Forget parts—focus on the process
April 11, 2006
Most manufacturers measure or test parts to verify that the parts meet quality standards. This conventional approach is time-consuming because testing adds steps and time to the production process. Furthermore, it is only as good as the sample size. A different approach to quality is to use a strain monitor to measure strain on the machine's frame. Comparing the strain with a reference (measured when the machine was known to be producing good parts) is a way to monitor the production process, and it doesn't require extra time or steps.
|The crimp force monitor is sensitive enough to indicate a process failure when a single sheet of paper disrupts the crimping cycle. This level of sensitivity isn't practical for most manufacturing operations, but it does demonstrate the process variation monitor's capability.|
It's time to take a fresh look at the way manufacturers monitor quality in their facilities. A new perspective is especially critical for simple processes such as tube bending, end forming, riveting, and crimping, in which the manufacturer must produce parts at "parts-per-second" speeds for only "pennies-per-part" prices. Failing to ensure absolute quality in such circumstances can result in significant penalties, containment, reduced profits, or lost business.
Most of the current quality assurance tests for forming, fabricating, and assembling processes are simply dimensional measurements or physical inspections. In some situations, inspection personnel use any of a variety of inspection devices to measure individual parts. The primary purpose of such inspection devices is to catch gross failures and prevent the affected parts from getting shipped to the customer. In some cases, quality control personnel take sample parts from a production batch and subject them to more accurate or thorough physical and dimensional inspections.
Either way, relying purely on dimensional measurements or visual inspections has some major inherent shortcomings. These processes don't necessarily detect material flaws (thickness or hardness variations or tiny cracks). For processes that include some assembly steps, a dimensional check doesn't ensure that hidden components are indeed assembled to the part. Such checks cannot uncover inadequate lubrication before stamping, bending, or forming, which may not affect the part's resultant dimensions, but certainly can affect the part's quality, reliability and life cycle. Intensive offline testing, either destructive or nondestructive, is necessary for uncovering these additional production quality deficiencies.
Sample testing offers a bit of added confidence in the product's quality, but if one part out of a hundred is bad, it is unlikely that a bad part will be selected for the sample. Even a large sample size, such as 50 samples out of 100 parts, leaves a 50 percent chance of discovering a problem in the one bad part, and then the entire batch would be considered suspect.
However, sample testing combined with some statistical analysis indicates a process's capability. In this case, the object is not to determine the presence of failed parts in the batch, but rather to determine the degree of variability in the parts produced. The smaller the variability, the more confidence the manufacturer has that the process is producing only good parts.
Monitoring the process for variations, rather than checking the parts for flaws, is an entirely new way of looking at quality assurance. The basic assumption behind the concept of process variation monitoring is simple: If a process (including inputs, force, and tooling) is known to be capable of producing only good parts, and the process is stable and repeatable, the output (good parts) should be stable too. The key is to determine whether the process is stable and, therefore, capable of producing good parts.
For many processes in which parts are formed, shaped, or fastened by applying force, a simple force-resistance-strain relationship exists. If the force is known to be constant (in a mechanical press, for instance), the strain on the machine is directly related to resistance to that force. The resistance to the force includes all the other process variables, such as the raw material (thickness and hardness), lubrication, tooling condition, and even machine condition and parameters (the condition of the bearings, clutches, and other moving parts and the setup variables).
Monitoring the process is a matter of measuring the strain on the machine's frame while it is producing good parts and recording the strain's signature. During subsequent production runs, the monitor uses this recorded signature as a reference. The monitor compares
This forming process has four distinct steps: 1. The tool contacts the chamfered tube. 2. The tube forms to the tool and begins to buckle. 3. The buckled material forms to the tooling. 4. The tooling withdraws.These forming steps put a specific strain on the machines's frame (plotted at right). The gray line represents strain while the process was producing good parts; it serves as a reference. The current cycle is plotted in green. Because the green trace closely resembles the gray trace, the sensor signals PASS to indicate a good part.
the strain measured during production with the reference signature. If the current signature closely resembles the reference, the process is stable and the monitor indicates the part is good (see Figure 1). On the other hand, if the current signature differs from the reference signature, the system signals the operator that the process has changed (see Figure 2).
Plotting the strain on an end forming machine's frame while it is manufacturing good parts (gray line) provides a reference. Comparing it to the strain of a subsequent manufacturing cycle (shown here in red) reveals that something in the process has changed substantially. The part is indeed bad, but it may have passed a visual inspection.
As a reliable quality assurance tool, modern piezoelectric strain sensors coupled with advanced signal monitors is one method for monitoring end forming, riveting, and crimping.
Intelligent signature monitors can learn good process signatures, and then employ a number of standard or application-specific discrimination algorithms, to determine whether the process has varied from its known, capable condition. Standard detection algorithms include simple peak, area, or envelope methods. However, these standard algorithms are good only for coarse variation detection unless the tolerances are set very tight. The drawback with stringent tolerances is the possibility of falsely labeling good parts as rejects. In many cases, it is necessary to use proprietary, application-specific algorithms.
Every production cycle creates a specific strain signature for every unique part produced. In many cases, one properly positioned, high-sensitivity piezoelectric strain sensor is sufficient. Other processes might require a second signal delivered by a second sensor.
Process variation monitoring has five unique characteristics that can benefit nearly all fabricators.
First, it monitors 100 percent of the production run without adding time or processing steps. Second, the early detection of a defective part prevents further processing, which saves production time and reduces scrap costs. Third, maintenance personnel eliminate the downtime associated with performing maintenance too frequently or not frequently enough. With a process variation monitor in place, they do maintenance when the machine actually needs attention, and not according to a worst-case schedule. This reduces maintenance costs and increases operational efficiency. Fourth, manufacturers replace tooling only when it needs to be replaced—not too early, when it still has some life left in it, and not too late, after it starts to affect quality or efficiency negatively.
Fifth, and most important, all personnel—maintenance, production, and management—focus squarely on the process, rather than on parts, which facilitates sustainable gains in efficiency, scrap reduction, and quality.