CN102430841A - Arc welding robot laser vision seam tracking control method based on offline planning - Google Patents

Abstract

The invention relates to an arc welding robot laser vision seam tracking control method based on offline planning. An online robot seam tracking task is completed by combining information such as a prior model of a target to be measured given by seam tracking offline planning, a planned seam tracking movement trace and the like and comprehensively using a sensor detection and predication mechanism, a transmission delay compensation mechanism, an online seam path filtering mechanism and an online 6-dimensional welding trace synthesis mechanism. The sensor planning technology is introduced into offline programming, so that the onsite teaching problem can be solved, and a template of a total solution to entire seam tracking can be provided from the macroscopic aspect on the principle that excellent sensor detection, welding gun welding and robot operating state are ensured by taking the sensor planning technology as a common manner of systematic problem solving. The sensor only needs to effectively correct a part of links of the template on line in real time to finally automatically and symmetrically solve the robot seam tracking problem.

Description

Arc welding robot laser vision Weld Seam Tracking Control method based on off-line planning

Technical field

What the present invention relates to is the method in a kind of robot self adaptation welding applications field, particularly a kind of arc welding robot laser vision Weld Seam Tracking Control method based on off-line planning.

Background technology

The current welding robot overwhelming majority belongs to on-the-spot teaching type of the first generation or second generation off-line programing type; But no matter be on-the-spot teaching type programming or off-line planning programming; A critical problem when practical application, all will facing is exactly the adaptability of the program of establishment for on-the-spot actual environment; The most important reason that this problem occurs then is that various actual features in the field Welding environment are with respect to the variation of corresponding ideal factor of when programming; Especially in the mode of production of mass; Welding object unpredictable error on pose and size then is a most important reason, and wherein existing processing and the position while welding that error caused of fit on and the static change of size also have workpiece in the welding process to be heated and the dynamic deformation of the welding bead that change caused of radiating condition in addition.Address the above problem two kinds of thinkings are generally arranged; A kind of is to improve the machining accuracy of workpiece, the assembly precision that improves frock clamp and the strict mode of controlling the robot teaching track through employing to reduce the error in environment and the application; To improve the manufacturing cost of enterprise significantly but do like this, and the time loss cost.Therefore the self adaptation welding manner just becomes economy and the important techniques that guarantees robot welding quality and further hoisting machine people Automation of Welding and intelligent degree, and THE WELD SEAM TRACKING TECHNOLOGY becomes basic and crucial technology in the self adaptation welding because of its problem that has solved " welding gun departs from weld seam ".

The welding sensing is the maximum technology of butt welded seam tracking technique influence on development.Wherein, Optical sensor is (applicable like MIG/MAG aspect applicable craft and object; TIG, multiple welding procedures such as plasma, Laser Welding, and the welding object of multiple material such as carbon steel, aluminium alloy); Aspects such as accuracy of detection and real-time (generally reaching 0.5mm levels of precision and 50HZ at least) and aspect, application scenario have advantage, are the soldered sensors of dominating THE WELD SEAM TRACKING TECHNOLOGY future development.And be that the visual sensing of initiatively light source has certain advantage with laser in practical application; Show that mainly it can carry out the measurement of weld seam three-dimensional information; Therefore can detect the various complicated weld seams that have curvature in theory, its three-dimensional measurement ability and robot welding are used and are comparatively adapted to (the general requirement of robot welding can be carried out the welding under the multiple pose) in addition, and laser-vision sensing is except being applied to weld joint tracking in addition; Weld seam is proofreaied and correct before can also being applied to weld; Robot welding line is from main programming, and the guiding of weld seam initial point is located, and many practical application such as welding quality Non-Destructive Testing of postwelding.

Laser vision weld joint tracking in the past generally can be divided into two big types, and a kind of being called as " on-the-spot teaching type ", on behalf of type, it mainly the MTR seam tracking system of Britain Meta Vision company is arranged; A kind of in addition then is to be called as " autonomous following-up type ", and on behalf of type, it mainly Digi-IROBONET-MASTER V300A is arranged.Wherein " on-the-spot teaching type " requires manual work to consider simultaneously at the scene under the situation of welding gun and sensing head constraint; Shortcomings such as butt welded seam carries out the segmentation teaching, and it is big therefore to demonstrate the teaching difficulty for some complicacies or multiple-pass weld, and the teaching workload is big; Its requirement takies field apparatus in addition; Actual teaching quality is relevant with operating personnel's working experience, makes the obvious lack of wisdomization of this kind mode, and is comparatively backward.And " autonomous following-up type " generally only needs after operating personnel are provided with according to the characteristics of field welding simply; Promptly guiding the tracking that welding gun is accomplished butt welded seam by sensing head; Though this kind mode can promote the intelligent degree of weld joint tracking on degree greatly; But needs of robot welding are taken all factors into consideration welding object, technology, the systemic problem of factors such as equipment and environment; " autonomous following-up type " then is that the main control power of whole welding robot has been transferred to sensor; Therefore but sensor can only be made a strategic decision in real time according to the detection information of part online and instructed the motion of welding gun, not only (collision of sensing head and weldment might occur, perhaps move towards and can't hold for the macroscopic view that has than the weld seam of deep camber aspect the welding surroundings assurance of periphery, having blind spot; Move and have little time to accomplish to follow the tracks of); And can't expect (it is spacing to be about to get into the joint like robot, perhaps is absorbed in the work singular point) for the state that will occur of robot, not enough for some comparatively complicated welding procedure adaptability in addition.

Summary of the invention

The object of the invention provides a kind of arc welding robot laser vision Weld Seam Tracking Control method based on off-line planning.

For achieving the above object, the technical scheme that the present invention adopts is:

A kind of arc welding robot laser vision Weld Seam Tracking Control method based on off-line planning, carry out according to the following steps:

(1), off-line planning: through planning weld joint tracking track, welding line joint skeleton pattern to be measured to obtain in advance;

(2), the current weld image of sampling, and carry out outline, to obtain current welding line joint profile with described welding line joint skeleton pattern to be measured;

(3), described current welding line joint profile is carried out continuous detecting, the weld seam path point of extraction two dimension;

(4), communication delay between the sampling time point of the described current weld image of compensation and the welding gun pose point time point that pairing with it vision controller obtains, and obtain the weld seam path point of three-dimensional through the sensing head measurement model;

(5), after the described three-dimensional weld seam path point process filtering of continuous extraction, obtain level and smooth actual welds pursuit path;

(6), combine described weld joint tracking track and the online movement locus of actual welds pursuit path generation robot welding line tracking planned in advance to weld with the drive machines people.

Preferably, in step (3), adopt the prediction search window that described current welding line joint profile is carried out continuous detecting.

Further preferably, described prediction search window is carried out according to the following steps:

(1), predicting tracing point position;

(2), predicting tracing point shape;

(3), generation forecast search window: synthetic and obtain prediction view profile collection with as the formation base of predicting the region of search through the position and the shape of trace point of prediction in the step (1), (2), and should predict that the region of search was defined as the prediction search window.

Preferably, in step (5), adopt the weld seam path filters to filter.

Because the technique scheme utilization, the present invention compared with prior art has advantage and effect:

1, the present invention is under the situation of obtaining the off-line programing result; Can realize robot self adaptation welding very simply based on laser vision sensor; On-the-spot only need be behind the pre-planning result who imports off-line programing artificially and obtained (difference between environment and the off-line programing simulated environment does not have under the situation of the bigger qualitative change of generation at the scene), robot will automatically implement the self adaptation welding step;

2, the invention is not restricted to common straight bead and follow the tracks of, suitable equally for plane curve and space curve weld seam, joint categories equally can be various even unconventional in addition, is some weld seams that in checking present technique scheme, adopted;

3, the present invention can guarantee reliability and the accuracy that butt welded seam is followed the tracks of under starting the arc situation, is carrying out under the curve weld joint tracking situation, on the basis that guarantees the effective shape of curved welding seam, can suppress effectively noise.

Description of drawings

Accompanying drawing 1 is the definition of weld joint tracking motion model relative coordinate system;

Accompanying drawing 2 is the Weld Seam Tracking Control block diagram;

Accompanying drawing 3 is contour prediction coupling block diagram;

Accompanying drawing 4 is outline position prediction sketch map;

Accompanying drawing 5 is that joint profile search window generates sketch map;

Accompanying drawing 6 is a system operation time model sketch map;

Accompanying drawing 7 is set up sketch map for the track correspondence;

Accompanying drawing 8 is a KUKA robot welding line tracking control flow chart.

The specific embodiment

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

A kind of arc welding robot laser vision Weld Seam Tracking Control method based on off-line planning:

As shown in Figure 1, set up corresponding coordinate system: comprise the world coordinate system W that the Z axle is opposite with gravity direction all the time; Having represented the robot basis coordinates of robot motion's working space reference is B; Represented the weld seam coordinate system J of weld seam characteristic; Having represented and being based upon the most advanced and sophisticated tool coordinates of welding wire is H; Having represented the sensor eye coordinates of weld joint tracking lower sensor viewpoint pose is S; Represented with the plane of delineation and formed single laser coordinate system L that should concern; Camera coordinate system C in the line-structured light sensing head and plane of delineation coordinate system I,

As shown in Figure 2, this method is carried out according to the following steps:

(1), off-line planning: accomplish and plan weld joint tracking track, welding line joint skeleton pattern to be measured in advance;

(2), the current weld image of sampling, and carry out outline, to obtain current welding line joint profile with welding line joint skeleton pattern to be measured;

(3), as shown in Figure 3: mainly through the prediction of outline position constantly (is designated as to t ^Ip _{P, t|t+1}) and the prediction of shape (be designated as ^IV _{PM, t|t+1}) the scope prediction of actual joint profile in new image that comprehensively provide (promptly predict search window, be designated as ^IA _{P, t|t+1}) to reduce the time and the space complexity of strong jamming bottom profiled coupling, specifically:

A, predicting tracing point position:

As shown in Figure 4: as be located at t+1 and obtain image constantly, and the weld seam filtering path that obtains simultaneously before this moment to do ^Bc _{FSP, t}, further corresponding predicted path is designated as ^Bc _{P, t|t+1}, and have ^Bc _{P, t|t+!}= ^Bc _{FSP, t}, obtain the welding gun pose of current time in addition

And current known structure light face parameter vector ^Cn _L, then do through trick relation and the structured light face parameter vector that the welding gun pose can be easy to obtain under the B coordinate system

$n_B=n_C\·T_H^C\·T_{t+1}^B{}_H,$

So according to the weld seam predicted path ^Bc _{P, t|t+!}With structured light face equation ^BN ^BThe p=0 simultaneous solution then can obtain the weld seam path point under the B coordinate system ^Bp _{P, t|t+1}Value, pass through formula so at last

$p_{D,t|t+1}^I=T_L^{I}T_H^{L}T_{t+1}^B{}_{H}p_{P,t|t+1}^B,$

Just can obtain the weld seam path imaging point on correspondence image plane, just outline position ^Ip _{P, t|t+1}

B, predicting tracing point shape:

Profile adopts the mode of directly inheriting, and predicts that promptly contour shape is the contour shape that previous frame successfully detects:

^IV _PM，t|t+1＝ ^Iv _D，t，

Wherein ^Iv _{D, t}Be t detection profile result constantly;

C, generation forecast search window:

The synthetic prediction view skeleton pattern collection that obtains through position and shape ^IQ _P=( ^IV _PM, ^Ip _P) with formation base, will predict that in addition the region of search unification is called prediction search window SW as the prediction region of search, it specifically can be expressed as ^ISW _P=( ^IQ _P, ^IA _P), wherein ^IA _PBe expressed as the size (being the scope of region of search) of SW,

For a skeleton pattern collection ^IQ _P=( ^IV _PM, ^Ip _P),, and expand to obtain search window as skeleton with it through the mode of Fig. 5 ^ISW _P=( ^IQ _P, ^IA _P), wherein ^IA _P=(a _X, a _Y), and a _XAnd a _YBe respectively along image coordinate system line direction and column direction region of search size parameter value,

Right in addition ^IA _PThe mode of the employing dynamic self-adaptingization of value, for t constantly promptly ^IA _{P, t-1|t}, get

^IA _P，t-1|t＝(1-f((t-t _L)/ΔT))A _PRI+f((t-t _L)/ΔT)A _PRX

A wherein _PRIAnd A _PRXBeing respectively minimum and maximum allows ^IA _PValue (this value actual needs with profile in whole tracing process in image change in location relevant), t _LThe detection success time point that the current future position of expression distance is nearest promptly satisfies t _L＜t, and Δ T is expressed as the maximum continuous detecting Time To Failure interval (general requirement can not surpass (future range)/(speed of welding)) that allows, and function f () just is expressed as the value proportion function, is taken as

$f\left(x\right)=\left\{\begin{array}{cc}{x}^3& x\≤1\\ 1& x>1\end{array}\right.,$

After successfully detecting the welding line joint profile, just can therefrom extract the weld seam path point of two dimension ^Lp _SP, next then need further it to be transformed among the operable working space B of robot, promptly

$p_{\mathrm{DSP},t}^B=T_t^H{}_{B}T_L^{H}p_{\mathrm{DSP},t}^L;$

(4), because the structural limitation of real system (being that vision controller and robot controller non-integral are integrated); Therefore make and strictly to be mapped because of the cause of big communication delay (current employing is based on the serial communication mode of RS232) between image acquisition time point and the robot end's pose acquisition time point; And then can cause measure error, and influence final tracking accuracy;

For correlation delay is compensated, need the time model of analytical system, consider system operation time model as shown in Figure 6, wherein T _RBe robot motion's servo period, Δ T _RBe the delay between the pose delivery time point of slave computer refresh time point and host computer, T _PFor the host computer pose transmits cycle, Δ T _PFor host computer with the transmission delay between the vision controller, T _VBe whole visual servo cycle, wherein T _GBe image sampling cycle, T _IBe visual processes cycle, t _{R, i}Be robot servo time point, t _{P, j}Be robot delivery time point, t ' _{P, j}Be the robot pose point time of advent, t _{G, k}Be image sampling time point, t ' _{G, k}For thread obtains the image time point,

Requirement is got up image acquisition time point and robot acquisition time point synchronously, adopts the timing of vision controller end can obtain t ' _{G, k}, so the estimated image sampled point

$t_{G,k}=t_{G,k}^{\′}-\frac{T_G}{2},$

Can obtain robot delivery time point t ' equally _{P, j}, then estimate robot pose time point

t _R，i＝t′ _P，j-ΔT _P-ΔT _R/2，

Further can obtain and t through linear interpolation _{G, k}Pairing robot pose point estimate

$T_t^H{}_B=t_{R,k}=(1-\mathrm{\λ})t_{R,i}+\mathrm{\λ}{t}_{R,i+1},$

Wherein

$\mathrm{\λ}=\frac{t_{G,k}-t_{R,t}}{t_{R,i+1}-t_{R,i}};$

(5), establishing t moment filter window size for the local detection weld seam path sequence of L does

^BPA _DSP，t＝{ ^Bp _DSP，i，i＝1，...，L}，

To adopt cubic polynomial ^Bc _{FSP, t}=a ₃t ³+ a ₂t ²+ a ₁T+a ₀Weighted fitting comes right ^BP _{DSP, t}Carry out smoothly, the optimization aim of match does

$\min \left(\underset{i=1}{\overset{L}{\mathrm{\Σ}}}{w}_{\mathrm{DSP},i}\right|p_{\mathrm{FSP},i}^B-p_{\mathrm{DSP},i}^B\left|\right),$

Specifically can adopt Generalized Inverse Matrix or based on the solving method of LQ decomposition algorithm, the weights sequence of test point is designated as in the wherein corresponding filter window

^BW _DSP，t＝{ ^Bw _DSP，i，i＝1，...，L}，

And each weights is calculated as

$w_{\mathrm{DSP},i}^B={\mathrm{con}}_{\mathrm{DSP},i}/\underset{i=1}{\overset{L}{\mathrm{\Σ}}}\mathrm{con}_{\mathrm{DSP},i}^B,$

Wherein ^BCon _{DSP, i, t}Be the normalization confidence level sequence in the filter window ^BCon _{DSP, t}= ^BCon _{DSP, i}, the i=1...L} element is specially ^BCon _{DSP, i, t}=| [c _Dc _T] ^T, wherein || || the length of vector, and path point confidence level vector [c _Dc _T] mainly by detection confidence c _DWith effective confidence level c of time _TForm,

And detection confidence c _DNeed take all factors into consideration in the present image signal to noise ratio to provide in real time; And for the time window that detects the path in the filter window ^BPA _DSP= ^BP _{DSP, i}, i=1...L}, effective confidence level c of time _TBe calculated as

$c_{T,i}=\frac{T_{\mathrm{ti}}-t}{T_{\mathrm{ti}}},$

T wherein _TiBe maximum effective time of window threshold value, must not surpass " future range/current maximum movement speed " as the one of which, promptly preposition tracking maximum effective period;

(6), obtain 2 adjacent on weld seam path points for filtering ^Bp _{FSP, i}With ^Bp _{FSP, i+1}, it can be by following formula

^Bc _FSP，i(λ)＝a _3，iλ ³+a _2，iλ ²+a _1，iλ+a _0，i，

Carry out the weld seam path curve and describe, satisfy following constraints simultaneously

^Bc _FSP，i(0)＝ ^Bp _FSP，i， ^Bc _FSP，i(1)＝ ^Bp _FSP，i+1，

$\stackrel{.}{c}_{\mathrm{FSP},i}^B\left(0\right)=\stackrel{.}{p}_{\mathrm{FSP},i}^B,$ $\stackrel{.}{c}_{\mathrm{FSP},i}^B\left(1\right)=\stackrel{.}{p}_{\mathrm{FSP},i}^{+1}{}_B,$

Wherein

$\stackrel{.}{p}_{\mathrm{FSP},i}^B=\left|\right|v(i,i+1)\left|\right|\frac{(v_{\mathrm{NORMAL}}^B(i,i+1)+v_{\mathrm{NORMAL}}^B(i-1,i))}{\left|\right|v_{\mathrm{NORMAL}}^B(i,i+1)+v_{\mathrm{NORMAL}}^B(i-1,i)\left|\right|},$

V (i, j)= ^Bp _{FSP, j}- ^Bp _{FSP, i}, v _NORMAL(i j) is v (i, normalized vector j).Unite above-mentioned constraints, and can calculate these 3 parameters of curve according to

and do

$a_{3,i}=\stackrel{.}{p}_{\mathrm{FSP},i+1}^B+\stackrel{.}{p}_{\mathrm{FSP},i}^B-2p_{\mathrm{FSP},i+1}^B+2p_{\mathrm{FSP},i}^B,$

$a_{2,i}=-\stackrel{.}{p}_{\mathrm{FSP},i+1}^B-2\stackrel{.}{p}_{\mathrm{FSP},i}^B+3p_{\mathrm{FSP},i+1}^B-3p_{\mathrm{FSP},i}^B,$

$a_{1,i}=\stackrel{.}{p}_{\mathrm{FSP},i}^B,$

d _0，i＝ ^Bp _FSP，i，

Then final weld seam path curve does ^BC _FSP= ^Bc _{FSP, i}, i=1...M-1}.

And for the motion mode (current various industrial machine philtrums carry out the motion mode of complicated track) that adopts the straightway piecewise approximation commonly, by the weld seam path curve ^BC _FSPSet out the path of welding point of its up-to-date generation ^Bp _{FH, j=J}Can be taken as at the weld seam path curve ^BC _FSPOn the upper edge welding direction of advance a bit, and this point need satisfy condition

| ^Bp _FH，J- ^Bp _FH，J-1|＝TH _L，

TH wherein _LDesired segment length when straightway approaches motion,

Further establishing the online robot welding track that has generated does

^BTR _H＝{ ^BT( ^Br _H，n， ^Bp _H，n) _H，n，n＝1...N-1}，

Therefore for the online weld joint tracking welding pose point that will generate ^BT ( ^Br _{H, n}, ^Bp _{H, n}) _{H, n=N}, wherein get the position ^Bp _{H, N}= ^Bp _{FH, J}, and its attitude ^Br _{H, N}Then by the pre-planning track ^BTR _HM= ^BT _{HM, k}, k=1...K} decides, and can problem separated be converted into " at track so ^BTR _HMLast search and tracing point position ^Bp _{H, N}Corresponding position ", as shown in Figure 7, can adopt " characteristic " that be used as weld seam path itself along " travel distance " with respect to the weld seam starting point on the direction of advance of weld seam path, therefore for weld seam path point ^Bp _{H, N}, establish corresponding pre-planning path ^BPA _HM= ^Bp _{HHM, k}, among the k=1...K} a bit do ^Bp _{HM, G}, then both should meet the following conditions

$\underset{i=0}{\overset{G}{\mathrm{\Σ}}}\left|\right|p_{\mathrm{HM},i+1}^B-p_{\mathrm{HM},i}^B\left|\right|=\underset{i=0}{\overset{N}{\mathrm{\Σ}}}\left|\right|p_{H,i+1}^B-p_{H,i}^B\left|\right|.$

The below for example practical implementation in the KUKA robot:

6DOF welding robot KUKA KR16 is adopted in practical implementation; Its controller is KRC2; For a weld seam object, concrete basic procedure comprise " pre-planning of robot welding line tracking task---seam tracking system imports the pre-planning result---search weld seam starting point before the actual welding---weld start position guiding---actual welds trackings---is carried rifle after welding completion automatically ".Wherein, as shown in Figure 8:

A, the pre-planning of robot welding line tracking task: in robot Off-line Programming System, design and build the robot welding scene; Utilize the weld joint tracking mission planning module of having developed to accomplish relevant planning action then, final output can supply the employed program results Parameter File of seam tracking system;

B, seam tracking system import the pre-planning result: before carrying out the actual welds tracking, through the process software interface operation in the vision control system, be the relevant configuration file that system imports corresponding welding scene by manual work;

C, search weld seam starting point: current mainly is to obtain actual weld seam start position through the mode of the automatic searching and detecting of sensor; The method that is adopted mainly is that sensing head carries out searching and detecting along the searching route (current employing straightway path) of pre-planning, and successfully to detect for the first time the weld seam path point that obtains as the weld seam starting point;

D, weld start position guiding: the robot welding gun is navigated to the weld seam starting point place of being detected, in this process, the leading portion weld seam is detected simultaneously, prepare thereby do initialization for weld joint tracking;

E, actual welds are followed the tracks of: adopt above-mentioned steps (1)-(6);

F, welding destination county are carried rifle: after welding was accomplished, system just moved in the other direction along the welding gun axis direction and puies forward distance on certain.

The foregoing description only is explanation technical conceive of the present invention and characteristics, and its purpose is to let the personage who is familiar with this technology can understand content of the present invention and enforcement according to this, can not limit protection scope of the present invention with this.All equivalences of doing based on spirit of the present invention change or modify, and all should be encompassed within protection scope of the present invention.

Claims (4)

1. arc welding robot laser vision Weld Seam Tracking Control method based on off-line planning is characterized in that: carry out according to the following steps:

(1), off-line planning: through planning weld joint tracking track, welding line joint skeleton pattern to be measured to obtain in advance;

(2), the current weld image of sampling, and carry out outline, to obtain current welding line joint profile with described welding line joint skeleton pattern to be measured;

(3), described current welding line joint profile is carried out continuous detecting, the weld seam path point of extraction two dimension;

(4), communication delay between the sampling time point of the described current weld image of compensation and the welding gun pose point time point that pairing with it vision controller obtains, and obtain the weld seam path point of three-dimensional through the sensing head measurement model;

(5), after the described three-dimensional weld seam path point process filtering of continuous extraction, obtain level and smooth actual welds pursuit path;

(6), combine described weld joint tracking track and the online movement locus of actual welds pursuit path generation robot welding line tracking planned in advance to weld with the drive machines people.

2. the arc welding robot laser vision Weld Seam Tracking Control method based on off-line planning according to claim 1 is characterized in that: in step (3), adopt the prediction search window that described current welding line joint profile is carried out continuous detecting.

3. the arc welding robot laser vision Weld Seam Tracking Control method based on off-line planning according to claim 2, it is characterized in that: described prediction search window is carried out according to the following steps:

(1), predicting tracing point position;

(2), predicting tracing point shape;

(3), generation forecast search window: synthetic and obtain prediction view profile collection with as the formation base of predicting the region of search through the position and the shape of trace point of prediction in the step (1), (2), and should predict that the region of search was defined as the prediction search window.

4. the arc welding robot laser vision Weld Seam Tracking Control method based on off-line planning according to claim 1 is characterized in that: in step (5), adopt the weld seam path filters to filter.

Images