WO1997023550A1

WO1997023550A1 - Automatic control of the degree of cure of prepregs

Info

Publication number: WO1997023550A1
Application number: PCT/US1996/020307
Authority: WO
Inventors: John R. Peterson; Alan J. Luse; Homer A. Mills; Amy Lynn Moe
Original assignee: Alliedsignal Inc.
Priority date: 1995-12-22
Filing date: 1996-12-20
Publication date: 1997-07-03
Also published as: AU1428597A

Abstract

A method of controlling the degree of cure of prepregs of the type used to make laminates and printed circuit boards. The degree of cure is measured and adjustments are made to the heat supplied to an incompletely cured resin-coated web to cause the overall average cure and the cure in predetermined regions of the web to be within a specified range about a target value. In a preferred embodiment, a continuous web of resin-coated fabric (10) is continuously measured to determine the degree of cure across its width, the values are averaged and then compared with the target value having a known standard deviation. Adjustments are made to the heat supply (16) and to the web speed to correct the deviation of the degree of cure from the target value.

Description

AUTOMATIC CONTROL OF THE DEGREE OF CURE OF PREPREGS Background ofthe Invention This invention relates in general to the manufacture of prepregs, which are used in making laminates and printed circuit boards. Typically, these are glass fabrics which have been coated with a resin, usually an epoxy resin, and then partially cured.

U.S. Pat. Nos. 5,142,151 and 5,457,319 disclose methods of measuring the degree of cure of the resins and the amount present on the fabric. It is suggested that such measurements may be used to control the curing process. The present invention relates to methods for achieving that control.

Summary ofthe Invention The invention in one aspect is a method of controlling the degree of cure of prepregs during their manufacture. A continuous fabric web, typically fabric woven from glass yams, is coated with a curable resin varnish, typically an epoxy resin varnish, and then heated to drive off solvents and to cause the curing reactions to occur. The degree of cure is measured within each of a predetermined number of regions across the width of the web. The measurements of the degree of cure in each region are averaged and compared with a predetermined target value having a predetermined standard deviation. Depending upon the extent of the deviation found by this procedure, the heat supplied to each region ofthe web is adjusted to bring the average value to within one sigma (as estimated by the standard deviation) of the target value. The process is continuously adjusted with the result that the prepregs closely approach the targeted degree of cure with a rninimum variation.

In one aspect, the invention is a method for controlling the degree of cure of prepregs during their manufacture by first controlling the average degree of cure by adjusting the speed of the resin-coated web, followed by controlling the degree of cure of predetermined regions ofthe web by adjusting the heat input to such regions. In a preferred embodiment where the measurements of the degree of cure are made continuously across the width of the web with the Dynamic Cure

Monitor described in U.S. Pat. No. 5,457,319, and the heat is supplied to the regions by electrically heated radiant coils, adjustment ofthe radiant coils is made by the following hierarchy of rules and using an average value for each region.

Rule I: adjust the temperature of the radiant coils by 16°F (8.9°C) if any one ofthe measured average values for a region is greater than 3 sigma;

Rule LI: adjust the temperature ofthe radiant coils by 8°F (4.4°C) if 2 of 3 ofthe measured average values are on one side of the target value and are greater than 2 sigma and Rule I does not apply;

Rule LTI: adjust by 4°F (2.2°C) if 4 of 5 of the measured average values are on one side ofthe target value and are greater than 1 sigma and Rules I and π do not apply; and

Rule TV: adjust by 2°F (1 1°C) if 8 of 8 of the measured average values fall on one side ofthe target value and Rules I-LH do not apply.

The foregoing rules apply to control the degree of cure in the predetermined regions across the web. If the overall average ofthe web is above or below the target value, a change to the speed of the web is made following similar rules. In a preferred embodiment, Rule I: if any value is greater than 3 sigma, the web speed is changed by

0.8 ft/min (0.24 m min);

Rule LI: if 2 of 3 values are on one side of the target value and are greater than 2 sigma then the web speed is changed by 0.4 ft/min (0.122 m/min) if Rule I does not apply; Rule LTI: if 4 of 5 values are on one side of he target value and are greater than 1 sigma then the web speed is changed by 0.2 ft min (0.061 m/min) if Rules I and II do not apply;

Rule IV: if 8 of 8 values are on one side of the target value then the web speed is changed by 0.1 ft m (0.030 m/min) if Rules I, II, and m do not apply. As in Rules I-IV above, these rules follow a hierarchy such that the largest changes are made first and the smallest changes last. 97/23550 PC17US96/20307

3

Description ofthe Drawings

Figure 1 is a simplified diagram ofthe prepreg manufacturing process. Figure 2 illustrates the regions measured across a moving fabric web. Figure 3 shows a typical correction made to the degree of cure across a moving fabric web.

Detailed Description ofthe Preferred Embodiments Statistical Process Control of Prepreg Manufacture

In a typical process for manufacturing prepreg an epoxy resin formulation (an epoxy varnish) is coated onto a fabric made of woven glass yarns and then partially cured to form the "prepreg", which is used later in making printed circuit boards. The degree of cure is important to the user of the prepreg. A typical prepreg will be about 40% cured, although this may vary, depending upon the intended use. Although the users specification may allow a variation of say, plus or minus 2%, with the introduction of automated continuous measurement ofthe degree of cure, such as is discussed in U.S. Pat. Nos. 5,142,151 and 5,457,319, it is practical to implement feedback control ofthe prepregging process. If periodic sampling and measurement are carried out manually, the amount of data is usually too small and too late to allow close control ofthe degree of cure. While the general technique will be familiar to those skilled in the art, the successful application to the prepregging process depends on consideration of a number of factors. The general principles of statistical process control will be briefly discussed first in order to establish the underlying basis for the process as applied to making prepregs. When one thinks of product specifications it is often in the context of a target value having an acceptable range within which all of the product should fall, say 40% ± 2% in the present context. However, although this means that a product which is somewhere within the range of 38 to 42% is acceptable, the product should have an median value of 40% with little variation, not just be within the range of acceptable values because random variations can cause a significant amount ofthe product to be outside the specification range. A method 7/23550 PC17US96/20307

4

first proposed by W. Shewhart has been applied to keep a product within the desired range and at the same time to provide a more consistent produα.

If a series of measurements is made on a material undergoing a manufacturing process, the values obtained typically are not identical. They vary because the process is changing or because the measurements themselves are imprecise. Thus, when such a series of measurements is plotted to show the number of times a value is obtained, a bell-shaped curve is typically obtained, which should be centered about the target value for the measurement. The curve presents the probability that a particular value will be obtained. It can be shown for an ideal bell curve that 68.26% of all the values measured will fall within a range of ± 1 sigma around the median or peak value ofthe curve. In practice, the value of sigma is approximated by calculating a "standard deviation" from the values which have been measured. Within the range of ± 2 sigma, ideally 95.44 % ofthe values will be found. Within the range of ± 3 sigma ideally 99.72 % of all the values will be found. Thus, if a value is measured which is outside the range of ± 3 sigma it is likely that the process has varied, since it is unlikely that such a value would be measured if the distribution is normal and centered about the target value. Correction of the process may be indicated, particularly if further measurements continue to show such variation from the target value. In applying these principles, a standard deviation is calculated (hereinafter generally referred to as "sigma") from the data available and that sigma value is used as a measure of the performance of the process and the corrective action which is needed.

Another way of looking at statistical process control is that keeping the measured values with the desired range about the target value is not enough. Instead, the objective is to continuously measure the desired property, in this case the percent cure, and, if required, to change the process to cause the product to have a median value at the target value and all measured values within ± 3 sigma. Thus, when an average measured value is greater than 3 sigma a large correction is made in order to move toward the target value. When an average value is between 2 and 3 sigma, a smaller correction is made. When an average measured value is between than 1 and 2 sigma, then a still smaller correction is made. Finally, when an average value is within one sigma, the smallest correction would be made. This is termed "proportional control" of the process. The control process attempts to place certain average values within the standard deviation, one sigma, and thus, at the same time, reduces the individual variations and narrows the standard deviation.

Making Prepregs

In the typical manufacturing process as shown in Figure 1, a roll of woven glass fabric 10 is coated with an epoxy resin varnish and then passed through equipment in which the coated fabric is dried and cured so that it is dry to the touch, although the epoxy resin is not fully cured. Such coated fabrics are called "prepregs" and they are said to have been "B-staged". When the prepregs are fully cured in making laminates or printed circuit boards, they are said to be "C- staged". The equipment in which the resin varnish is coated onto the fabric and partially cured is commonly referred to as a "treater", generally designated 12. There are various types of equipment used for curing the coated fabrics. Such equipment is controlled according to the invention to provide a consistent degree of cure.

In the typical electrically heated treater the fabric 10 is passed through a pan 14 of epoxy resin vamish and then through metering rolls (not shown) to adjust the amount ofthe vamish. Then, the coated fabric is passed into a region in which heat is supplied by electrical elements 16 to remove solvents and to raise the temperature of the varnish so that the curing reactions take place. This process is rather complex. Heat may be supplied to the coated fabric (the "web") either by radiant energy or by convection. Depending upon the frequency of the radiant energy, the heating may occur in the vamish or in the glass fabric, or both. In contrast, in convection heating the transfer of energy takes place by convection from hot air to the web and by convection or conduction within the resin varnish and the fabric. Since the first step is to remove solvents which have been included to dissolve the components ofthe varnish and to adjust its viscosity, it is important to provide heat of vaporization while not overheating so that boiling of the solvent occurs. Further, it is important to remove the solvents so that flammable mixtures do not occur within the enclosed spaces of the treater. Consequently, the use of hot air as a heating means serves to both heat and to carry off the evaporated solvents. If radiant heating is used, it is necessary to provide a flow of gas (not shown), usually air, to remove the evaporated solvents. In either case, after the solvents have been removed, the temperature of the web will be increased so that the curing process proceeds.

Many treaters are designed so that the heating occurs while the web is moving in a vertical direction. Some others carry the coated web horizontally, but most commercial treaters are of the vertical design. In general, the solvents are removed while the web moves upward and the major portion of the curing occurs while the web moves downward again. Typically, heat is provided separately for the upward and downward passes. Ideally, the heat would be supplied and controlled independently to each of a large number of relatively small sections of the web so that solvent evaporation and curing can be closely controlled. As a practical matter, compromises must be made. However, if heat is supplied by electrical radiant coils 16, independent adjustment of the heat input supplied to the many zones is relatively simple. This flexibility is not as readily available in treaters which supply radiant heat from panels heated by hot oil or other heat transfer fluids, or in treaters where the heat is supplied by hot air. Consequently, treaters which are heated electrically are more readily adapted to the methods ofthe invention, although treaters heated by hot oil or by hot air may also be controlled, even though the details ofthe procedure may differ.

In the preferred embodiment, measurements are continually being made of the degree of cure and of the amount of resin on the web at the outlet of the treater, using the Dynamic Cure Monitor 18, as described in U.S. Pat. No. 5,457,319. Alternatively, measurements may be made using FTLR equipment as described in U.S. Pat. No. 5,142,151. It will be appreciated that since the web is continually moving, typically at about 3 to 35 ft/sec (0.92 - 11.6 m/sec). If measurements are taken sequentially by a sensor moving at constant velocity back and forth across the web, the readings will represent values along a zig-zag shaped path 13 along the web 12, as illustrated in Figure 2. If measurements are taken frequently, there may be upwards of 100 taken during each traverse of the web. These measurements are averaged and the heat supplied to the web is adjusted to control the degree of cure.

It has been found by continuously monitoring of the degree of cure that variations occur across the web. These may result from a natural scatter of the measurements, which are affected by the local conditions, for example the relative amounts of glass fibers and the resin at a particular spot but, even after applying data smoothing algorithms, such as multi-point moving averaging, to minimize the noise in the data, variation is still observed, which is attributed to varying states of cure and resin content occurring across and along the web. The degree of cure will be affected by the amount of heat received at a particular location and the heat needed to evaporate the solvent and to raise the temperature ofthe resin and the glass fabric. Adjustments to the heat supply must be made since it is not possible to change the amount of resin or the amount of solvent at particular locations on the web. In practice, the width of the moving web is divided arbitrarily into regions which will be controlled as illustrated in Figure 2. The data taken by the cure monitoring equipment is averaged over a predetermined length ofthe web and the heat input to each region is corrected as required by the data.

In a typical example, a 50 inch wide web (1265 mm) is divided into nine regions, 12A-L, a 2-inch (50.8 mm) wide strip at each edge (A and I) and seven uriiforrnly wide regions (-6.58" or 166 mm) (B-H). Within each region about thirteen measurements will be made in each pass of the cure monitoring equipment across the web and an average value computed (represented by circles 12 A-I). Three passes are averaged for each region (e.g. the three values within the oval at 12C) to provide a value which is compared with the target value for the degree of cure. Adjustments are made using a hierarchy of rules based on the extent of the deviation measured. If the average value of the degree of cure is greater than 3 sigma (3 times the standard deviation), then the largest correction is needed. If 2 of 3 ofthe average values ofthe degree of cure are on one side of the target value and are greater than 2 sigma (2 times the standard deviation) from the target value, then a large correction is needed, provided that the previous rule does not apply. If 4 of 5 of the average values ofthe degree of cure are on one side of the target value and are found to be greater than 1 sigma, a smaller correction is needed, provided that the previous two rules do not apply. If 8 of 8 of the average values ofthe degree of cure are on one side of the target value, again, provided that the previous rules do not apply, the smallest correction is needed.

Deviations from the targeted degree of cure can also be corrected generally by adjustment of the speed of the web, thus effectively increasing or decreasing the average heat input. For this purpose, the overall average of the sections B-H may be used. Changes to the web speed will follow rules similar to those used to adjust the heat input to other individual regions. In fact, in the preferred embodiment such corrections are made before any fine (lining of the heat input to the web are made. Thereafter, the heat supplied to each predetermined region ofthe web is adjusted, depending on the deviation from the target value. In the example being described, the temperature of electrical heaters serving each region is adjusted in response to data received from the cure monitor.

If the deviation is due to a change in the amount of solvent or some mechanical problem with the metering rolls or the treater such problems must be dealt with by changes other than those made to the heat input or web speed according to the invention.

One preferred application where the heat is supplied by electrically heated radiant coils follows these rules: take an average of three successive values of, for example, 12C and use that new average (shown by the oval in Figure 2) as one data point. Rule I: adjust the temperature of the radiant coils by 16°F (8.9°C) if any data point is greater than 3 sigma from the target value;

Rule II: adjust the temperature ofthe radiant coils by 8°F (4.4°C) if 2 of 3 data points are on one side of the target value and are greater than 2 sigma and Rule I does not apply; Rule EH: adjust by 4°F (2.2°C) if 4 of 5 data points are on one side ofthe target value and are greater than 1 sigma and Rules I and II do not apply; and Rule IV: adjust by 2°F (1.1°C) if 8 of 8 data points are on one side the target value and Rules I-OI do not apply.

It will be appreciated that such rules apply to a particular type of electrically heated treater. The amount of adjustment may differ if the design of the treater differs. Other rules will be needed if a different type of heat is supplied. The general principles will apply. That is, the correction will be proportional to the deviation of the measured value from the target value as compared with the standard deviation.

In a similar way, the web speed may be adjusted following similar rules: take an overall average of regions B-H for three scans ofthe web.

Rule I: adjust the speed by 0.8 ft/min (0.24 m/min) if any data point is beyond 3 sigma;

Rule II: adjust the speed by 0.4 ft/min (0.122 m/min) if 2 of 3 data points are on one side of the target value and are beyond 2 sigma and Rule I does not apply;

Rule ELI: adjust the speed by 0.2 ft/min (0.061 m/min) if 4 of 5 data points are on one side ofthe target value and are beyond 1 sigma and Rules I and LT do not apply;

Rule IV: adjust the speed by 0.1 fVrnin (0.030 m/min) if 8 of 8 data points are on one side of the target value and none of the previous rules are being violated.

The heat input and web speed preferably would not be changed simultaneously. In practice, it has been found that the web speed should be adjusted first since it is based on an overall average value for the degree of cure. Once the overall average is within control, the individual regions A-I can be adjusted to that the degree of cure is uniform across the web. Typically, the web speed will be adjusted and then a certain time delay must occur before the web speed can be adjusted again. In the meantime, heat input to the individual regions can be adjusted. PC17U 96/20307

10

Example 1 Figure 3 illustrates a typical correction made in an electrically heated treater using the method described above on a 45 inch (1140 mm) wide web. The range of the desired cure is shown by dashed lines H and L. It is clear that two regions at the left side of the diagram have exceeded the desired degree of cure while the central region of the web is approaching an under-cured condition as shown by the curve labeled "before". A correction was made following the rules described above with the results that the curve labeled "after" shows that the region which had been over-cured is within the desired range and the central region has increased the degree of cure so that the web is more uniformly cured.

Claims

CLAIMS We claim

1. A method of controlling the degree of cure of prepregs during their manufacture comprising (a) coating a continuously moving fabric web with a curable resin vamish;

(b) heating said coated fabric of (a) to remove solvents in the varnish and to partially cure the resins;

(c) measuring the degree of cure of the partially cured resin in each of a predetermined number of regions across the width of said web; (d) deterrrining the deviation of the degree of cure from (c) with respect to a predetermined target value having a predetermined standard deviation (1 sigma); and

(e) adjusting the heat supplied to each region of said web in proportion to the deviation of the degree of cure measured in (c) as determined in (d) with respect to the predetermined standard deviation.

2. The method of Claim 1 wherein said coated fabric web in (b) is heated while said web is moving in a vertical direction.

3. The method of Claim 2 wherein said coated fabric web in (b) is heated by electrically heated radiant coils disposed to heat each of said regions of said web.

4. The method of Claim 3 wherein the temperature of said radiant coils for each region is adjusted according to the following hierarchy of rules:

Rule I. adjust by 16°F (8.9°C) if the deviation ofthe average value of the degree of cure is greater than 3 sigma; Rule II: adjust by 8°F (4.4°C) if the deviation of the average value of the degree of cure is on one side of the target value and greater than 2 sigma and Rule I does not apply;

Rule III: adjust by 4°F (2.2°C) if the deviation ofthe average value ofthe degree of cure is on one side of the target value and greater than 1 sigma and Rules I and II do not apply;

Rule IV: adjust by 2°F (1.1°C) if all ofthe average values of the degree of cure are on one side ofthe target value and Rules L, π, and UJ do not apply.

5. The method of Claim 4 wherein the temperature of said radiant coils of each region are adjusted according to the following hierarchy of rules:

Rule I: adjust by 16°F (8 9°C) when any average value is greater than 3 sigma; Rule LI: adjust by 8°F (4 4°C) when 2 of 3 average values are on one side ofthe target value and are greater than 2 sigma and Rule I does not apply;

Rule HI: adjust by 4°F (2.2°C) when 4 of 5 average values are on one side ofthe target value and are greater than 1 sigma and Rules I and II do not apply; Rule IV: adjust by 2°F (1.1°C) when 8 of 8 average values are on one side ofthe target value and Rules L, LI, and LTI do not apply.

6. A method of controlling the average degree of cure of prepregs during their manufacture comprising

(a) coating a continuously moving fabric web with a curable resm vamish; (b) heating said coated fabric of (a) to remove solvents in the vamish and to partially cure the resins;

(c) measuring the degree of cure ofthe partially cured resin across the width ofthe web and calculating an average degree of cure;

(d) determining the deviation of the average degree of cure measured in (c) with respect to a predetermined target value having a predetermined standard deviation (1 sigma);

(e) adjusting the speed of said web in proportion to the deviation of the average degree of cure as measured in (c) as determined in (d) with respect to the predetermined standard deviation.

7. The method of Claim 6 wherein the speed of said web is adjusted according to the following hierarchy of rules:

Rule I: adjust by 0.8 ft min (0.24 m/min) if the deviation of the average degree of cure is greater than 3 sigma;

Rule T: adjust by 0 4 ft/min (0.122 m/min) if the deviation of the average degree of cure on one side of the target value and is greater than 2 sigma and Rule I does not apply; Rule UJ: adjust by 0.2 ft/min (0.061 m/min) if the deviation ofthe average degree of cure on one side of the target value and is greater than 1 sigma and Rules I and II do not apply;

Rule IV: adjust by 0.1 ft/min (0.030 m/min) if the deviation of all values of the average degree of cure are on one side of the target value and Rules I, LI, and III do not apply.

8. The method of Claim 7 wherein the speed of said web is adjusted according the following hierarchy of rules:

Rule I: adjust by 0.8 ft/min (0.24 m/min) when any average value is greater than 3 sigma;

Rule LI: adjust by 0.4 ft min (0.122 m/min) when 2 of 3 average values are on one side of the target value and are greater than 2 sigma and Rule I does not apply;

Rule HI: adjust by 0.2 ft/min (0.061 m/min) when 4 of 5 average values are on one side ofthe target value and are greater than 1 sigma and Rules I and II do not apply;

Rule TV: adjust by 0.1 ft/min (0.030 m/min) when 8 of 8 average values are on one side ofthe target value and Rules I, LI, and LTJ do not apply.

9. A method of controlling the degree of cure of prepregs during their manufacture by supplying heat to cure a resin varnish on a continuously moving fabric web comprising

(a) controlling the average degree of cure of said resin by adjusting the speed of said web; then

(b) controlling the degree of cure of predetermined regions of said web by adjusting the heat input to said regions; and

(c) repeating steps (a) and (b) to maintain the degree of cure within predetermined target values.