Poka-yoke examples: preventing the error where it occurs

Why it matters

If a screw can be tightened even though a cover has been placed reversed, the defect cannot be secured just by asking the worker to be more careful. Error-proofing, commonly called poka-yoke, studies arrangements that make a wrong action harder to occur, or a defective result harder to move forward, by preventing it or making it visible. Error and defect are separated: reversed placement is the wrong action; a product that leaks or closes incorrectly may be the result. To find the most useful point of intervention, one goes back from the final check to the work where the error occurs. Device selection starts by describing clearly which action can be done wrongly under which condition. A general defect name does not by itself form a design requirement. The condition in which the error becomes easier is also observed. Parts that look alike, symmetric seats or seating surfaces that cannot be seen can encourage a wrong choice. Giving the worker a reminder to be careful does not change these conditions. The design should make it easy to bring the right part to the right position while determining how the wrong option is handled. When the target error is clearly defined, it also becomes clear which acceptable and unacceptable samples are needed for the trial. Finding a defect afterward and preventing it from occurring in the first place are recorded as separate outcomes.

Detection produces the information that tells an unacceptable condition apart. Prevention makes the wrong placement or the next operation impossible unless a certain condition is met. A warning produces a signal for a person to see, hear and respond to appropriately. These do not by themselves provide the same level of assurance. A sensor can see wrong placement; but whether this information stops production is examined separately. In a device that only sounds, if the worker misses the sound, the product can move on. A physical stop can close off the wrong orientation; but it may not test another error such as a missing part at all. It must be explained which error condition the arrangement does not cover as well as what it does.

In real applications, detection and prevention can be combined in the same device. For this reason, when counting types, the classification unit and rule are written in advance. Adding one trial as detection, prevention and warning all together does not increase the trial count; it counts the same event three times. In the count in this lesson, each trial is placed in a single class according to the primary function being tested. If functions are combined, the auxiliary features are kept as notes. A table that closes the trial count is not the same as a success rate. That a type has been tested many times does not prove it is more reliable. The calculation of the false-signal rate and choosing verification frequency by risk are the subject of M17. Detection, physical prevention and warning are the training classes used here for the primary function; they are not a universal, mutually exclusive technical classification of all error-proofing arrangements.

Error-proofing is not made a substitute for the protective devices needed for worker safety. If a fixture that protects quality creates a hand-trapping hazard, the design is not complete. Testing with a defective part is also done under conditions that will not mix into production and will not put a person at risk. Known defective samples are clearly separated and kept under control when the work is finished. After maintenance, a product changeover or a fixture adjustment, the earlier successful record alone may not be sufficient. The affected function is verified again. Continuing production by disabling the sensor does not solve an unwanted stoppage; it creates a new way of working in which a defect can pass silently. Data notice for this lesson's figures, examples, counterexamples and three short quiz questions: This dataset was generated synthetically for training; it does not represent a real business, employee or customer study.

Formula