Process Capability for Deep Hole Drilling: Using Cp and Cpk to Measure Your Process

A customer walked into my shop five years ago and asked for capability data on a hydraulic spool bore. I had been drilling deep holes for fifteen years and could hold 18 mm +/- 0.02 mm in my sleep, but I had never calculated a Cp or Cpk in my life. I fumbled through that first report with a spreadsheet and a statistics textbook open on my phone. The numbers that came back surprised me – and they changed how I set up every job from that point forward.

Cp and Cpk are not just numbers you put on a PPAP to satisfy a customer requirement. They tell you whether your process is actually centered in the tolerance band and whether the natural variation of your process fits inside the print limits. For deep hole drilling, where the tool is cutting blind and the bore gets measured after the part comes off the machine, those numbers are the closest thing you have to real-time feedback on whether your setup is working.

What Cp and Cpk Actually Measure

Cp measures how wide your process spread is compared to the tolerance band. It is the tolerance width divided by six times your process standard deviation. If your Cp is 1.0, your process variation takes up the entire tolerance band. If it is 1.33, your variation uses 75% of the band. If it is 1.67, you are using 60%.

Cpk adds a second layer: it also accounts for whether your process average is centered in the tolerance band. You can have a Cp of 2.0 – plenty of room – but if your process average is shifted off center, your Cpk will be lower. The formula is the smaller of two values: (USL minus mean) divided by 3 sigma, or (mean minus LSL) divided by 3 sigma.

The difference between Cp and Cpk tells you how much room you have to improve just by centering better. If Cp is 1.8 and Cpk is 0.9, your process spread is fine but your average needs to move. If both numbers are low, your machine or process has too much variation and centering alone will not fix it.

I check both on every capability study. Cp tells me whether my tool, machine, and parameters are capable of holding the print. Cpk tells me whether I have dialed in the setup correctly.

Sample Size and Data Collection for Deep Hole Drilling

The textbook says 30 parts minimum for a capability study. I have found that 25 to 30 holes is enough for deep hole drilling, provided you measure each hole at the right locations. Running fewer than 20 holes and your confidence interval is too wide to trust the result.

The more important question is which measurements to take and where. For deep holes, a single diameter measurement per hole hides taper and ovality. I measure at three depths – 25%, 50%, and 75% of the bore length. At each depth I take two readings 90 degrees apart to catch ovality. That gives me six diameter data points per hole, which I average for the hole-level Cpk calculation.

For straightness and surface finish I take fewer readings per hole because those characteristics vary more between holes than within a single hole. Straightness in particular is dominated by machine alignment and tool geometry, not by small variations along the bore.

I measure every hole in the study consecutively from the same bar stock batch. Changing material batches mid-study introduces a variable you cannot separate from the process. I also run the study on the same machine, same operator shift, and same tool design to isolate the process variation.

Calculating Capability for Diameter, Straightness, and Surface Finish

I calculate three separate Cpk values for every deep hole drilling job: one for diameter, one for straightness, and one for surface finish. They rarely come out the same, and that is the useful part.

Diameter Cpk is the most straightforward because the tolerance is well defined. Most deep hole prints call an H7 or H8 fit, which gives a clear upper and lower spec limit. I take the 18 diameter readings from 25 holes, calculate the grand average and pooled standard deviation, and plug them into the Cpk formula. I have seen diameter Cpk values from 0.8 on a marginal gun drilling setup up to 2.1 on a well-tuned BTA machine with fresh guide pads.

Straightness is harder because it is one-sided. There is no lower spec limit – the hole cannot be too straight. I use the one-sided Cpk formula, which is essentially (USL minus mean) divided by 3 sigma. A straightness Cpk of 1.33 means the worst hole in the study is still well inside the print limit.

Surface finish Cpk follows the same one-sided approach for a maximum Ra requirement. I calculate it separately from diameter because the variables that drive surface finish – feed rate, coolant pressure, tool edge condition – are different from the variables that drive diameter.

I record all three in my process documentation and track them over time. A drop in diameter Cpk paired with a stable straightness Cpk usually points to guide pad wear. A drop in surface finish Cpk alone points to the cutting edge geometry.

Interpreting Capability Results

Here is the table I keep taped to my toolbox. It has not changed in five years because the math does not change, only the process does.

I do not trust a single Cpk calculation from one study. I run three consecutive studies on the same job and look at the trend. If all three come back above 1.33, I lock in the setup parameters and move to reduced sampling. If the numbers bounce around – 1.6 one week, 1.1 the next – I have an unstable process that needs root cause work before capability matters.

Different industries have different minimum requirements. I have been asked to meet all four of these at various points:

The aerospace and medical numbers look the same on paper but the difference is in the process validation. Aerospace customers typically want to see the study protocol and the raw data. Medical customers want to see the gage R&R and the measurement uncertainty budget alongside the Cpk.

What to Do When Cpk Is Too Low

Low Cpk breaks down into two cases: the spread is too wide, or the process is off center. I check the Cp first because it tells me which case I have.

If Cp is low, the natural variation of the process exceeds what the tolerance allows. For deep hole drilling, the usual suspects are:

• Tool condition: a worn gun drill or BTA head produces larger diameter variation. I regrind or replace the tool and rerun the study.
• Machine alignment: headstock-to-bushing misalignment creates tapered bores. I check the alignment with a test bar.
• Coolant pressure and filtration: inconsistent coolant delivery causes chip packing and diameter spikes. I verify the pressure at the tool tip, not at the pump gauge.
• Workpiece clamping: a part that moves during the cut introduces variation that looks like process spread. I check the clamping setup.

If Cp is acceptable but Cpk is low, the process average is shifted. I adjust the tool offset or the reaming allowance and recenter. A 0.01 mm offset in a 20 mm bore with an H7 tolerance (0.021 mm) can turn a 1.6 Cpk into a 0.8 Cpk. Centering is free and it has the biggest impact on Cpk of any single adjustment.

I also check the measurement system before changing the process. An air gage with a dirty filter or a micrometer that has not been calibrated in six months will produce a low Cpk that does not actually exist. I run a gage R&R study annually and before any customer-facing capability report.

How Process Capability Changes With Tool Wear

This is the part that surprised me when I started tracking capability. A new gun drill might give you a Cp of 1.8 on the first 10 holes. By hole 80 on the same tool, Cp can drop to 1.0. The process spread widens as the cutting edge wears and the guide pads lose their burnished surface.

I track Cpk over tool life on every production job. The data lets me set tool change intervals based on capability instead of guessing. On a hydraulic valve body job running 4140 pre-hard, I found that diameter Cpk dropped below 1.33 at exactly 120 holes and surface finish Cpk followed at 150 holes. I changed the tool change frequency from 200 holes to 120 holes and the scrap rate dropped from 2% to 0.3%.

The pattern is consistent: tool wear shifts the process average first, then widens the spread. The shift happens because the cutting edge erodes and the effective diameter shrinks. On a gun drill, the diameter drops by roughly 0.002 mm to 0.005 mm over the life of the tool depending on the material. That shift alone cuts Cpk in half on a tight tolerance bore.

The spread widens because the worn tool interacts differently with material variations. A new tool cuts the same whether the material is at the low end or high end of the hardness spec. A worn tool cuts differently on hard spots and the diameter variation increases.

Tracking capability over tool life has saved me more money than any single process change I have made. It tells you exactly when to change the tool, which is never at the same interval the tool supplier recommends.

For more detail on what tolerances you can expect from each deep hole drilling process, read my guide on deep hole drilling accuracy. For the full SPC methodology including control charts and out-of-control rules specific to deep hole drilling, see the SPC for deep hole drilling guide.

Key Takeaways

• Cp measures process spread versus tolerance width. Cpk adds centering. The gap between them tells you which problem to fix.
• Run 25 to 30 consecutive holes from the same material batch. Measure diameter at three depths and two orientations per hole to catch taper and ovality.
• Calculate separate Cpk values for diameter, straightness, and surface finish. Different variables drive each one.
• A Cpk of 1.33 is the minimum for most production work. Aerospace and medical typically require 1.67.
• When Cpk is low, check Cp first to decide if the problem is spread or centering. Fix the measurement system before fixing the process.
• Tool wear shifts the process average and widens the spread over time. Track Cpk over tool life and set change intervals based on capability data, not guesswork.