A Study of Granulation Methods for Lincoln ElectricStudent names: Charlotte Nottingham, Samantha Nobbe, Bryan Shaw, Simon Ray, and Keaton Jacobi Graduate Student: Line KoleilatFaculty Advisors: Professor Paul MortIndustrial Sponsors: Eric Gulliver

Project Background

The senior design team working on the Lincoln Electric project is seeking to assist Lincoln Electric indeveloping improvements to the granulation processes of SAW granular fluxes to increase efficiencyin production, specifically to reduce the mass fraction of particles that are too small (fines) or toolarge (overs), and thus fail to meet strict sizing criteria. Granulation processes such as high-shearmixing, Forberg paddle mixing, and fluidized bed granulation were tested to determine whichprocess would result in prototype granules best matching the samples produced by Lincoln Electric.

Experimental Procedure

Results and Discussion

Recommendations

Summary of Effects

MSE 430-440: Materials Processing and Design

This work is sponsored by Lincoln Electric, Cleveland, OH.

Lincoln Electric is considering an upgrade to itsmanufacturing process for its product family of bondedSubmerged Arc Welding (SAW) fluxes and is seekingan analysis and prototype demonstration using variousgranulation methods available at Purdue’s Center forParticulate Products and Processes (CP3). BondedSAW fluxes are typically a blend of powders (alumina,silicates, fluorspar, magnesia, and other minerals),bonded together with aqueous silicates in agranulation process.

The goal of this project is to determine the bestgranulation method to reduce out of size granuleswhile maintaining current product properties, such asshape, size, and composition.

The Forberg mixer (above) was chosen for testinggranulation processes due to its fluidization zonewhere binder could be injected, and the ease ofinstallation in Lincoln Electric’s current facility.

Lincoln Electric providedsamples to use as abaseline to work towards.

Of the many granulation processes we tested withdifferent parameters, prototype “X” had the mostsimilar size distribution of accepts to our target with13% overs and φd of 45.0°. “X” was made with 18%“Binder / Powder, pre-cap” ratio, 1% “cap - powder”,3% “Acid/Silicate” ratio added before powder cap.

A Canty SolidSizer was used to collect size and shapedata on the granules through image analysis, and aFlowdex was used to determine the angle of repose.

This plot is an example ofthe images collected bythe SolidSizer. Here, theparticles are plotted bytheir elliptical form factorand aspect ratio. Thecloser both variables areto 1, the more sphericaland smooth the granulesare; leading to increasedflowability.

As samples were created,they were characterizedin a similar manner.

Early samples werevery promising andwithin the acceptableranges, except for thenumber of overs aseach batch containedbetween 20% and30% overs.

Since the granules need to be able to flow well andpile up over a weld joint, the drained angle of repose(φd) needs to be between 40-45 degrees. This targetwas achieved in most prototype batches.

After the addition of the binder,several capping methods aretested by the addition of the fifthraw material powder and the acid.

1. Cap with 5th material, no acid2. Cap with 5th material, add acid3. Add acid, cap with 5th material

The five raw material powdersused are weighed and added tothe mixer. The materials are dry-mixed to ensure evendistribution. A select fraction ofthe fifth powder material is leftout and is used later to cap thegranules due to its high surfacearea.

After dry-mixing, the sodium silicatebinder is injected into the fluidizationzone in the center of the mixer. Thegel produced by the sodium silicatesolution can be “set” with theaddition of a weak acid. Acid is not atypical component and is mainlyused in our prototype batches toexpedite their preparation forcharacterization.

Sodium Silicate Solution

Hydrogel

Addition of 5% acetic acid

Once the granulation process isfinished, the granules are dried andsieved to remove any granuleslarger than 1700μm which areconsidered “overs” and cannot beused for welding. These aretypically recycled and reworked inproduction. Granules smaller than1700μm are termed “accepts.”

Compiling the processing parameters with theassociated changes in particle characteristics, wecreated a linear multivariate least squares fitregression model with JMP analytical software. Withthe JMP Prediction Profiler, general trends in effectson granule characteristics created by changes in theprocessing factors independent of each other areobserved.

The goal of this analysis is to determine whichprocessing parameters will yield granulations withreduced overs without creating accepts that have toowide variations and too low average sizes.

Contour plots derived from the model show theprocessing parameter values interacting with oversand geometric mean particle size. Reducing the“Binder / Powder, pre-capping” ratio can reduce thenumber of overs but is also correlated with adecrease in dg of accepts. Increasing the “Capping, %of Total Mixture” allows for an increased “Binder /Powder, pre-capping” ratio without increasing overswhile also increasing the dg of accepts. An increased“Binder jet / tip speed” ratio correlates with increaseddg of accepts with little effect on overs. Overs couldalso be reduced by increased concentration of silicatein the binder solution.

Granulations with φd >45° are capped withconcentrated acid andhave lower FF* values

Based on these findings, Lincoln Electric shouldseriously consider a modified Forberg mixer as afeasible candidate to replace their current granulationmixer. We believe this will increase productionefficiency, reduce cost, and produce a product thatfalls within their ideal design criteria.