Controlling diversified gum calenders using gamma backscatter thickness measurement.

Many of today's tire plant gum rubber calenders are operated in a very flexible manner. Some are large with multiple nips. On some, multiple strips are produced from each calender nip in various widths. In some cases, contoured profiles and multiple layer constructions are produced at the calender. As these calenders produce more specialized and complex products, their importance to the production of tires increases. Accurate measurement and control of the calendered thickness thus has become a key ingredient for successful tiremaking. While accommodating flexibility and providing accuracy, a system to measure and control the product of the gum calender must also be sufficiently inexpensive to justify its purchase. Our search for a way to satisfy all of the above requirements led us to decide to develop a system based on gamma backscatter measurement. The total development project included the design of this system and the initial applications of it.

Experimental

A measurement and control system utilizing stationary gamma backscatter sensors was developed for gum calenders. The configuration employed for beta testing is given in figure 1. There are two sensors measuring the left and right sections of the product that emerges from each of two roll gaps. Based on the inputs from all four sensors, the system controls the product thickness for each strip independently by driving four motors that adjust the roll gaps. The operator views the measurement information and inputs product parameters on a color, touch screen monitor.

Results and discussion

Type of measurement

The gamma backscatter technique, based on Compton photon backscattering principles, offered the advantages listed in table 1. Furthermore, the operating principles of a gamma backscatter sensor appeared to satisfy the requirements for accurate measurement.

As shown in figures 2a and 2b, the radioactive isotope, a small, sintered pellet contained in a sealed, stainless steel capsule which is, in turn, sealed in a tungsten collimator, emits 60 Kev photon. When these photons are directed at materials of low atomic number, some interact with the material, or scatter. the scattered photon, called a Compton photon, has a lower energy than that emitted by the source. The change in energy is a function of the angle of scattering with the maximum occuring at 180[degrees], or in the direction of returning to the sensor.

Surrounding the source assembly is a clean, sodium iodide (NaI) crystal, called a scintillation detector because it gives of flashes of light (scintillates) when struck by the backscattered photons, Thus, the crystal converts photons in the gamma region of the electromagnetic (EM) spectrum into photons in the light (visible) region on the EM spectrum. The visible photons are converted to electrons when they strike the photocathode of a photomultiplier tube (PMT).

The electrons are amplfied (caused to increase in number) by a chain of dynodes in process called "stimulated emission," producing a "shower" of electrons that makes an electrical pulse. The pulses are then counted and processed, producing an output for the sensor in direct engineering units such as mils or grams per square meter. The number of pulses per unit time (pulses per second) is proportional to the mass, or weight per unit area, of the material. The assumption that this technique would meet or exceed all accuracy specifications was tested both in the laboratory and on the production line.

Measurement testing

The laboratory calibration, over a typical range, is nicely represented by a typical, first order exponential curve, mass = f (natural log of counts), which is shown in figure 3.

When the range was extended to much heavier sheets, the fit of the calibration curve remained excellent. Figure 4 shows heavier sample tests with a correlation between sensor and lab weights to an r-squared greater than 99.9. This level of correlation has been confirmed on many different compounds from a number of different companies.

However, most plants prefer to have their online measurement expressed in thickness units. Since the conversion from weight to thickness involves density, the ability to associate a particular product with its nominal specific gravity was identified as an essential part of the system. This is accomplished in the system that was developed by including specific gravity in the product recipe data list. The recipe feature of the system is discussed later.

Online tests were conducted over a several-month period. Readings on a reference sample were checked weekly. After three months, the reference checks showed the sensors to be repeatable to 0.30% of the known value of the sample. Actual repeatability was probably much better but could not be confirmed due to the experimental errors inherent in the procedure that was followed.

To enable rapid compensation for any thickness upsets, the sensors are located as closely as possible to the nips. At the beta site, the distance from the nips is within three meters. The resulting transport delay between the nips and the sensors is therefore only four seconds at the normal calender running speed. At this location, the temperature of the face of the sensor was 135[degrees] F when the calender was running. To prevent eventual heat damage to the crystal, the sensors were enclosed in a water cooling jacket. The cooling reduced the temperature under the acceptable limit of 120[degrees]F. After one year of operation, there is no evidence of heat damage to the sensor.

It is known that a thickness change occurs between the calender and the windup. With the sensors located near the calender nips for fast control, it is necessary to have additional sensors at the windup for final product measurement unless the change is predictable. To test the hypothesis that windup measurement is unneccessary, a fifth sensor was temporarily installed to one of the windups. The data gathered from that test proved that the drawdown ratio varied by less than 1% of its mean value. On the basis for this test, the sensor at the windup was removed.

Operator usage

A number of operator interface design features were adopted to make the screen format simple and easy-to-use. The video monitor is a large high-resolution, color model.

Color coding is used to distinguish between labels, targets, data, control outputs, enterable and nonenterable values. Entries are made by breaking the plane of an infrared touch frame that surrounds the monitor. Making inputs by touching the display was determined to be more natural and less complicated than typing on a keyboard or selecting from a panel of dedicated pushbuttons.

With the "soft" touch buttons, a large number of options or an extensive database can be accessed quickly with just a

Table 1 - summary of gamma backscatter advantages

Advantage                          Reason
No concern with source             Half life of 455 years for
decay                              Am-241
Easy to install                    Single-sided, weight 1
                                   Kg
Low composition sensitivity        f(atomic number/atomic
                                   weight)
No need for offline                Not affected by dirt,
standardization                    temperature and not subject
                                   to drifting
Small amount of radiation          Only requires 150 mCi
                                   source

few buttons.

Only the required buttons are presented to the operator at any given time. For example, the numeric and alphabetic keypads ar displayed only when a value to be changed is touched. The display options are presented in a step-by-step manner that guides the operator. The split screen allows sensor data to be continuously viewed even while changing targets or modifying a recipe.

Sensor data is displayed both numerically and in graphic form. Figure 5 shows a display with these two forms of data presentation selected on the upper and lower screens. The simplicity of the operator interface enabled all operators, including those with no previous experience in computer system use, to quickly become proficient in interfacing with the system.

Process control

The thickness of each strip is controlled by feedback to the screwdown motor on that side of the calender for that roll nip. Moving one side of a roll up or down will affect the thickness of the rubber on the opposite side, as well as on the same side. Therefore, adjustments to a screwdown are decoupled for the other side. The control move is adjusted to account for the distance from the screwdown to the measurement point by solving the expression:

Change in left screwdown = left thickness error + (right thickness error - left thickness error)

X (Distance from left screwdown to left measurement/distance between measurements)

Changeovers can occur frequently and may require a large gap change. To minimize the time required to reach target after a product change, a control feature, called Delta Preset, was developed. When this feature is invoked, a control move is made to all gaps immediately after a new recipe is loaded. Larger moves are permitted than when running in a steady state condition. A gain multiplier is provided to allow a larger initial move than would be called for by the normal tuning. During the product-to-product transition, the material is automatically considered scrap until the readings are within specification.

Data capture

For analysis on a roll by roll basis, data is transferred to a host computer over a 9600 baud serial communication link. A host computer, such as a personal computer, can easily acess data from the calender system with a simple querying protocol.

Operators and management personnel employ the hard copy job report which summarizes a run of a particular combination of strips. X-bar and R charts are plotted for any measured variable. The charts contain a data section which includes X-bar bar, sigma and CPK.

Recipe management

Separate targets, specific gravity, control parameters, alarm limits, graphic display scaling and SPC parameters can be saved and recalled for 57 different products. For four roll calenders, a product can be defined with different thickness targets for the top and bottom gaps.

Although it will be required infrequently, a database management system was developed to extend the number of recipes available to the system to 10,000. Running on a personal computer, this system displays the recipe data in the same format as the process control system. Extensive editing and search capability is available. Recipes are transmitted to and received from the process control system via the series communication link.

Results

The system that has been discussed was installed last year on a relatively large, four roll gum calender. Installation, calibration and control tuning were all complete in one week. Data gathering trials were successfully conducted during the same week using a personal computer. A second installation in the spring of 1991 similarly started up quickly. The customer verified significant savings on that calender within two months.

The trend plots in figure 5 show the control effectiveness that was achieved at the beta site at the outset. To date, thickness variations have been significantly reduced from a three sigma of 7% to three sigma of 1%. The second installation showed similar improvement. Forty--five days after installation, the three sigma variation, had been reduced from 7.70 to 2.50.

Conclusion

The operation of a gum calender can be greatly improved by a thickness measurement and control system. The process visibility and the operator time made available by automatic control enables closer attention to the important aspects of the calender that are not controlled by the system. It has been shown that these benefits are available even to calenders which produce multiple strips simultaneously.