Modeling the Inspection Task

Ted Schorn

To improve on the results observed in the measurement of visual inspection effectiveness, it is important to understand the inspection process more clearly.

During typical training associated with quality control, the inspection process is treated as an indivisible unit of activity. When the quality of the inspection is discussed, the topics usually are regarding gage error (such as gage repeatability or reproducibility studies) or establishing acceptance standards and criteria. The act of inspection is rarely investigated or discussed in the training of quality engineers, but this causes inspection to be viewed as a “black box.” Unfortunately, when guesses are made about what might be inside the box, much time is spent guessing about the impact of inspector diligence or integrity on the results. This is misguided thinking and quite harmful. Questioning the inspector’s diligence or integrity is demeaning and demoralizing. It also provides a convenient escape route for organizational accountability for the inspection results: little can be done by management to improve employee diligence or attention to detail so banners and speeches substitute for true process improvement.

One needs to understand that as a quality practitioner one is responsible to create a positive environment for the employees and to equip them for the tasks that are included within the production plan. The employer has the responsibility for the final product quality results and the care of the individuals, who are hired to get the work done. The decision to use humans as inspectors ought to be deliberate, and employers should plan for their capabilities, capacities and limitations.

Blaming the inspectors—identifying them as the root cause of the problem—is irrational and a poor root cause analysis by any measure. To do a proper job in the analysis of inspection error improvement opportunity means one must understand what is in the “black box”—how inspection proceeds and what influences it. To do this, the inspection task needs to be broken down into its elements.

Inspection begins with the selection and physical movement of the part to the inspection location. This may require the part to be pulled out of a bin, from a conveyor or some other container. This part may be obtained as a deliberate selection process, such as a sampling plan or some other identified routine for randomized selection, or it may be a 100% inspection of all parts. Identifying this task as an element in the process may appear trivial, but it is important to note. Proper training for this manual function is important for its relationship to operator fatigue due to poor ergonomics, errors in omitting parts from the inspection process, or inspecting the wrong lot or wrong quantity from the lot. Eventual decision making on lot quality may depend on the selection randomization and fairness. There are some inspections where this operation is absent, such as viewing parts on a moving conveyor belt or performing inspections on objects too large to physically handle. In such cases, the inspection routine begins with the inspector entering the work area and presenting him/herself to the part(s) rather than the other way around.

After the part is brought into the inspector’s workspace for viewing, (e.g., on an inspection table or stand) it must be oriented in some manner. This “sets up” the initial look at the part. Often this occurs very quickly and without actual planning.

With experienced and well-trained inspectors, the entire set of part movements is choreographed and the part is always placed on the table or held in the same way to begin the inspection. It is at that point that the eyes scan the part from the vantage offered by the initial orientation. Visual search is the mental and physical process of visual probing that allows the eye to have an opportunity to put the anomaly of interest into the active field of view. Understanding the function of the human visual sensory system is necessary here and many people have inaccurate views on just how people capture images and “see.”

For example, often the word “scanning” is used to describe visual search, a mental picture of an inspector’s eyes following a raster path (a straight line, back and forth path like that used in reading) as they search the part’s surface area visual. That mental image is far from the truth. Inspection search paths are more similar to random wandering unless specific training or other guidance is provided.

Searching is limited by the visual sensory system of the inspector—eyesight. The brain must receive the signal that something is present in the field of view. This includes characteristics of static and dynamic visual acuity, color vision and depth perception. Once the brain acknowledges the sensory input, the mind must then identify and classify the finding. Mentally or sometimes aloud, the inspector goes through a moment of identification and classification such as, “That is an inclusion” or “That looks like a paint run.”

Once the discontinuity or item of interest has been so classified, the brain once more must decide—“Is this okay or not and what do I do with this?” It is important to distinguish between the two distinct mental processes here. The mind must understand what is seen before it can decide if that anomaly is okay.

Disposition decisions are quite unlike classification decisions; they are more about judgment than recognition. As a result, they are influenced by a whole host of subjective factors as well as the objective information presented in standards, acceptance criteria and limit samples.

Depending on the complexity and physical geometry of the part, the inspector may view it from multiple angles or orientations, each time looping through the operations of searching, seeing, classifying and deciding status. Good quality control practice dictates that the entire part be inspected before a final disposition is made (lest the part be cycled to a repair center only to have a later examination reveal other defects, or perhaps a fatal flaw, resulting in greater waste of time and revenue).

Once the inspection views are completed and a disposition is made of the entire article the part can be dispatched. The dispatch of the part includes the ergonomics of physically moving the part and handling it to avoid damage or deterioration by returning it into the production stream or physically segregating it into some location or holder as nonconforming. The recording and information task includes the identification and labeling of nonconforming or suspect parts, documenting the nonconformance and perhaps notification or alarm steps to provide immediate feedback to the production process.

The Nature of Inspection Tasks
The task analysis includes two major skill types, manual and perceptual. Among the perceptual skills, the mental attributes of attention, memory, recognition, classification and judgment are required.

Manual tasks properly performed:

  • Minimize fatigue and stress that can impair judgment, reduce focus and create errors of both false alarms and misses.
  • Minimize damage to product associated with handling defects.
  • Establish an efficiency of movement associated with lower overall “floor to floor” cycle times.
  • Lead to a higher incidence of effective control of nonconforming product flowing out from the inspection operation.
  • Assist in the maintenance of the planned sequence of views of the part during the inspection.

The perceptual tasks requiring involvement of the cognitive functions are those that contribute substantially to the errors made in inspection. Specifically, two elements hold the key to visual inspection improvement, visual search and deciding the status of the item. These are also the two elements toward which most of the visual inspection literature is aimed, both in understanding the factors that influence these elements but also in the proper approach towards the training and selection of inspectors for the job.

It may come as no surprise that decision making is of crucial influence, but visual search? The model, which Wang and Drury have developed and many others have utilized, highlights the need for inspectors to put their eyes on the anomalies before making a judgment. Thinking about experience with misses and what type of parts are returned by customers, it is common to think that the inspector decided to let something go or “wasn’t paying attention” when missing something. Defects missed often are just that; the inspector never saw them because a consistent visual search routine was not followed. An inspector may not be wrong when saying, “I never laid eyes on that defect,” after being shown a customer return. Developing and standardizing a visual search routine can have a most significant impact on visual inspection effectiveness.

Introducing the tasks that represent visual inspection can also be important when considering the job description and qualifications for inspectors. These include: specific physical capabilities (i.e., lifting a certain weight repetitively); fine motor skills, standing for long time periods; certain sensory capacities (i.e., a specific visual acuity or color discrimination capability); and specific mental abilities (i.e., visual and content memory, classification capability and decision making based on general criteria).

It is assumed that each element of the inspection task is important and worth discussing. The model forces one to treat the subject of human factors, which are not usually well-treated in classic quality control texts. The model also requires us to understand inspection from new perspectives, such as decision-making theory. The model assumes that improvement will also be the result of a solid foundation in ergonomics and industrial engineering, aspects of which are now much in vogue as part of lean manufacturing, though other aspects are ignored and at times are viewed as archaic, such as time and motion studies.

This model provides a better explanation for the actual inspection results. It ought to explain empirical results if it is useful as a model—it ought to explain real life. Contemplating the focus of the model on the human activity (and not just standards and criteria) forces an understanding of what is happening to the inspectors within their heads and hearts. This is a more compassionate view than many take in classic quality assurance/quality control (QA/ QC), and this is the right direction. Yet, it is not only compassionate, but more realistic. Human factors study is not just about how humans go about doing what they do, but also about why they make mistakes. It is known that “to err is human” but coming to grips with human error and working to minimize it while keeping the human at the center of the decision-making apparatus is both right and humbling. The author is not an advocate of error proofing techniques and the practice the Japanese methods, called poka yoke. These approaches remove the human and institute fail safes that anticipate human mistakes and engineer them out or 100% automatically check for them and in many cases, correct them automatically if they are detected. But all of this hardly applies in human visual inspection. The state-of-the-art in sensory inspections of all kinds (e.g., a mechanical wine-tasting machine) is such that machines cannot do what humans can do in this arena. Therefore, one must learn new things and adopt a very humble and personal approach to this task of visual inspection improvement.

One must be ready to increase planning activity as new areas are found where control, which has not been noted previously, needs to be exercised. There must be readiness to dig into the factors that influence visual inspection effectiveness.  

This article is based on an excerpt of Schorn’s book Improving the Effectiveness of Visual Inspection, and material from Schorn’s course he will be holding Feb. 26-27 at AFS headquarters in Schaumburg, Illinois. To sign up, visit www.afsinc.org. 

Click here to see this story as it appears in the February 2019 issue of Modern Casting