Imagine running a race and only looking at the ground in front of your feet. You may not trip, but you also may not see the snarling dog at the end of the block. To reach the finish line in one piece, you need to focus on the big picture as well as what's right in front of you.

Western Digital Develops One Solution for Online and Offline SPC (below)

That's exactly the situation with statistical process control (SPC) and statistical quality control (SQC). In the quest for better yields and higher quality, manufacturers must consider these statistical techniques as standard tools for understanding, validating, and improving processes in all areas of manufacturing as well as health care, distribution, transportation, and accounting. Using online software, SPC gives you the close-up view. Using offline software, SQC shows you the big picture.

Most people are unaware that these software packages hold two different, but necessary, functions. Many opt for only one package and, consequently, compromise their company's success. To get where you're going in terms of improving quality and increasing profits, you need to look at both views.

Why the Confusion?

Before the widespread implementation of supervisory control and data acquisition (SCADA) and human-machine interface (HMI) systems, most SPC and SQC was performed by quality-control departments as an offline process. Data was collected from test stations, laboratories, and so on, and statistical analysis was performed later.

SCADA/HMI systems, however, have made it feasible to provide plant-floor SPC charts using data collected in real time directly from the process. As different departments (e.g., quality and production) now handle the offline and online applications of SQC and SPC with different priorities, the charts and their applications are often not comparable.

Companies that want to standardize SQC and SPC and increase their use find they need two vital functions:

• Provide the plant floor with SPC charts
• Make data collected by SCADA systems available for offline analysis

Today, online SPC and offline SQC software are available to support these efforts, but companies believe they only need to provide basic SPC charts to the plant-floor HMI displays; thus, they opt to purchase only the online SPC software.

As mentioned previously, choosing one over the other is a mistake, because the bulk of SPC's value is derived from process improvements driven by offline analysis. Online SPC offers substantial benefits, but its limitations render it inadequate for understanding the overall process.

An SPC Close-Up

Online SPC software excels at monitoring production processes in real time to give you a close-up view of what's currently going on. It uses parameters that have a direct, understandable effect on the process.

The classic use of online software is to keep plant-floor operators from overadjusting their machines. It also helps them see when a significant shift in the process happens and whether it requires action.

A typical example is where product samples are weighed on an electronic scale that sends the information directly to the online software. Because it's in real time, the online software automatically displays only the current data in control charts for the operators and their supervisors to analyze.

Online software usually provides an interface that controls exactly what is presented to the user. It lets you choose the point in the process where you're putting the monitoring station, configure the user interface with a tailored screen, and preselect the charts you want to display.

Online software requires little or no input from the operator, making training and use easy. To some extent, it prevents operators from influencing the data to skew results. Some companies, for example, have reported incidents where operators repeatedly weigh perfect samples to disguise a problem. In this case, there is a management-not software-issue at stake. Online software isn't designed to keep bad operators from doing wrong things. It is simply designed to aid the operator in controlling the process.

What online software can't do is show you the big picture. It won't let you see changes over time, look at a complete process, or compare multiple production lines. And it's often hard to incorporate offline analysis, such as from laboratory information management systems (LIMSs), raw materials, final product checks, and so on. It's also especially difficult to analyze defect data due to the difficulty in forming meaningful samples large enough to create usable control charts.

Another weakness is that sampling and subgrouping of data points is often inflexible or hard to change. This is a serious issue with SCADA systems because the systems frequently collect too much data to make it usable for SPC charting. For these tasks, you need offline SQC software.

The Big Picture

Offline SQC software is a decision-support tool that lets you analyze process behavior. It lets you answer all kinds of questions about your process. Is the process stable and predictable day in and day out? From supplier to supplier? From shift to shift? Could you increase yields? Lower production costs? Can you improve your process in ways that have an impact on your bottom line? What was the impact of previous process improvements?

Unlike online software, offline typically has a spreadsheet interface. You own the data and can manipulate it in any way you want. You can reorganize the data file, key in additional data, calculate new variables from existing data, and change how charts are configured.

Offline SQC is essential to detecting differences over time such as shift changes, day-of-week problems, and differences between suppliers. The importance of this broad time horizon cannot be overstated. Many companies have used offline SQC to detect problems related to seasonal weather and even an individual's vacation schedule.

In addition to allowing comparisons among lines, operators, shifts, and suppliers, offline SQC lets you more easily combine plant-floor data with results from test stations and the laboratory. For example, most data may be collected from online sensors, but some results are determined in the laboratory.

Once your process is in control (i.e., statistically predictable), offline SQC helps you make long-term process improvements. It can uncover relationships, monitor the results of process changes, and provide various other decision-support functions (e.g., it can determine whether a process can deliver product within specifications). This is where the major bottom-line benefits are achieved.

The Data Challenge

The biggest challenge with offline software is data availability, especially from SCADA systems. Because these systems collect huge amounts of data, they provide challenges and opportunities for improving the process.

The challenges often relate to having too much data. SCADA systems can collect at a minimum one data point per second and sometimes much more. When you're measuring data so close together, you can't see how the process flows. Often you can't get a statistically valid sample, and the control charts won't work. In addition, these systems often store data in a variety of forms, which makes accessing it difficult.

The best opportunities lie with the new SCADA systems that store data in a central process information database. These systems are gold mines for offline SQC, which lets you access these "vaults" of process data and mine them for information leading to process improvements. However, in light of the too-much-data problem, your success in accessing and mining the data lies in whether the offline software will let you filter for valid, representational data and then specify meaningful subgroupings of this data.

The Best of Both Views

So, offline SQC belongs to the quality department, and online belongs to the process people, right? Wrong. For many of the reasons described above, process engineers should also look at the data offline to see the big picture. To achieve SQC's fullest potential, these engineers should analyze data in the same way as does the quality department.

As SCADA/HMI systems proliferate and generate huge amounts of data, it's the process engineers and the people who design these systems that are in charge of it. Their view of SPC is a process view, which to them naturally means real-time control. They don't necessarily see how—with offline's big-picture view—the data gets used effectively.

The answer is for everyone (especially those close to the process and those involved with decisions about SCADA systems) to understand the strengths and weaknesses of both online and offline software.

Together, they are a powerful team that can help you cut costs, reduce rework and scrap, increase yields, reduce process variation, meet tighter specifications, and improve the bottom line. That's a race everyone can win.

Author Information

Jeffery L. Cawley is vice president of Northwest Analytical, Inc. (NWA), which provides analytical software for understanding processes and improving quality. NWA's suite of quality management software includes NWA Quality Analyst and NWA Quality Monitor. Cawley holds B.S. degrees in chemistry and geology from Juniata College in Huntington, Pa., and did graduate work in geochemistry at Princeton University.

Western Digital Develops One Solution for Online and Offline SPC Although you can't use online statistical process control (SPC) software for offline analysis, sometimes you can use a flexible offline software package for both. Western Digital, a thin-film disk manufacturing facility in Santa Clara, Calif., chose Northwest Analytical's Quality Analyst offline SPC software and combined it with custom programming (created with Microsoft's Visual Basic) to serve both needs. The system has an online function that immediately alerts plant-floor operators if a process goes out of control. It also has an offline function that lets process engineers and managers mine the SPC data for continuous process improvements. For online use, each workstation has a stand-alone module (SAM) that automatically collects process data from test devices. The operator then enters an alphanumeric identifier code that describes the demographics of the sample (i.e., manufacturing system, product type, and so on). The identifier code is transmitted with the test data to the SPC software, which allocates the data to the correct data file and generates the appropriate control charts. If the process goes out of control, the system alerts the operator and records it in an event log. The system then provides an online response action plan sheet, which gives the operator a step-by-step script for reacting to the out-of-control event. For offline use, the SAMs are networked with a remote server. An internal timer in Visual Basic code triggers transfer of all local data files to the external server once every hour. Engineers and product managers use the SPC software to access and analyze the data on the external server for in-depth process analysis. With access to the full range of data (which, at worst, is an hour old), they can quickly identify problems and make improvements to reduce process variation. For Western Digital, this system provides a cost-effective SPC solution that serves both the line operators and those who need to see the big picture for process improvement.

back to top