{"id":2336,"date":"2026-09-17T10:10:44","date_gmt":"2026-09-17T02:10:44","guid":{"rendered":"https:\/\/www.jccncmachining.com\/?p=2336"},"modified":"2026-09-15T16:26:12","modified_gmt":"2026-09-15T08:26:12","slug":"cnc-process-capability-cpk","status":"publish","type":"post","link":"https:\/\/www.jccncmachining.com\/ru\/blog\/cnc-process-capability-cpk\/","title":{"rendered":"\u041f\u0440\u043e\u0448\u0435\u0434\u0448\u0430\u044f \u0434\u0435\u0442\u0430\u043b\u044c \u0432\u0441\u0451 \u0435\u0449\u0451 \u043c\u043e\u0436\u0435\u0442 \u0441\u043a\u0440\u044b\u0432\u0430\u0442\u044c \u043d\u0435\u0441\u0442\u0430\u0431\u0438\u043b\u044c\u043d\u044b\u0439 \u043f\u0440\u043e\u0446\u0435\u0441\u0441 \u0427\u041f\u0423"},"content":{"rendered":"

<\/p>\n

CNC process capability describes whether a stable machining process can repeatedly hold a defined specification; Cpk estimates how well the observed process is centered within its upper and lower limits.<\/strong> A first article can pass while the process behind it is drifting, off-center, or producing too much variation. That is why one conforming part and a capable production process are different claims.<\/p>\n

For a buyer, the useful question is not \u201cWhat is your Cpk?\u201d in isolation. It is: for which characteristic, from which machine and setup, over what period, using what sample, with what measurement system, and after what stability checks? A polished index calculated from mixed or manipulated data can conceal more risk than it reveals.<\/p>\n

Three parts tell three different stories<\/h2>\n

\"three-process-stories\"\nImagine three production lots for the same shaft diameter. Each lot contains a part measuring exactly at nominal. In the first lot, all results cluster tightly around nominal. In the second, the average sits near the upper specification limit. In the third, values swing in a repeating pattern as the machine warms and the tool wears. The nominal part from each lot looks identical, but the processes are not equally predictable.\n\nCapability analysis uses a distribution of results because manufacturing decisions depend on variation, not on a showcase sample. It becomes especially valuable for repeat production, automated runs, mating features, and characteristics where a small shift can create assembly failures. The first article inspection<\/a> establishes an approved starting point; capability work examines whether routine output can stay there.<\/p>\n

Cp measures room; Cpk measures room and position<\/h2>\n

\"cp-cpk-visualized\"\nWhen a process is approximately stable and the analysis assumptions are appropriate, Cp compares the specification width with the estimated natural process spread. Cpk also considers how close the process mean is to the nearest specification limit.\n\nCp = (USL \u2212 LSL) \/ 6\u03c3<\/strong><\/p>\n

Cpk = minimum of [(USL \u2212 mean) \/ 3\u03c3] and [(mean \u2212 LSL) \/ 3\u03c3]<\/strong><\/p>\n

USL and LSL are the upper and lower specification limits, and \u03c3 represents estimated process standard deviation. A high Cp with a much lower Cpk signals that the spread may fit inside the tolerance but the process is not centered. If the average continues to move, conforming output can become nonconforming even though the theoretical spread looks narrow.<\/p>\n\n\n\n\n\n\n\n\n
Pattern in the data<\/th>\nWhat the indices may show<\/th>\nEngineering interpretation<\/th>\n<\/tr>\n<\/thead>\n
Tight and centered<\/td>\nCp and Cpk are similar<\/td>\nVariation is small and the mean is reasonably positioned.<\/td>\n<\/tr>\n
Tight but near one limit<\/td>\nCp higher than Cpk<\/td>\nCentering, offset strategy, or drift requires attention.<\/td>\n<\/tr>\n
Wide but centered<\/td>\nBoth indices low and similar<\/td>\nThe process spread consumes too much tolerance.<\/td>\n<\/tr>\n
Trend, cycle, or sudden shift<\/td>\nA single index can be misleading<\/td>\nInvestigate stability before making a capability claim.<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n

Cp and Cpk do not change the drawing tolerance. They summarize observed process behavior relative to those limits. A supplier should never widen a specification, delete inconvenient data, or mix unlike process streams merely to improve the number.<\/p>\n

Do not calculate capability until the process is stable<\/h2>\n

\"stable-process-pattern\"\nCapability assumes the data represent a consistent process. CNC machining often contains time-related signals: spindle and coolant temperatures rise, cutting edges wear, offsets are adjusted, bar stock changes, fixtures are reloaded, and operators restart production after interruptions. If these events create trends or shifts, one average and standard deviation may not describe what the next part will do.\n\nA time-ordered run chart or appropriate control chart should be reviewed before interpreting Cpk. Points outside control limits, repeating cycles, long trends, or distinct clusters suggest special causes or mixed populations. The correct response is to understand those causes, not to calculate a more favorable index from the same unstable set.\n\nSubgrouping also matters. Combining results from two machines, several cavities or fixture stations, different material lots, or day and night shifts may create a distribution that belongs to no actual process. The analysis plan should preserve the sources a buyer would need to control.\n\n\n

The measurement system can inflate or hide variation<\/h2>\n

\"measurement-system-variation\"\nObserved variation contains both manufacturing variation and measurement variation. If an instrument lacks suitable resolution, the method is sensitive to operator pressure, or the fixture bends the part, the calculated spread may describe the inspection process as much as the CNC process.\n\nBefore requesting a formal capability result, confirm:\n\n\n