An interesting manufacturing tool is emerging as more companies seek to mine, or take advantage, of the process data their production systems generate. The tool, called Process Monitor (PM), uses statistical analysis methods to crunch the manufacturing process data, and predict whether or not the production system is operating within preset control limits that indicate an acceptable product outcome.
A potential Process Monitor user may have available existing data gathered from various points or sensors in a manufacturing line, usually due to local control of individual process subsystems. The question that must be answered is:
Will this data, when input into the PM and processed through the statistical analysis that the PM provides, yield meaningful output that can be utilized to improve the manufacturing outcome?
Here are some tips and recommendations about things you should know if you are considering a PM system. Start by asking yourself these questions:
How complex is your process?
Is your production yield high, and are your individual process steps fairly robust?
It will take an investment of capital, time, and internal resources to implement a PM tool, which would need to be funded by the value of recovered waste, production time, or increased sales/income due to saleable quality improvements. Your investigation should start with what your forecasted gain will be worth, which will form the justification for the investment in PM.
Another thought here is that you can first begin by manually gathering process data, and control chart that data visually on the shop floor, to determine where your process variability occurs, and then what corrective actions may need to be taken to get your manufacturing steps under control. You may be able to make needed process adjustments and reduce your variability (and the resulting waste) without the help of a higher level, powerful database that the PM tool utilizes.
More importantly, this step will help you eliminate non-essential process steps and clarify which process parameters are affecting your product outcome, in the event that you do decide to move forward with a PM system implementation.
Experiencing either high waste, or poor overall equipment efficiency and/or low production line effectiveness?
So, you know that there would be significant payback if you could reduce variability and improve product yields and/or machine uptime. Further, you have manually gathered and control charted what you assume are your salient process data, and tried several process adjustments … but you are still experiencing variability and waste.
The next step that should occur is to establish some root cause and effect relationships. For example, if your production is chemical batch in nature, it will be important to understand which process parameters (heat cycle, reactant addition, mixing, etc.) are the most significant influencers of product outcome. Are these key parameters being measured and logged real time? Are their interdependencies understood? If your production line is discrete assembly in nature, your key process parameters may be speed, acceleration, force, position, etc. and the interdependencies will likely involve time and geometry/space.
Where I am going with the discussion about cause and effect when it comes to defects or interruptions (variability) is to establish a connection between a particular process parameter (as noted above, and what we would consider to be the cause) and its ability to influence (usually negatively, in the form of variability) your product quality, production rate, etc. (what we would consider the effect).
Once you have established the specific process conditions that have been shown to be able to alter your production outcome, you are now better prepared to take advantage of what the PM is designed to do…monitor these parameters through appropriate instrumentation and data collection equipment, and input the date into the PM for statistical analysis.
For the PM tool to be effective, we first need to feed it the pertinent process data. The assessment and determination of the key process parameters, and the instrumentation to measure them, are provided by the manufacturer, not by the PM system.
You need to be able to explain how your product is built, and how you measure a successful build, so you can provide useful and germane information for the PM to operate on.
With the implementation of PM, those identified parameters are now tracked real time, and alarms are sent when process control limits are violated. Additionally, the SQC capability of the tool allows the user to understand where within the control limits the process is running, and how much of the range is being used (which is directly related to process capability, or CpK). The output from the PM is useful here as it allows the user to fine tune his process to optimize performance and yields.
Where PM becomes more valuable is when we have a complex manufacturing system with overlapping process parameters (meaning process manufacturing conditions are changing concurrently) and it is difficult for us from an observatory standpoint to determine which variable caused the defect.
If this type of system is properly instrumented, and the data is streamed real time and analyzed, each parameter (channel) can be statistically compared to all other events for that channel, and a projection can be made for which process condition(s) have wandered furthest from the calculated running mean. This information is extremely valuable when troubleshooting the process to make the needed adjustments required for waste elimination.
Another significant function that a PM system can provide is to be able to overlay the data stream from multiple channels in real time. This overlay creates a multiple dimensional picture of the predefined window in which the product is supposed to be (or is expected to be) manufacturable. Through creating this multidimensional picture of repeated events, the domain of failed product runs can be compared to the domain of acceptable product.
When the cause of the failed product is the result of interactions of two or more process parameters, this may be the only way to debug the problem, and make the necessary process adjustments, to improve the outcome.
One additional benefit of implementing a PM system is that over time the statistics that are gathered for the parameters being measured and tracked will yield data that allows control limits to be more centered, which will also serve to reduce variability and increase yields. Many times the product design, or the manufacturing process design, assume perfect (or typical) conditions as inputs to their constructs.
The measured performance of the various critical process points as input to the PM system are more realistic, as they take into account these unforeseen factors, losses, etc. Thus, expected control limits driven by either product or process design can be properly adjusted to allow for these other influencers that occur naturally in all manufacturing systems.
Thus, we see that with a methodical approach and understanding of where the causes of variability are originating, a PM database and statistical analysis software program can be an effective aid in troubleshooting and better controlling those process parameters that have the most influence on product outcome. This ultimately will result in higher product yields and a more robust manufacturing capability.