Wireless Data Acquisition for Machinery in Motion: Improving PET Blow
Mold Bottle Production
Asa Wilson and Greig Latham of Keeva, L.C.
Category: Industrial Automation
Products Used: National Instruments LabWindows/CVI, Keeva, L.C. Wireless
Data Acquisition System
The Challenge: Obtain accurate, high-resolution manufacturing data from a
rotating bottle blow mold wheel. Move it into an advanced data analysis
and visualization environment where it can be used to reject substandard
products, optimize quality, anticipate maintenance, and support R&D. Provide an
inexpensive solution that is easy to install with minimal downtime.
The Solution: Use the Keeva, L.C. Data Acquisition Unit to collect
deterministic (± 5 ms), 1,000 sample/s data from sensors on the wheel bottle
forming stations. Send the data via wireless Ethernet to a LabWindows/CVI
workstation where it can be analyzed and displayed. Use a digital interface to
reject bad bottles.
As the Polyethylene Terephthalate (PET) bottle industry expands into new
markets, the increased demand makes it important to ensure continuous
high-quality production runs. New requirements in bottle size, shape,
durability, and shelf life, however, make these production runs more difficult
In the PET blow mold production cycle, a preheated test tube-shaped preform
enters a chamber where a descending plunger uses high-pressure air to rapidly
expand the thick plastic wall of the tube until it conforms to the mold.
For continuous rapid production, several bottle-forming stations are mounted
on the rim of a rotating wheel. As the wheel turns, each station moves through
the steps of picking up a preform, expanding the bottle, and discharge.
The physical characteristics of the resulting bottle are critically dependent on
the proper execution of the forming cycle. If malformed bottles are not
rejected, they could cause an entire truckload of bottles to be refused by the
customer, resulting in substantial costs. In order to monitor the forming cycle,
a high-performance data acquisition system must be mounted on the wheel that can
transfer detailed process measurements to an off-wheel analytical workstation.
As shown in Figure 1, the Keeva, L.C. Wheel Watcher system consists of two
computers communicating via TCP/IP over a wireless Ethernet link. The Data
Acquisition Unit (DAU) is a Pentium class embedded computer mounted on the wheel
and equipped with air pressure and rod displacement sensors placed at each
bottle forming station. The DAU makes high-speed deterministic readings from
these sensors and assembles them into a data frame that is sent to the Data
Management Unit (DMU) located off the wheel. The DMU is equipped according to
the processing load of the LabWindows/CVI application and in consideration of
customer preferences. For this application, a Pentium II 350MHz platform running
Windows NT was used.
Figure 1. Keeva, L.C. Wheel Watcher System topology
For this application the DAU was configured to sample both air pressure and
rod displacement for a 24-station wheel at a rate of 1 sample/sensor/ms. Since
each sample yields a two-byte data value, this requires a hefty (2 x 24 x 1000 x
2) » 100,000 bytes/s data transfer rate. In order to moderate the demand that
this rate places on the workstation and data storage, the DAU is equipped with
two features: 1) data frames are consolidated on a per-bottle basis to maximize
the communication overhead-to-payload ratio, and 2) each station follows a
'sampling profile' that gathers data at various resolutions during the cycle
according to downstream analytical requirements. We picked LabWindows/CVI
for developing the DMU software because of its maturity, its industry-proven
reliability, and its rich graphic development environment. These characteristics
served to streamline our efforts and allowed us to concentrate on the special
requirements of our product and the application. The modular architecture and
message-based strategy of LabWindows/CVI enabled us to develop components that
are both reusable and multipurpose in nature.
Once the data reaches the DMU, the advanced multithreading and
instrumentation driver features of LabWindows/CVI are used to distribute the
information according to a client/server strategy in support of the
visualization, quality control, and archival activities. Additional clients
could include a relational database, an SPC application, or other special
The following screen shots show some of the system features. All
presentations include an area on the right side (blue) that shows current
The 'cycle display' (Figure 2) allows the user to view six curves derived
from any station, signal, or mathematical transform. The air pressure (red) and
rod displacement (cyan) are displayed along with the first order derivative of
air pressure (green). The first and second order derivatives of the sensor data
are used to amplify variations in the signal.
Figure 2. Cycle Display
In the Process Display (Figure 3) the user can set 'warn' (yellow) and
'error' (red) limits for each station. When the signal goes outside of a limit,
a tally is kept and, if the error limit is exceeded, the product is rejected.
Each limit can be given a sliding 'detection window' that specifies both a span
and intensity setting when testing the signal.
Figure 3. Process Display
The Design of Experiment Display (Figure 4) provides a setting in which the
user can examine and analyze a specific signal. The LabWindows/CVI graphic
cursor tools are used to provide instantaneous and interval readings of signal
magnitude and angular position.
Figure 4: Design of Experiment Display
Two more displays (not shown) round out the system features. The Calibration
Display simplifies calibration of sensors by allowing the system to 'learn' the
maximum and minimum signal values from incoming data, which are then paired with
the corresponding engineering values. The Indices Display presents mathematical
coefficients that serve to characterize the quality of the forming cycle, such
as the timing and impact of air delivery, the smoothness of displacement rod
movement, and so forth.
The Bottom Line
The process solution in this application has three significant economic
· Dramatically lower installation costs – this system requires two days
to install, compared to ten days for other
solutions, yielding a 500 percent saving in downtime that can pay for the
· Significant operational savings – client estimates indicate the return on
investment to be about 180 days, representing a net present value of over
300 percent based on a 25 percent internal rate of return
· Capable and efficient network administration – the Ethernet network allows
distributed network maintenance expected by leading edge users
The economic benefits accrued to this system are a result of employing
advanced software such as LabWindows/CVI, high-speed hardware such as Pentium™
class embedded controllers and capable networking solutions such as Ethernet.
Moreover, and of quantum significance, these choices allow rapid deployment of
this core technology to a variety of industries.