extreme difference histogram should the correspondence rule.

Fig. 6 is the statistical distribution of spot welding displacement signal standard deviation and spot size.Statistical sample is identical with Fig. 5 with the spot-welding technology standard.Fig. 6 spot-wedling electrode displacement signal standard deviation can distinguish incomplete fusion or incomplete penetration solder joint and splash, qualified solder joint.Near diameter (Φ 6mm) qualified and critical point that welding spot size is less than normal overlaps.

The 2nd kind of method: obtain characteristic information based on hardware or software or the direct filtering of software and hardware combining

Three kinds of representative point welding position shifting signals adopt DB3 wavelet decomposition result as shown in Figure 4.Can determine that the frequency range of a7, d7, d6, d5 and d4 correspondence is respectively [0,39Hz], [39Hz, 78Hz], [78Hz, 156Hz] and [156Hz, 312Hz] according to sample frequency (5000Hz) and each layer of wavelet decomposition frequency structure relation.

Find that by wavelet analysis there are corresponding relation in spot welding displacement signal characteristic frequency information and splash.Fig. 7 further utilizes LABVIEW to realize displacement signal Butterworth midband pass filter.Fig. 7 (a, b, c) is respectively the local midband pass filter of Fig. 4 (a, b, c), and frequency range is consistent with wavelet analysis about the extreme difference 0.5V of 40～80Hz. Fig. 7 (b) splash solder joint midband pass filter.Therefore, can extract the splash feature by hardware design.

In like manner, can adopt with quadrat method and handle the lifetime of resistance spot welding electrode pressure signal, characteristic frequency is lower than 25Hz and is higher than 312Hz at low frequency and high band.

The 3rd kind of method: based on LABVIEW multisensor resistance spot welding feature extracting method

Adopt distributed multi-sensor synchronous acquisition system to realize the aluminium alloy pinpoint welding quality monitoring.Utilize LABVIEW virtual instrument graphical language to work out the related data process software, this software has possessed functions such as data requirementization, digital filtering, data statistics, feature extraction, grade estimation and classification and information stores.Fig. 8 resistance spot welding splash solder joint voltage signal (a), displacement signal (b) and pressure signal (c).And voltage derivative signal (d), displacement differential signal (e) and welding foroe partial-band resistance filtering signal (f).Feature extraction sees attached list 1.

Table 1

The 4th step: data processing centre (DPC) carries out statistical study to every kind of sensor special time period and frequency band signal, further improves the signal to noise ratio (S/N ratio) of characteristic information;

The 5th step: data processing centre (DPC) uses statistical analysis technique or through differentiating every kind of sensing data of analytical, obtains the statistical information of statistics extreme difference, average, variance, standard deviation, mean absolute deviation;

The 6th step: data processing centre (DPC) realizes the online quality judging of direct labor or judges that by means of machine each critical point is implemented in line mass and judges according to the characteristic information and the statistical information that obtain.

The 7th step: the data analysis processing unit adopts principal component analysis (PCA), neural network method or fuzzy mathematics to realize the fusion of characteristic information in conjunction with neural net method, estimate the internal relation between a plurality of information characteristics of resistance spot welding, carry out the on-line prediction and the classification of resistance spot welding quality.

Below with principal component analytical method to prediction of quality and sorting technique.

Principal component analysis (PCA) is a kind of effective ways of analyzing many statistics, by structure and the irrelevant new proper vector of existing characteristic information, examines the information of acquisition closely, reaches and reduces information characteristics vector dimension, simplifies the purpose of characteristic information.

If x ₁..., x _mSample through standardization (average is 0)

x _kj，k＝1，…，n；j＝1，…，m

Calculate sample correlation matrix

$R=\left(\begin{array}{cccc}1& r_{12}& \&CenterDot;\&CenterDot;\&CenterDot;& r_{1n}\\ r_{21}& 1& \&CenterDot;\&CenterDot;\&CenterDot;& r_{2n}\\ \&CenterDot;& \&CenterDot;& & \&CenterDot;\\ \&CenterDot;& \&CenterDot;& & \&CenterDot;\\ \&CenterDot;& \&CenterDot;& & \&CenterDot;\\ r_{n1}& {\mathrm{<figure-callout\; id="2"\; label="r\; n"\; filenames="A20061013012600191.png,A20061013012600192.png"\; state="\{\{state\}\}">r</figure-callout>}}_n& \&CenterDot;\&CenterDot;\&CenterDot;& 1\end{array}\right)---\left(8\right)$

$r_{\mathrm{ij}}=\frac{1}{m-1}\underset{k=1}{\overset{m}{\mathrm{\&Sigma;}}}{x}_{\mathrm{ik}}{x}_{\mathrm{jk}}=r_{\mathrm{ji}},i,j=1,\&CenterDot;\&CenterDot;\&CenterDot;,n$

Secular equation is arranged

|R-λI|＝0(9)

Promptly $\left|\begin{array}{cccccc}1-\mathrm{\&lambda;}& r_{12}& \&CenterDot;& \&CenterDot;& \&CenterDot;& r_{1n}\\ r_{21}& 1-\mathrm{\&lambda;}& \&CenterDot;& \&CenterDot;& \&CenterDot;& r_{2n}\\ \&CenterDot;& \&CenterDot;& & \&CenterDot;& & \&CenterDot;\\ \&CenterDot;& \&CenterDot;& & \&CenterDot;& & \&CenterDot;\\ \&CenterDot;& \&CenterDot;& & \&CenterDot;& & \&CenterDot;\\ r_{n1}& {\mathrm{<figure-callout\; id="2"\; label="r\; n"\; filenames="A20061013012600191.png,A20061013012600192.png"\; state="\{\{state\}\}">r</figure-callout>}}_n& \&CenterDot;& \&CenterDot;& \&CenterDot;& 1-\mathrm{\&lambda;}\end{array}\right|=0$

Obtain n non-negative real root, be actually the R correlation matrix is made svd, obtain diagonal matrix, and arrange as follows according to the size of value:

λ ₁≥λ ₂≥…≥λ _n≥0

With the following equation of characteristic root substitution, ask proper vector α _i(i=1 ..., r):

$\left(\begin{array}{cccccc}1-{\mathrm{\&lambda;}}_l& r_{12}& \&CenterDot;& \&CenterDot;& \&CenterDot;& r_{1n}\\ r_{21}& 1-{\mathrm{\&lambda;}}_l& \&CenterDot;& \&CenterDot;& \&CenterDot;& r_{2n}\\ \&CenterDot;& \&CenterDot;& & \&CenterDot;& & \&CenterDot;\\ \&CenterDot;& \&CenterDot;& & \&CenterDot;& & \&CenterDot;\\ \&CenterDot;& \&CenterDot;& & \&CenterDot;& & \&CenterDot;\\ r_{n1}& {\mathrm{<figure-callout\; id="2"\; label="r\; n"\; filenames="A20061013012600191.png,A20061013012600192.png"\; state="\{\{state\}\}">r</figure-callout>}}_n& \&CenterDot;& \&CenterDot;& \&CenterDot;& 1-{\mathrm{\&lambda;}}_l\end{array}\right)\left(\begin{array}{c}{\mathrm{\&alpha;}}_1\\ {\mathrm{\&alpha;}}_{\mathrm{<figure-callout\; id="2"\; label="l"\; filenames="A20061013012600191.png,A20061013012600192.png"\; state="\{\{state\}\}">l</figure-callout>}2}\\ \&CenterDot;\\ \&CenterDot;\\ \&CenterDot;\\ {\mathrm{\&alpha;}}_{\ln }\end{array}\right)=\left(\begin{array}{c}0\\ 0\\ \&CenterDot;\\ \&CenterDot;\\ \&CenterDot;\\ 0\end{array}\right)---\left(10\right)$

It should be noted that and to obtain all proper vectors that r principal character vector gets final product before only obtaining.The method of determining r is to make

$\underset{k=1}{\overset{r}{\mathrm{\&Sigma;}}}{\mathrm{\&lambda;}}_k/\underset{k=1}{\overset{n}{\mathrm{\&Sigma;}}}{\mathrm{\&lambda;}}_k\&ap;90\%$

Can construct the major component matrix according to proper vector:

$\left(\begin{array}{ccccc}{z}_{11}& \&CenterDot;& \&CenterDot;& \&CenterDot;& z_{1m}\\ \&CenterDot;& & \&CenterDot;& & \&CenterDot;\\ \&CenterDot;& & \&CenterDot;& & \&CenterDot;\\ \&CenterDot;& & \&CenterDot;& & \&CenterDot;\\ {\mathrm{<figure-callout\; id="1"\; label="z\; r"\; filenames="A20061013012600171.png,A20061013012600172.png"\; state="\{\{state\}\}">z</figure-callout>}}_r& \&CenterDot;& \&CenterDot;& \&CenterDot;& z_{\mathrm{rm}}\end{array}\right)=\left(\begin{array}{ccccc}{\mathrm{\&alpha;}}_{11}& \&CenterDot;& \&CenterDot;& \&CenterDot;& {\mathrm{\&alpha;}}_{1n}\\ \&CenterDot;& & \&CenterDot;& & \&CenterDot;\\ \&CenterDot;& & \&CenterDot;& & \&CenterDot;\\ \&CenterDot;& & \&CenterDot;& & \&CenterDot;\\ {\mathrm{\&alpha;}}_{\mathrm{<figure-callout\; id="1"\; label="r"\; filenames="A20061013012600171.png,A20061013012600172.png"\; state="\{\{state\}\}">r</figure-callout>}1}& \&CenterDot;& \&CenterDot;& \&CenterDot;& {\mathrm{\&alpha;}}_{\mathrm{rn}}\end{array}\right)\left(\begin{array}{ccccc}{x}_{11}& \&CenterDot;& \&CenterDot;& \&CenterDot;& x_{1m}\\ \&CenterDot;& & \&CenterDot;& & \&CenterDot;\\ \&CenterDot;& & \&CenterDot;& & \&CenterDot;\\ \&CenterDot;& & \&CenterDot;& & \&CenterDot;\\ {\mathrm{<figure-callout\; id="1"\; label="x\; n"\; filenames="A20061013012600171.png,A20061013012600172.png"\; state="\{\{state\}\}">x</figure-callout>}}_n& \&CenterDot;& \&CenterDot;& \&CenterDot;& x_{\mathrm{nm}}\end{array}\right)---\left(11\right)$

To the many proper vectors in the subordinate list 1, adopt principal component analysis (PCA), first three major component proportion of the new proper vector that makes up is respectively 54.4%, 32.7% and 6.2%, amounts to 93.3%.In other words, can utilize original nine characteristic informations of three proper vector approximate expressions of new structure.Thereby reach the purpose of many information fusion and data compression.

Fig. 9 has realized that for utilizing first principal component and Second principal component, point quality is judged and classification.Different solder joints are distributed in three zoness of different respectively.205 in sample, 106 of wherein qualified solder joints, 54 of interior splash solder joints, 45 of incomplete fusion or incomplete penetration solder joints.

Drop on totally 95 of the interior solder joints of circle among Fig. 9, have only 1 interior splash mistake to fall wherein, judging nicety rate is 98.9%; Circle left side triangle is interior splash solder joint, has only 1 incomplete fusion solder joint to be included in wherein in 54 solder joints, and judging nicety rate is 98.1%; Circle right side five-pointed star is incomplete fusion or incomplete penetration solder joint, comprises 12 qualified solder joints in 56 solder joints, and judging nicety rate is 78.6%.

Figure 10 is the oscillograph monitored signal, as can be seen the relation between quality of welding spot and electrode displacement, the welding foroe.

By said method, can effectively extract the characteristic information of resistance spot welding quality and merge, and on this basis, realize on-line prediction and classification resistance spot welding quality

Advantage of the present invention and good effect are:

1. structure a kind of resistance spot welding multisensor quality monitoring device and the feature information of the present invention's proposition are carried Get and information fusion method, quality-monitoring and judgement that can the canbe used on line resistance spot welding, and can be to point Feature information in the weldering process is analyzed effectively, and feature information signal to noise ratio height, the feature extracted are bright Aobvious, good stability.

2. the resistance spot welding multisensor quality monitoring device that proposes of the present invention only adopts four kinds of sensors or its In certain several sensor or a kind of sensor wherein, it is reasonable in design, is convenient to on-the-spot I﹠M, because of This its practical, implementation cost is low, be easy to apply.

3. the resistance spot welding quality monitor sample extracted of the present invention has advantages such as being convenient to expansion, is suitable for model Enclose wide.

Claims (9)

1. resistance spot welding quality monitoring device, constitute by resistance spot welding equipment body, last horn, following horn, top electrode, bottom electrode, it is characterized in that: a displacement transducer is installed in the rigid support top that top electrode installs, the lower surface of this displacement transducer is parallel with the metal plate that last horn outside level is stretched out, a pressure transducer is installed in bottom at following horn, and displacement transducer and pressure transducer all are connected with equipment data acquisition analyzing by shielded cable.

2. resistance spot welding quality monitoring device according to claim 1, it is characterized in that: a voltage sensor also is installed on top electrode and bottom electrode, in the bottom electrode outside one current sensor is installed also, the two all is connected with equipment data acquisition analyzing by shielded cable.

3. resistance spot welding quality monitoring device according to claim 1 is characterized in that: described displacement transducer is an eddy displacement sensor; Described pressure transducer is highly sensitive quartzy electrostrictive strain extreme pressure force transducer, and the two all is connected with equipment data acquisition analyzing by shielded cable.

4. resistance spot welding quality monitoring device according to claim 2 is characterized in that: described voltage sensor is a LEM Hall effect electrode voltage sensor; Described current sensor is that a skeleton adopts flexible material, the air core coil of the number of turn between 500 to 50000 circles.

5. resistance spot welding quality monitoring device according to claim 1, it is characterized in that: described equipment data acquisition analyzing comprises that data acquisition process unit, Data Transmission Controlling unit and data analysis processing unit constitute, wherein, the data acquisition process unit is made of displacement transducer unit, voltage sensor unit, current sensor unit, pressure sensor unit, these four sensor units all are connected with the Data Transmission Controlling unit by serial line interface, and this Data Transmission Controlling unit is connected with data processing centre (DPC) by parallel interface.

6. resistance spot welding quality monitoring device according to claim 5 is characterized in that: described serial line interface can be the combination of RS232 interface, RS485 interface or RS232 and RS485 interface.

7. the monitoring method of a resistance spot welding quality as claimed in claim 1 is characterized in that: this method realizes by following step:

(1) the data acquisition process unit gather in real time displacement, voltage, electric current, four kinds of data of pressure or wherein the starting point of the pinpoint welding procedure that arrives of certain several data or wherein a kind of data based sensor monitors and terminating point as the reference point of data cutout, intercepting and normalization sensing data;

(2) synchro control of sensor data acquisition is realized data acquisition process unit is carried out in the Data Transmission Controlling unit, and normalized sensing data is transferred to data processing centre (DPC);

(3) data processing centre (DPC) extracts the point quality characteristic information according to the position and the scope of every kind of sensor characteristics information time of origin of sensing data analysis and frequency;

(4) data processing centre (DPC) carries out statistical study to every kind of sensor special time period and frequency band signal, further improves the signal to noise ratio (S/N ratio) of characteristic information;

(5) data processing centre (DPC) uses statistical analysis technique or through differentiating every kind of sensing data of analytical, obtains the statistical information of statistics extreme difference, average, variance, standard deviation, mean absolute deviation;

(6) data processing centre (DPC) realizes the online quality judging of direct labor or judges that by instrument each critical point is implemented in line mass and judges according to the characteristic information and the statistical information that obtain;

(7) the data analysis processing unit adopts principal component analysis (PCA), neural network method or fuzzy mathematics to realize the fusion of characteristic information in conjunction with neural net method, estimate the internal relation between each information characteristics of resistance spot welding, carry out the on-line prediction and the classification of resistance spot welding quality.

8. the monitoring method of resistance spot welding quality according to claim 7, it is characterized in that: the mode of described collection voltage also can directly be got voltage between spot welding top electrode and the bottom electrode as voltage signal, usually secondary voltage is less than 10V, behind electric resistance partial pressure, can be connected by direct and follow-up modulate circuit.

9. the monitoring method of resistance spot welding quality according to claim 7 is characterized in that: the method for described extraction point quality characteristic information can use time-frequency domain analytical approach based on small echo, based on hardware or software or the hard soft method that combines with software, based on LABVIEW multisensor resistance spot welding feature extracting method.

Images