Source: ASSURED BIOTECHNOLOGY CORPORATION submitted to NRP
REAL-TIME PREDICTION OF FORMALDEHYDE (H2CO) EMISSIONS DURING WOOD-BASED PANELS MANUFACTURING
Sponsoring Institution
National Institute of Food and Agriculture
Project Status
COMPLETE
Funding Source
Reporting Frequency
Annual
Accession No.
0217848
Grant No.
2009-33610-19639
Cumulative Award Amt.
(N/A)
Proposal No.
2009-00233
Multistate No.
(N/A)
Project Start Date
Jun 1, 2009
Project End Date
Jan 31, 2010
Grant Year
2009
Program Code
[8.1]- Forests & Related Resources
Recipient Organization
ASSURED BIOTECHNOLOGY CORPORATION
228 MIDWAY LANE, SUITE B
OAK RIDGE,TN 37830
Performing Department
(N/A)
Non Technical Summary
Formaldehyde (H2CO) has been classified as carcinogenic to humans by the International Agency for Research on Cancer (IARC) since 2004. Formaldehyde is the most important aldehyde produced commercially, and is used in the preparation of urea-formaldehyde and phenol-formaldehyde resins. One source of formaldehyde in homes, schools, and offices comes from furniture and cabinets manufactured from composite wood products such as particleboard (PB), medium density fiberboard (MDF), and hardwood plywood (HWPW). Composite wood products use urea formaldehyde (UF) based resins to bind fibers together. Formaldehyde in composites is significant enough that the California Air Resources Board (CARB) recently issued regulations to cap formaldehyde emissions from composite panels used in finished consumer products made from these panels. EPA has been petitioned to make the CARB standard into a federal regulation and is expected to publish an Advanced Notice of Proposed Rulemaking on this issue this fall. The EPA, the Consumer Products Safety Commission and the Centers for Disease Control have identified levels above 0.1 ppm, as a concern for exposure by sensitive populations. Existing standards and regulations (ANSI A208.1 and A208.2, HUD 24CFR3280.308) limit formaldehyde emissions for HWPW, PB, and MDF panels. However, formaldehyde levels in numerous FEMA trailers issued to hurricane Katrina victims greatly exceed the 0.1 ppm benchmark with some even exceeding the OSHA limit of 0.75 ppm, indicating that many high H2CO emitting products go undetected during manufacturing. As H2CO emission standards become more stringent (e.g., CARB regulations, potentially EPA), there is an urgent need for continuous monitoring of formaldehyde emissions during manufacture. Accurate on-line monitoring will allow manufacturers to gain tighter control of emission levels, reduce rejected lots, reduce claims, and improve profitability. Ultimately, real-time formaldehyde emissions monitoring will improve indoor air quality. Assured Biotechnology Corporation proposes to develop a continuous on-line monitoring system that will merge near infrared (NIR) measurements with plant process parameters to achieve high levels of prediction of H2CO. The objective of this Phase I research is to further evaluate an alpha-prototype real-time NIR spectroscopy hardware and software system to accurately detect H2CO emissions. Assured Biotechnology Corporation will use multivariate methods with the NIR spectroscopy data fused with critical process variable data to generate accurate calibration models. A commercial-ready system will be identified from Phase I results. A real-time H2CO on-line monitoring device will have great commercial potential with U.S. manufacturers and with foreign importers to the U.S. as CARB and other standards are enforced on imported products that emit H2CO.
Animal Health Component
100%
Research Effort Categories
Basic
(N/A)
Applied
100%
Developmental
(N/A)
Classification

Knowledge Area (KA)Subject of Investigation (SOI)Field of Science (FOS)Percent
1410410202050%
5110650202050%
Goals / Objectives
The goal of this SBIR Phase I research and development effort is to provide a proof of concept for a continuous monitoring system that will merge NIR spectral data with plant process variables to achieve high prediction results of H2CO emissions during wood composite panels manufacture. The proof of concept will demonstrate implementation of this system for medium density fiberboard (MDF) manufacture. MDF is a wood composite panel that uses a high resin to wood ratio and is one of the highest H2CO emitting products. The first step in Phase I will consist of improving the industrial strength of an alpha-prototype device that includes a full range NIR spectrometer, automation components for light source and referencing. Several Visual Basic programs will be enhanced to control the device, merge NIR spectral data with associated plant process variables and H2CO emission test results, store those data in a database, build partial least squares calibration models and predict H2CO emissions real-time. As soon as the first calibration models are built, validation of the constructed models will be assessed. The impact of the addition of the plant process variables to the NIR spectral data will be measured by comparing the validation performances of calibration models built with NIR spectral data only and models built with NIR spectral and plant process data. Calibration models will allow plant technical managers to identify which process variables are critical for H2CO emission variation. Near the end of the Phase I field study, stored data will be retrieved to assess other multivariate methods such as kernel partial least-squares and kernel ridge regression The overall best performing method will be implemented in the commercial ready system for Phase II. Given that the cost of a full range NIR spectrometer (1000-2500 nm) is substantive, calibration models will also be built and validated using a reduced spectral range (1000-1800 nm). A NIR spectrometer operating in that reduced range is typically one-half to one-third the cost of a full range NIR spectrometer. Findings could have a significant impact on the overall cost of the final H2CO monitoring system during commercialization.
Project Methods
Phase I of this project will require the construction of a custom near-infrared (NIR) spectroscopy device that will be installed in a medium density fiberboard (MDF) or particleboard (PB) production facility with the ultimate goal of predicting long term formaldehyde emissions of finished panels, real-time. A full three months plant trial will be considered. The spectroscopy device will be comprised of a NIR spectrometer operating in the 1000-2500 nm range, an automated white reference material apparatus operated by air cylinders, a tungsten halogen light acting as an excitation source for the wood furnish, and an air flushing system to maintain critical components free of dust. The aforementioned components will be housed in a compact aluminum housing. The device will be operated by a computer that will store collected NIR spectra into a Microsoft SQL database. The custom device will be placed on the side of the forming line, collecting NIR spectra of the wood furnish, before the furnish is being pressed into a panel. A NIR spectrum will be collected every minute, adjusting the collection time based on the production line speed. Results from the production facility testing laboratory that performs formaldehyde tests on panel samples using a small scale chamber (ASTM D6007-02) will be merged with NIR spectra collected on the furnish corresponding to the tested panel samples. As soon as a sufficient number of NIR spectra and corresponding formaldehyde test results will be gathered (ASTM E1655-05), a partial least squares calibration model will be built, allowing the NIR device to start predicting real-time the long term formaldehyde emission of panels. Accurate real-time predictions will allow the plant manager and personnel to gain knowledge of the formaldehyde emissions of finished panels several minutes before being produced and more than 24 hours before being laboratory tested. Most importantly, formaldehyde emissions of the produced panels will be known in between consecutive laboratory tests, closing an information gap of several hours. An attempt will also be made to merge spectroscopic data with plant process variables data, and relate to formaldehyde test results. Improved prediction performance is expected from the added process parameters. MDF and PB plants typically produce several different products within a 24 hours period. Individual calibration models will be constructed for the two most produced products. An attempt will also be made to build a calibration model for all product types. Formaldehyde emission values predicted by the NIR device will be compared to the values found by the small scale chamber tests (performed several times a day, at least 24 hours after the panel sample has been extracted from the production line). Correlation coefficient, root mean square error and average error will be evaluated, providing an unequivocal measure of the NIR device performance.

Progress 06/01/09 to 01/31/10

Outputs
OUTPUTS: The most significant outputs included: 1) The plug in play installation of the near infrared (NIR) hardware. Manufacturing MDF lines run continuously, but we were able to physically install the device during a fifteen minute shutdown period and minimize downtime losses. 2) VPN/IT access and successful implementation of visual basic code and algorithms that enabled real-monitoring remotely and collection of calibration data. 3) Dissemination of data to the management of the test plant, investment groups and to sales and manufacturing parties. Objective 1 (Achieved). Development of a 24/7 NIR spectroscopy device (hardware only). The device housing consists of aluminum framing components and sheet metal paneling. The device is mounted horizontally on one side of a medium density fiberboard (MDF) forming line at one of the outfeeds of the forming bin. Objective 2 (Achieved). Development of a Visual Basic application for Calibration Model Generation. Newly constructed calibration models are stored in a separate SQL table. The model identification number, regression coefficients, calibration and cross-validation correlation coefficients, root mean square errors of calibration and cross-validation are stored. As soon as the calibration models are built, formaldehyde emissions are predicted real-time using the regression coefficients and new incoming NIR spectra. Calibration models are updated if the residuals between predicted and actual formaldehyde emission values exceed 10% of the actual value. Application has been configured with VPN/T-1 access. Objective 3 (Achieved). Collection of data for calibration model construction and real-time prediction. Application has been configured with VPN/T-1 access for remote control and user adjustment. Objective 4 (Achieved). Post-processing of the collected data to assess other multivariate methods and NIR spectral ranges. This validation task was completed at the University of Tennessee. Kernel partial least squares (KPLS) and kernel ridge regression (KRR) methods are employed to build calibration models on the same complete records that were used at the test site by the PLS models. Validations of the KPLS and KRR models were compared to those of the PLS models. Validation results were compared using prediction correlation coefficients and root mean square errors of prediction. Dissemination. Achievements were reported to the MDF plant manager via e-mail, on-site discussions, and presentations. Objectives were also reported to Technology 2020 a local Tennessee Venture placement organization. Limited information was released to Alexeter Technologies and ASDI Corp. Alexeter Technology specializes in the sales of chemical and biological detectors, while ASDI Corp. manufactures NIR spectrophotometers for industrial applications. Both companies express interest in the technology and fulfill the role of distribution and manufacturing, respectively, and are a potential source for future funding and strategic partnership. PARTICIPANTS: (1)Edward Sobek, PhD, was the PI on the project. He traveled to field test sites, worked with MDF plant managers,organized devise installation, reviewed data collections and prepared interim and final reports. (2)Lyn Pope drafted budget and completed all financial reports and tracked spending. (3)Merissa McGraw organized data packages and deliverables. (4)University of Tennessee, Forestry products division. Dr. Tim Young and his post doc Nicholas Andre worked as subcontractors to Assured Biotechnology Corp on advanced statistical algorithm development and data capture technology. (5)MDF manufacture in South Carolina assisted with installation and daily maintenance checks of the beta unit. The plant manager did not want to have the name of the company released. TARGET AUDIENCES: The target industry is the medium density fiberboard manufacturing sector. PROJECT MODIFICATIONS: Not relevant to this project.

Impacts
Change in Knowledge: Environmental conditions in medium-density fiberboard (MDF) manufacturing significantly distorted the imaging matrix. The beta unit was designed with a glass matrix that served as a barrier to prevent plant dust and debris from entering the sealed spectrophotometer enclosure. The design incorporated air-knife technology to keep the glass plate clean of dust during sampling. The system was pre-tested in a wood shop before transport to the beta test site. It performed to specification. However, when installed in the MDF plant, a thin layer of debris accumulated on the glass plate. It was identified as the resin mix used in MDF manufacturing, and it acted like an adhesive on the glass surface; preventing the air-knife from removing all of the debris. The debris significantly degraded the spectral signal, both quantitatively and to a lesser degree qualitatively. Near infrared (NIR) unlike Fourier Transform Infrared (FTIR) is extremely sensitive to signal degradation because detection is based on a set of multivariate statistical algorithms, and not a database query. To remedy the situation, the plant manager was consulted. Unfortunately, the heat originating from the spectrophotometer's halogen light source was a fire hazard when exposed to plant dust and the glass could not be removed. Since the project was already two months behind (explanation in interim), additional funds were not available to continue renting the spectrophotometer equipment. The data that was collected with the glass plate in place indicated that formaldehyde detection, in a continuous MDF plant, was only 80%-85% accurate with the real-time formaldehyde detection system. This value falls below the threshold of acceptability (≥90%) that was predetermined before testing began. This threshold was 5% greater than the 85% detection accuracy required for standard small chamber testing, the current industry standard. The higher standard was derived previous bench scale tests. Change in Action: The knowledge gained, served to initiate the design of a filtered windowless pathway from the spectrophotometer to the MDF surface. This unique design may lend to a patentable, windowless light path that will significantly improve both qualitative and quantitative spectra, without the danger of explosion or fire in the MDF plant environment. The light pathway will be both chemical and debris free via a unique lateral flow filtration design. In the final product a filter pack will be sold as a consumable that must be replaced after a predetermined number of run hours. Once design is complete, Assured Biotechnology Corporation will fund a short test run with the updated system. The MDF test site is extremely interested in Assured Biotechnology's real-time formaldehyde detection technology and they agreed to leave the mounting framework for the formaldehyde detector in place plus provide support during the mounting process of the modified instrument. This will allow the collection of data to position the project for USDA phase II funding application in 2011.

Publications

  • No publications reported this period