Overcoming pierce point error: lines do not converge

[louisd]

Hi guys, looking to make a robust method to prevent the "lines do not converge error" on some constructed pierce points (line into circle).

I've got pics here https://photos.app.goo.gl/1cbhG8y8X4DRHVzg6 (kirbs, work your magic to post please, as I have 1kb left of my 1mb allowance. lol.
two arcs, four lines to construct four pierce points. The parts are small (0.09" diameter hole)
I think some parts have flash on the edges of the line-arc intersect point, so the probe is shanking out and producing completely incorrect circles.

I want to know if there's a way for me to anticipate lines do not converge error, then add a loop to grow the diameter, or shift Y value of circle, until lines do converge.

Here's the code I tossed together, forces generic diameter and YZ locations based upon robust inner circle that's not segmented (ID1), if any of the measured values become outliers (>0.003" or so).

CIR1 =FEAT/CONTACT/CIRCLE/DEFAULT,CARTESIAN,IN,LEAST_SQR
THEO/<0.148,0,0>,<1,0,0>,0.0985
ACTL/<0.148,0,0>,<1,0,0>,0.0985
TARG/<0.148,0,0>,<1,0,0>
START ANG=342,END ANG=378
ANGLE VEC=<0,1,0>
DIRECTION=CCW
SHOW FEATURE PARAMETERS=NO
SHOW CONTACT PARAMETERS=NO
CIR2 =FEAT/CONTACT/CIRCLE/DEFAULT,CARTESIAN,IN,LEAST_SQR
THEO/<0.148,0,0>,<1,0,0>,0.0985
ACTL/<0.148,0,0>,<1,0,0>,0.0985
TARG/<0.148,0,0>,<1,0,0>
START ANG=18,END ANG=-18
ANGLE VEC=<0,-1,0>
DIRECTION=CW
SHOW FEATURE PARAMETERS=NO
SHOW CONTACT PARAMETERS=NO
$$ NO,
Below code does the following (with goal to prevent erroneous intersect points).
V5 & V6= Finds smallest Y location deviation from zero, between cir1 and cir2, and outputs that value (used to locate ID2 Y value if MID_IDCIRS is > 0.004" off).
V7= Finds the average diameter of cir1 and cir2. if either cir is different from ID1 cir by 0.003" or more, it forces ID1's measured value to create the average.
V8= Finds the average Z location of Cir1 and Cir2. If either circles are off by more than 0.003" it forces the Z value to 0 for that circle z.
ASSIGN/V5=MIN(ARRAY(ABS(CIR1.Y),ABS(CIR2.Y)))
ASSIGN/V6=IF(V5==ABS(CIR1.Y),CIR1.Y,CIR2.Y)
ASSIGN/SMALLDIA1=SORTUP(ARRAY(CIR1.DIA,ID1.DIA))
ASSIGN/SMALLDIA2=SORTUP(ARRAY(CIR2.DIA,ID1.DIA))
ASSIGN/C1_DIAM=IF(ABS(SMALLDIA1[2]-SMALLDIA1[1])>=0.003,ID1.DIA,CIR1.DIA)
ASSIGN/C2_DIAM=IF(ABS(SMALLDIA2[2]-SMALLDIA2[1])>=0.003,ID1.DIA,CIR2.DIA)
ASSIGN/V7=(C1_DIAM+C2_DIAM)/2
ASSIGN/V8=(IF(ABS(CIR1.Z)>=0.003,0,CIR1.Z)+IF(ABS(CIR2.Z) >=0.003,0,CIR2.Z))/2
MID_IDCIRS =FEAT/POINT,CARTESIAN,NO
THEO/<0.148,0,0>,<1,0,0>
ACTL/<0.148,0,0>,<1,0,0>
CONSTR/POINT,MID,CIR2,CIR1
ASSIGN/MIDLOC=IF(ABS(MID_IDCIRS.Y)>=0.004,V6,MID_IDCIRS.Y )
ID2 =GENERIC/CIRCLE,DEPENDENT,CARTESIAN,OUT,$
NOM/XYZ,<0.148+0,CIR1.TY,CIR1.TZ>,$
MEAS/XYZ,<0.148+0,CIR1.Y,CIR1.Z>,$
NOM/IJK,<1,0,0>,$
MEAS/IJK,<1,0,0>,$
RADIUS/0.0985+0,C1_DIAM
ID3 =GENERIC/CIRCLE,DEPENDENT,CARTESIAN,OUT,$
NOM/XYZ,<0.148+0,CIR2.TY,CIR2.TZ>,$
MEAS/XYZ,<0.148+0,CIR2.Y,CIR2.Z>,$
NOM/IJK,<1,0,0>,$
MEAS/IJK,<1,0,0>,$
RADIUS/0.0985+0,C2_DIAM
MID_ID =GENERIC/CIRCLE,DEPENDENT,CARTESIAN,OUT,$
NOM/XYZ,<0.148+0,MID_IDCIRS.TY,MID_IDCIRS.TZ>,$
MEAS/XYZ,<0.148+0,MIDLOC,V8>,$
NOM/IJK,<1,0,0>,$
MEAS/IJK,<1,0,0>,$
RADIUS/0.0985+0,V7
AXISLINE =FEAT/LINE,CARTESIAN,UNBOUNDED,NO
THEO/<0.03,0,0>,<1,0,0>
ACTL/<0.03,0,0>,<1,0,0>
CONSTR/LINE,BF,3D,ID1,MID_ID,,
OUTLIER_REMOVAL/OFF,3
FILTER/OFF,WAVELENGTH=0
MOVE/INCREMENT,<-0.4,0,0>
MOVE/INCREMENT,<0,0,0.2>
PART_ALIGN_6DOF=ALIGNMENT/START,RECALL:STARTUP,LIST=YES
ALIGNMENT/LEVEL,XPLUS,AXISLINE
ALIGNMENT/ROTATE,YPLUS,TO,INIT_ROTLINE2,ABOUT,XMINUS
ALIGNMENT/TRANS,YAXIS,ID1
ALIGNMENT/TRANS,ZAXIS,ID1
ALIGNMENT/TRANS,XAXIS,FRONT_PLANE
ALIGNMENT/END

ID_LINE1 =FEAT/CONTACT/LINE/DEFAULT,CARTESIAN,UNBOUNDED
THEO/<0.148,0.0466,0.053>,<0,-0.0436,-0.999>,<0,-0.9990482,0.0436194>,<-1,0,0>,0.025
ACTL/<0.148,0.0466,0.053>,<0,-0.0436,-0.999>,<0,-0.9990482,0.0436194>,<-1,0,0>,0.025
TARG/<0.148,0.0466,0.053>,<0,-0.0436,-0.999>,<0,-0.9990482,0.0436194>,<-1,0,0>
SHOW FEATURE PARAMETERS=NO
SHOW CONTACT PARAMETERS=NO
ID_PNT1 =FEAT/POINT,CARTESIAN,NO
THEO/<0.148,0.0481,0.086>,<0,-0.0436194,-0.9990482>
ACTL/<0.148,0.0481,0.086>,<0,-0.0436194,-0.9990482>
CONSTR/POINT,PIERCE,ID_LINE1,ID2
ID_LINE2 =FEAT/CONTACT/LINE/DEFAULT,CARTESIAN,UNBOUNDED
THEO/<0.148,0.0466,-0.053>,<0,-0.0436,0.999>,<0,-0.9990482,-0.0436194>,<-1,0,0>,0.025
ACTL/<0.148,0.0466,-0.053>,<0,-0.0436,0.999>,<0,-0.9990482,-0.0436194>,<-1,0,0>,0.025
TARG/<0.148,0.0466,-0.053>,<0,-0.0436,0.999>,<0,-0.9990482,-0.0436194>,<-1,0,0>
SHOW FEATURE PARAMETERS=NO
SHOW CONTACT PARAMETERS=NO
ID_PNT2 =FEAT/POINT,CARTESIAN,NO
THEO/<0.148,0.0481,-0.086>,<0,-0.0436194,0.9990482>
ACTL/<0.148,0.0481,-0.086>,<0,-0.0436194,0.9990482>
CONSTR/POINT,PIERCE,ID_LINE2,ID2
ID_LINE3 =FEAT/CONTACT/LINE/DEFAULT,CARTESIAN,UNBOUNDED
THEO/<0.148,-0.0466,-0.053>,<0,0.0436,0.999>,<0,0.9990482,-0.0436194>,<-1,0,0>,0.025
ACTL/<0.148,-0.0466,-0.053>,<0,0.0436,0.999>,<0,0.9990482,-0.0436194>,<-1,0,0>,0.025
TARG/<0.148,-0.0466,-0.053>,<0,0.0436,0.999>,<0,0.9990482,-0.0436194>,<-1,0,0>
SHOW FEATURE PARAMETERS=NO
SHOW CONTACT PARAMETERS=NO
ID_PNT3 =FEAT/POINT,CARTESIAN,NO
THEO/<0.148,-0.0481,-0.086>,<0,0.0436194,0.9990482>
ACTL/<0.148,-0.0481,-0.086>,<0,0.0436194,0.9990482>
CONSTR/POINT,PIERCE,ID_LINE3,ID3
ID_LINE4 =FEAT/CONTACT/LINE/DEFAULT,CARTESIAN,UNBOUNDED
THEO/<0.148,-0.0466,0.053>,<0,0.0436,-0.999>,<0,0.9990482,0.0436194>,<-1,0,0>,0.025
ACTL/<0.148,-0.0466,0.053>,<0,0.0436,-0.999>,<0,0.9990482,0.0436194>,<-1,0,0>,0.025
TARG/<0.148,-0.0466,0.053>,<0,0.0436,-0.999>,<0,0.9990482,0.0436194>,<-1,0,0>
SHOW FEATURE PARAMETERS=NO
SHOW CONTACT PARAMETERS=NO
ID_PNT4 =FEAT/POINT,CARTESIAN,NO
THEO/<0.148,-0.0481,0.086>,<0,0.0436194,-0.9990482>
ACTL/<0.148,-0.0481,0.086>,<0,0.0436194,-0.9990482>
CONSTR/POINT,PIERCE,ID_LINE4,ID3

[andrewb91]

You might try creating reverse lines.

[louisd]

all the line vectors point towards the center of the part, so that shouldn't have any impact.

[AndersI]

You could check the shortest distance between cir2 and each line (in that order). If dist < cir2 radius the line will pierce cir2, if dist == cr the line will be a tangent, if dist > cr the line will not touch cir2 at all.
(If I have understood the problem correctly)

[louisd]

Upon further investigating I found the star probe was bent a bit, causing shank-hits. I replaced the probe, and it seems to have entirely mitigated the issue. I like your concept though.

[Scott Staral @ WAFCO]

You can create an alignment at the center of the circle and do alignment lines. I do this all the time.

[louisd]

Ok re-hashing my post... Circle math is good, but the line vectors are periodically flipping on me enough to make the math failed errors continue.

I think I figured out a solution. If the J vector of the line is less than we will say, 25% of theoretical (0.03 J) it should produce the math error.
So I'm adding an if statement. If J is > (-0.03), and shift the line and remeasure it 0.008" closer to the known parting line, which will increase the vector, to be back in line with theoretical/expectation.

Big question now is, if I added this code, creating duplicate auto-line feature ID's like this, it will overwrite all values of the first execution, upon the IF statement being true, and skipping entirely, upon the IF statement being false, correct?! If i'm wrong, what's the workaround?

[COLOR=#FF0000][B]ID_LINE1[/B][/COLOR] =FEAT/CONTACT/LINE/DEFAULT,CARTESIAN,UNBOUNDED
THEO/<0.148,0.0466,0.053>,<0,[COLOR=#FF8C00][B]-0.0436[/B][/COLOR],-0.999>,<0,-0.9990482,0.0436194>,<-1,0,0>,0.025
ACTL/<0.148,0.0466,0.053>,<0,-0.0436,-0.999>,<0,-0.9990482,0.0436194>,<-1,0,0>,0.025
TARG/<0.148,0.0466,0.053>,<0,-0.0436,-0.999>,<0,-0.9990482,0.0436194>,<-1,0,0>
SHOW FEATURE PARAMETERS=NO
SHOW CONTACT PARAMETERS=NO
[B]IF/[COLOR=#FF8C00]ID_LINE1.J [/COLOR][/B]> (-0.03)
[COLOR=#FF0000][B]   ID_LINE1[/B][/COLOR] =FEAT/CONTACT/LINE/DEFAULT,CARTESIAN,UNBOUNDED
     THEO/<0.148,0.0466,[COLOR=#008000][B]0.045[/B][/COLOR]>,<0,-0.0436,-0.999>,<0,-0.9990482,0.0436194>,<-1,0,0>,0.025
     ACTL/<0.148,0.0466,[COLOR=#008000][B]0.045[/B][/COLOR]>,<0,-0.0436,-0.999>,<0,-0.9990482,0.0436194>,<-1,0,0>,0.025
     TARG/<0.148,0.0466,[COLOR=#008000][B]0.045[/B][/COLOR]>,<0,-0.0436,-0.999>,<0,-0.9990482,0.0436194>,<-1,0,0>
     SHOW FEATURE PARAMETERS=NO
     SHOW CONTACT PARAMETERS=NO
[B]END_IF/[/B]

[TK_Fire]

Loop until you get a line that works maybe?

Label
evaluate/adjust line
Line
If bad GOTO Label

may need to add variable in there for the first run

[louisd]

I like it!
assign variable to be initial line Z nominal,
do loop,
Measure line with variable assigned to Z.
re-assign variable value to shift Z closer to flash/parting line
until J vector of line is > 0.03

that makes one feature ID to measure, until it lines up. brilliant!
I'll have to add some default case to protect it from falling too low, but its a great plan. thanks for the suggestion TK_Fire