Automating and optimizing autoclave cure

Autoclave processing remains the backbone of advanced composite structure production. One of its primary goals is to fully cure a prepreg’s thermoset polymer matrix by initiating and sustaining chemical reactions that reduce its viscosity from B-stage (semi-solid) to liquid, and then increase it through gel to final vitrification (typically measured in terms of modulus of elasticity). Historically, these changes have not been “visible” during the cure cycle. Manufacturers typically perform a large array of preproduction tests to outline this complex change in viscoelastic properties, which occurs over time and as temperature increases, simplifying it into ramp rates, hold temperatures and dwell or “soak” durations.

In the autoclave, the part is assumed to have reached viscoelastic goals (full cure) when these secondary time/temperature goals are achieved. In consequence, safety margins must be built into time/temperature calculations to ensure full cure, and the process must be tightly controlled. Conventional autoclave control systems, therefore, are hardwired to the equipment and operated by a technician, who must monitor data readouts throughout each cure cycle (anywhere from 3 to 12 hours). Although the technician can manually adjust temperature, time and pressure controls to keep process variables within specified parameters, lagging thermocouples, high thermal mass tooling, out of boundary temperature and/or pressure conditions and other issues can arise, calling into question whether parts have, indeed, reached the viscoelastic threshold. Such conditions, if severe enough, can be cause for stopping a cycle and potentially scrapping a very expensive part. Because multiple parts often are cured in a single cycle to save time and amortize cost, process anomalies pose very significant risks.

Faced with this risk reality four years ago, Helicomb International (Tulsa, Okla.), a world-renowned, FAA-certified military and commercial helicopter repair facility, sought a better solution. “We started looking at autoclave process control software because it was becoming very difficult to fully meet all of our requirements with old technology,” says Brady Stephens, manager of Helicomb’s Manufacturing Div., which supplies autoclave-cured composite and metal-bonded structures to Boeing, Lockheed Martin, Northrop Grumman, GE Aviation Engines and other major aerospace OEMs. “We were not only getting ready for Nadcap accreditation but also preparing for an audit by Boeing to achieve D-16925 approval and BAC 5555/5514 certifications,” he adds, “and we knew we simply could not achieve these without new controls.”

Visualizing viscoelastic feedback

Stephens found his solution, unexpectedly, while attending a “Manufacturing Practices” class at Abaris Training Resources Inc. (Reno, Nev.). The instructor, Lou Dorworth, was using CSS300 process control software developed by AvPro Inc. (Norman, Okla.) to run his autoclave. During the class, Dorworth pointed out that traditional “time-temperature” cure cycles leave processors blind. “We are currently locked into legacy specifications,” Dorworth explained, “and, thus, the actual process is trapped in a box. If we can take a look at what we are trying to do in that box and bring visibility to the actual … viscosity and change in modulus during the process, then we can … have confidence in the end result.”

Although other digital autoclave control systems are available today, and most will readily improve process control, quality and recordkeeping, Stephens says the system demonstrated that day differed from most because it relies on what AvPro calls the principles of Material State Management (MSM). As the name implies, MSM is not dependent solely on time and temperature information. Instead, it is a viscoelastic feedback-based control system that measures the actual cure state of the polymer material and then uses that information to manage time and temperature parameters. Dorworth demonstrated this distinction, using a simple rheometer to sense viscoelastic change. A part was prepped conventionally for autoclave cure: Thermocouples were placed across the part, and the part was vacuum-bagged. A small, circular sample of the part (1.5-inch/38.1-mm diameter is of sufficient size) was placed in the rheometer‘s closed-cell mold chamber, which is capable of duplicating the autoclave’s thermal cycle. Another sample was placed in a differential scanning calorimeter (DSC), which measures and provides a verification of glass transition temperature (Tg). Dorworth then used the CSS300 software to initiate and control the autoclave cure, sending all sensor data to both the rheometer and DSC. Thus, the sample and part cured simultaneously and, as the cycle progressed, the rheometer gauged the viscosity change in the sample’s matrix as it progressed from initial B-stage through flow, gel and vitrification, recording and analyzing data throughout the process. The DSC used the same data to verify the temperature of vitrification and Tg. Stephens and the other attendees watched — on screen, in real time — the curing material’s viscoelastic change.

Dorworth also demonstrated (see “Steps” on p. 84) how this ability to “see” the change in viscoelastic behavior can enable quick assessment of whether an out-of-specification condition — a broken thermocouple, temperature spike, pressure fluctuation — is cause to scrap a part: A lagging thermocouple (T/C), outside the parameters of the specified cure temperature during another process, indicated that the part might not reach full cure. (Leading and lagging T/Cs correspond to the hottest and coolest areas of the tool.) Dorworth again placed a small circular sample of the part’s layup in a remotely located rheometer and then used the CSS300 software to send data via a wireless connection from the lagging T/C to the rheometer, which processed the sample, using the actual temperature as measured in real time from the problematic cure cycle (the software collects data every half second throughout the cure). The data received back from the rheometer included both viscosity and modulus measurements, graphed to show the viscoelastic change in the laminate sample as based on the lagging T/C and recorded for further analysis and database archive. Because the CSS300 software saves all of the sensor data points, the operator can simply retrieve the run data from the database and send it to the analytical equipment for postprocess evaluation anytime after the run.

It was a situation with which Stephens was all too familiar. With his existing control system, it was unlikely that all T/Cs would stay within the specified temperature range when autoclave loads included both lightweight composite tools, which heat relatively quickly, and heavy aluminum tools, which heat more slowly. The operator had to continuously monitor the T/C readings and manually adjust the heating elements to stay within process specifications. In the case of a T/C failure, Helicomb’s engineers were typically forced to abort the run, pull the part with the failed T/C, install a new one and then, if possible, restart the run. The AvPro system, however, was designed to put the malfunctioning T/C into a “monitoring stage” so that it is no longer controlling the cure, and then continue the run, using the remaining T/Cs and recalculating and resetting new leading and lagging T/Cs.

Traditionally, such a part would be sent to a Material Review Board (MRB) for quality assurance review, where engineers would decide whether to scrap the part or try to duplicate the laminate and process conditions via multiple test coupons and/or try to get core samples of the problem area of the part for DMA or DSC evaluation. (DMA, or dynamic mechanical analysis, determines the modulus of viscoelastic materials by applying an oscillating force and mea-suring the resulting displacement. When a change in temperature is applied and measured, the Tg also can be identified.) With the AvPro software and a rheometer, Dorworth explains, “I can conduct my review of the part in minutes and have the part back in production today versus weeks later, which is what it would take to complete traditional testing.” Another advantage of MSM processing is that cycles can be ended as soon as the viscoelastic goal is achieved — so no safety margins are necessary. Conversely, if areas of the part have not reached goals, the MSM-based control system ensures that the cure cycle continues until they do.

“I was impressed with its ability to improve part quality and better document autoclave runs,” Stephens recalls. “From the software’s graphs of the data, you could actually see where the material changed in viscosity to gel and then to asymptote modulus.” (Asymptote modulus refers to the point on the graph of time, temperature, viscosity and modulus where the modulus value plateaus. See Step 4 and Step 5 on page 84.) Stephens contracted with AvPro to replace Helicomb’s existing system — separate controllers hardwired to each of its two 6-ft by 20-ft (1.8m by 6.1m) autoclaves — with a single CSS300 software package installed on the company’s in-house computer network, enabling both autoclaves to be operated from any Helicomb workstation and, if need be, from Stephens’ personal computer at home as well. (It should be noted that AvPro MSM software is not limited to autoclave curing. It can also be used to control oven cure, resin transfer molding and infusion processes.)

According to Stephens, the trans-formation was immediate and still continues today. Helicomb quickly realized all the benefits demonstrated by Dorworth. Notably, even the multiple ramps and soaks of stepped cures required to cure in-house composite tooling and parts for the F-35 Joint Strike Fighter (JSF) program, which were previously the most difficult for operators to maintain, now pose few difficulties. “Now the manufacturing engineer loads the autoclave, hits the start button and the cure cycle is completely computer-controlled,” says Stephens. In the past, the only break an operator got was during the soak periods of a cure. With digital control, the software now knows to give only 30 percent of heating capacity or 100 percent during ramp and when to back off to avoid overshooting the specified soak temperature while simultaneously adjusting the pressure and vacuum, if needed. Because cure cycles average five to six hours, this has resulted in a dramatic elimination of non-value-added man-hours for Helicomb. Additionally, the time required to train an autoclave operator has dropped from six to eight weeks to one week.

Beyond the expected

Stephens can catalog a lengthy list of other advantages, many of them unexpected. “One of the big benefits for us has been that our manufacturing engineers no longer have to sit and babysit the autoclaves,” he says. Now they spend their time on value-added pursuits, such as maximizing autoclave throughput: The AvPro CSS300 software is used to build a thermal profile for each tool when it is built and again if it is modified. The OEM process spec for a part will detail placing a T/C every so many feet, usually totaling five or six for large parts (up to 6 ft by 10 ft/1.83m by 3.05m). When the T/Cs are linked to the system and the cycle begins, the software determines the leading and lagging T/Cs. This information aids engineers in assessing where T/Cs need to be attached for future runs on this tool. The engineers then assign a profile number to that cure cycle on that tool. “Initially, the profiles helped us with quality,” explain manufacturing engineers Heath Million and Kenny Torres, “because the AvPro software won’t start the autoclave unless the cure cycles for all of the parts being run together are compatible.” For example, if three parts that require soaks at 180°F/82°C, 250°F/121°C and 200°F/93°C, respectively, are mistakenly loaded into the autoclave together, the software will not initiate the run because the profiles are not compatible. According to Stephens this has drastically reduced scrap parts.

Over the past four years, Helicomb has built a thermal profile database on more than 200 tools. Although this is only about half of the company’s tool inventory, this ongoing project has already improved autoclave throughput by 35 to 40 percent. Million and Torres have been able to manage the database to assign as many parts as possible to each cure cycle profile, thus increasing the number of parts that can be run in the autoclave at one time. “At an average cost of $1,500 per autoclave run, we are adding directly to our bottom line by spreading that cost over seven to nine parts per run where we previously would have had only one or two,” says Stephens. “Plus it allows better service to our customers, with faster turnaround times.”

Because Helicomb operates its autoclaves 24 hours a day, five days a week, curing about 12,000 parts each year, the company has amassed a large database, containing complete records of hundreds of cure cycles for each prepreg and tool the company uses. With such data, says Dorworth, “you can use the software to do statistical analysis.” He gives an example: “Say that we have seen an out-of-parameter condition multiple times, but according to the database, all of the parts’ viscoelastic properties were within range for the parts to be considered good. Thus, every time we see that same out-of-spec condition in the future, upon reviewing the data, we can confidently say that no additional testing is even needed.” Such a database effectively replicates the preproduction OEM testing that established the legacy cure specification, in that, instead of simply time and temperature, the database now shows a broadened set of conditions that will allow this part to be cured effectively.

Viscoelastic feedback also gives the manufacturer greater process flexibility. For example, the sensor data in the database from the history of runs on a part can be used to model an alternative set of parameters. This offers increased flexibility for the manufacturer: A shorter cure cycle at a higher temperature could be calculated to meet a very tight customer deadline. Alternatively, throughput might be maximized by lowering cure temperature and holding longer to combine more parts per run.

A most unanticipated plus is that the AvPro software has been an effective marketing tool. Stephens points out that OEMs immediately feel comfortable with Helicomb as a subcontract man-ufacturer because he can e-mail them a complete cure-cycle record for every run. “Now we have something we can show the customer that is hard data, not interpretation,” he says. “It shows them that we have the controls in place to ensure the manufacturing quality and reliability that they must deliver to their customers.”

Moreover, digital documentation makes tracing the infamous “aerospace paper trail” a thing of the past. Previously, says Stephens, “if a customer came back to us two to three years after we made a part and wanted to look at the cure data, we would have to search through boxes of strip-chart recorder rolls. And then if it was an out-of-spec situation, the method of analysis was to use a ruler on the graph and interpret whether or not the sensors were within range.” Today, data for every cure — including readings from every thermocouple and the pressure gauge taken each minute during cure — are recorded and stored, complete with a graph of the entire cycle, and the data file is automatically backed-up on the company server. For part or process audits, Stephens’ team can pull complete data in seconds for every part made with the AvPro system. “We can tell you exactly how many minutes the ramp lasted, exactly what the ramp rate was and exactly what any thermocouple read at any time,” he maintains.

MSM software also can be used to simplify and reduce the cost of quality assurance testing on incoming materials. Legacy methods that are currently the standard for establishing material properties and acceptance criteria involve multiple physical coupon tests. However, with MSM, a single cure cycle is run on a small sample when each new material shipment arrives from the supplier. The material’s state response is then compared to a baseline established using MSM on an earlier shipment. The new material can be accepted or refused based on whether or not the viscoelastic changes (per the rheometer) and the chemical reaction (via the DSC) of the incoming materials are within acceptable tolerances.

An MSM-based future?

“We made the investment in the AvPro digital controls in order to position ourselves for the major aircraft plat-forms that were coming online, like the Boeing 787, the F-22 and the F-35,” says Stephens. He also believes MSM will give Helicomb an edge on its competitors, by enabling the company to work directly with OEMs to develop the best materials and process for a given structure, and that it may even be required in the future. Apparently, that belief is shared within the U.S. Air Force where Material State Management could become a contract requirement. “MSM is definitely how we want to be managing our composites processing,” says Frank Bruce, a materials engineer and member of the Air Force Research Lab (AFRL) Advanced Composites Office at Hill Air Force Base (Roy, Utah). Hill AFB is home to the Air Force Materiel Command (AFMC) Ogden Air Logistics Center, which is the worldwide manager for a wide range of aircraft, engines, missile and accessories components. Bruce has been involved with the development of the AvPro MSM process control software for more than five years and is currently working to validate the technology for the Air Force. His goal is to implement it for all composites production, first at Hill AFB and then at the other depot-level production centers, Tinker AFB (Midwest City, Okla.) and Robins AFB (Warner Robins, Ga.). “Using the MSM process to fabricate composite parts has the potential,” Bruce claims, “to save the government millions of dollars every year and improve the quality and reliability of parts used by our airmen.”