The case for quality and for continuous quality improvement in manufacturing is very well understood. Achieving it, however, is still a challenge for many manufacturers.
The advent of lean manufacturing has eliminated many positions once responsible for quality control, or has left just one individual to monitor an entire manufacturing line or plant. At the same time, maximizing yield—getting the most high-quality product off a line—is critical. In an era where price-conscious consumers can choose from an array of global manufacturers, dependable quality can be a potent competitive differentiator. And if a defective product does make it to market, recalls can be expensive and damaging to a company’s brand.
What tools are available to manufacturers to help them achieve their quality goals? One answer is manufacturing intelligence, or MI, a combination of deploying in-process testing with the analysis of data collected during testing to achieve several key quality objectives.
Properly done, in-process testing refers to the deployment of sensors at as many points along the manufacturing process as possible so that a complete view of that manufacturing process can be gained. It’s less important how those data are collected than the extent to which they are collected.
Data must be collected widely enough; that is, across the entire manufacturing process, including the whole manufacturing line, feeder lines, assembly, sub-assembly, machining and so on. And data must be collected deeply enough; that is, with sufficient sensitivity and precision so the entire process is depicted. Figure 1 is an example of this.
To view diagram larger, click diagram.
Figure. 1: This signature represents a typical vacuum test. Most systems would use the only measurement at the end point (A) to determine whether the part passes or fails. However, this type of analysis would pass subtle anomalies (B) which clearly do not match the repeatable pattern of “good” parts.
Using Process Signature Verification
Some test equipment now on the market collect sufficient data to allow what is known as a process signature verification, or PSV, to be developed for every manufacturing process. Whether they use PSV or not, most manufacturers are collecting data of some sort. The key is to collect enough data of the right type, and then to store it properly and make it available for intelligent analysis. This second half of the manufacturing intelligence equation, analyzing the data and acting on the intelligence gleaned from doing so, is where the real returns on quality investments are made.
In its most fundamental role, manufacturing intelligence can be deployed to find and fix the root causes of manufacturing defects in the plant. Quality problems often seem to occur randomly during production, and can be caused by many different factors, including problems with raw materials, faulty parts, machinery that falls out of alignment or wears down, process errors, human error or environmental factors such as humidity or temperature.
Any of these can mysteriously cause a defect; manufacturing intelligence solves the mystery. Because data have been collected depicting the entire production process, analysis of those data can rapidly identify exactly where the defect entered the process. Was it a new batch of raw material or a new supplier? Were all the defects produced on just one of a parallel set of lines or by just one of two or more machines, or during just one of two or three shifts through the day? Was there a new operator? Did the environmental conditions in the plant change in some meaningful way?
Comparing the process verification signature of the defective part to that of quality parts will rapidly identify which process in the manufacturing chain was responsible and the right fix can then be made.
Manufacturing intelligence can also help determine exactly what constitutes a right part. This may sound obvious, but it can mean more than simply a part that passes or fails a series of tests. Rather, a right part is one that fits as tightly as possible within a set of test parameters, and the proper analysis of in-process test data such as process signature verification helps establish exactly where those parameters should be set. Set them too tightly and false rejects occur, with implications for manufacturing yield, wasted materials and, potentially, unnecessary retooling of the manufacturing line or process. Set the parameters too widely and defects get through. (Figures 2 and 3.)
Figure 2: Limits before changes – The red line shows the limit set for a particular process. Each data point on the graph represents parts that have been tested along the line.
To view diagram larger, click diagram.
Figure 3: Limits are tightened – The graph shows the impact of tightening the limits using the historical test data. Some of the parts that previously passed the test, and may have caused issues downstream in the assembly process, would fail with the new limits.
Once the optimal parameters defining a right part have been established, they can be replicated across a plant, or used to characterize and debug a new plant.
In a zero-tolerance manufacturing environment such as is found in many parts of the medical device industry, manufacturing intelligence not only supports the achievement of a zero-defect product, it provides objective, verifiable evidence that regulatory and manufacturing quality standards have been met.
The capture and storage of process verification signatures or some other set of comprehensive in-process test data establishes an audit trail of every step in the manufacturing of a pacemaker or some other high-risk device. This can help a manufacturer meet regulatory requirements, guarantee the quality of parts shipped to a customer and mitigate liability exposure in case of a device failure.
The quality imperative does not necessarily diminish once parts have been shipped. Nor does the power of manufacturing intelligence to continue to drive returns lessen. Indeed, for some manufacturers, the real returns from their investments in manufacturing quality will be realized only in the case of a product failure triggering a possible recall.
Despite the best efforts of quality managers and the deployment of the best manufacturing intelligence systems available, unexpected defects can still occur. When a customer uncovers such a defect, entire shipments can be quarantined, massive product recalls can be triggered and reputations and revenue can be threatened.
Manufacturers with manufacturing intelligence systems at their disposal can respond more quickly and effectively to such events. The collection and storage of complete in-process test data means such manufacturers can swiftly establish whether the parts were damaged in shipment or through the customer’s own processes, or whether they are genuinely defective products for which their plants were responsible. And if it is a genuine defect, is it a universal defect or can it be isolated to a single machine, a single shift or some other limiting factor?
Besides the obvious economic benefit of being able to reduce the scope of a product recall, often by dramatic proportions, the swift and certain response by a manufacturer to a customer’s quality concerns can only enhance the valuable relationship between them.
Manufacturing intelligence can be a potent tool in the ongoing challenge to achieve and continuously improve upon quality. It relies in the first instance on collecting and storing enough of the right kind of data to provide clear insight into every aspect of the manufacturing process.
The intelligence contained in those data is then unleashed by the proper analytical tools, yielding compelling returns as manufacturers can now properly monitor and analyze their processes, find and fix the root causes of defects, define what constitutes a right part and replicate that standard across their entire manufacturing environment, comply with regulatory requirements in a zero-defect industry, and contain the fallout from a quality spill.