Plotting Proctor curves 1

[kentuckyengineer]

I recently ran into a situation where we performed modified Proctor testing and the best fit parabolic curve as determined from gINT was less than the maximum point that was plotted. It was probably 3 or 4 pcf less than what the maximum point on the best fit curve was. What are some tolerances or standard rules of thumb for this situation.

 

[dgillette]

3 or 4 pcf seems like a lot to miss by. I don't think there is any physical reason or ASTM requirement that it has to be a parabola. I did a quick look in Joseph Bowles' 1978 lab testing book (my copy is about to fall apart from use), Jack Hilf's chapter in Winterkorn and Fang (1975 edn.), and the USBR testing manual, and did not see any mention of fitting a parabola. I've always seen the technician just push his (occasionally her) pencil in a smooth curve through the dots, which is what the books all seem to call for. This may be a situation where the "convenience" of GINT drawing the curve for you is no convenience at all. It doesn't take much effort for a human to do it, and it adds one more place for someone to look at the reasonableness of the data.

If the material has very few fines (on the edge of where the Proctor test applies), you may not get a distinct peak. In this case, the fitted parabola may bear little resemblance to the data. Even more reason to have a human draw the curve.

 

[oldestguy]

dqillette is correct.

Matter of fact, most Proctor curves have a right side parallelling the zero voids theoretical curve. That bends the other way, opposite from a parabola's right side line.

Modern stuff can't match mother nature too well.

 

[GeoPaveTraffic]

I agree with the others, plot the curve by hand.

Never believe what a software package is telling you if you "just know" it is not right.

 

[Panars]

I haven't used gINT, but why would it insist on fitting a parabolic curve to your data? I did write an Excel macro program for plotting proctor points and fitting a SPLINE curve to the points. The curve goes right through each point. Basically, the macro fitted a quadratic to each set of three points and then blended the different curves together. I was then able to convert the curve to a dry density curve and determine the point of max dry density.

My point is, you don't have to do it by hand, but if you try to do it mathematically, you need to know what you are doing.

 

[oldestguy]

One more way.

Do it on Autocad and see what it does for curves. Probably like Panars so nicely explained. I use Autocad for contours and they work out well hitting all the points.

 

[BigH]

I like to draw this curve (and grain size curves) by hand - computers always seem to be somewhat off on their plotting (although I admit it might be my less than sophisticated way with them. I suggest, though, and it wasn't mentioned, that the ZAV (zero air voids line - i.e., 100% saturation line) be plotted on ALL compaction plots. If your curve goes onto the other side (right side) of it, something is amiss. A very well respected geo-firm once got hit up on this in a major way.

 

[Ron]

I agree with the others about software plotting. gINT is actually pretty good, but someone else touched on a key to a correct curve...the zero air voids curves. The right side of optimum (wet side) should be nearly parallel to the curves. The left side of optimum (dry side) can take a variety of shapes, even double peaking, depending on the material and gradation.

By DEFINITION in ASTM D1557, the curve is to be plotted as a "smooth curve connecting the points". The maximum dry density and optimum moisture are graphically determined points from the plotted curve. In software, the curve is a mathematical "best fit" with the point of inflection (1st derivative of the slope) being the MDD.

In hand plotted curves, the MDD is usually higher than any of the plotted points, because we try to bracket the optimum moisture, not hit it directly, in the test.

I'll take my trusty little 6-inch pointy french curve over the software every time.

You won't generally see that kind of variation you mentioned between two different people doing the test on a split sample (in the same lab).