There are few things more satisfying in a chemical engineer’s career than when the first drop of at-rate, at-cost, on-spec product comes off a commercial-scale system. That first drop is the culmination of tireless efforts to scale up from idea to reality, and declares you the champion for that moment. 

Scale up begins by defining the factors, levels and responses captured from the bench development effort. A factor is a controllable variable, or knob, that can be turned to a level, or set-point. The response is the characteristic product measurement resulting from the factor at a specific set-point. In feedstock characterization, a common factor is the moisture content, the level is the specific moisture percentage tested in the process system, and the response is the ultimate affect on product conversion.  A designed experiment is a series of tests measuring factor interaction at varying levels to extract meaningful correlations on response affects. This allows an engineer to set boundary limits on both incoming streams and process run conditions for the operators to follow to insure on-spec product. 

Factors and responses are measurements that require validated test methods to ensure that decision making information is based on statistically proven repeatability and accuracy against a known standard. 

A scale-up ratio defines the geometric, kinematic, and dynamic similarities between scales. Scale-up ratios are primarily selected based on the technical novelty of the system and the acceptable level of project risk determined by the investors. Typical scale-up ratios range from 5-1 for a first of its kind application up to 100-1 for a tried-and-true technology.

The factors, levels and responses create a two dimensional analysis of variance (ANOVA) or a three dimensional response surface originating from the bench scale. A model is developed from the correlations of the ANOVA. The first cut of the model is used to screen for main effects. The main effects are the factors that are the controllable, critical few. 

Using scale-up ratios, the pilot-scale is constructed and the predicted model from the bench-scale ANOVA is compared against actual pilot-scale run data. If the predicted values and the experimental data have a strong correlation, then the risk in using the same model and scale-up factors as a basis for the commercial scale is low. If the predicted values and the actual data have a weak correlation you must go back to the bench-scale, adjust the model and note the differences until you develop a correlation that matches the acceptable risk profile of the project. 

In an emerging field such as biorefining, companies compete to bring their technology first to market. The first rush is to prove the technology in a beaker. Occasionally investors see a success on the lab bench and assume the scale-up should just be a matter of routine. This is where many efforts fail. The scale-up is rushed, steps are skipped because the understanding of the complexities of scale-up had not been clearly articulated and thus expectations are not met. The carnage begins. The unfortunate part is that the small failures along the path of the scale-up effort are where the big breakthroughs occur. However, if the team doing the scale-up is either underfunded, or over stressed trying to fit a preassumed, inflexible scale-up expectation, the real pot of gold can be missed. A scale-up plan must be a fluid, agile path that adapts as new information emerges. 

In developing a plan for an entire plant, each major unit operation must be individually scaled with clearly defined input and output streams. 

Output stream specification in biorefining begins with understanding the ASTM requirements for saleable product. Every renewable fuel has a few ASTM specification components that are difficult to meet. It keeps the riff raff out. These components need to be accurately measured to ensure that the material is within specification at the transfer of sale. 

Input streams are qualified using feedstock characterization. To avoid a possible “garbage in garbage out” scenario, the output stream ASTM requirements should be referenced with specific focus on the difficult components. The variations in a feedstock source are best identified by regular and rigorous sampling, testing, and processing of the actual feedstock that will be utilized at the plant. As the feedstock is processed through each unit operation, the collective input variable factors typically reduce in number as waste materials are removed. Thus, garbage in, and good stuff out. 

Feedstock composition is source dependent and material properties must be empirically measured at process conditions to understand variations. Most biorefining feedstocks are mixtures of many constituents and these influence the overall bulk properties of the fluids. For example, manure is a mixture of volatile organic solids, water and other constituents, and published data on manure viscosity and density may not measure up to what is available at the plant. Viscosity and density are two common engineering fluid properties that are affected not only by the temperature and shear rate, but also greatly in the mixture composition that is present for the fluid.

Historically the reactor has been the focus of scale-up. Biorefining presents challenges for input and output unit operation scale-up that could rival the reactor complexities. Three common reactors are a stirred tank batch reactor, a continuous flow batch reactor, and a plug flow reactor. An ethanol fermentor is a pH and temperature controlled stirred tank batch reactor. Yield is the typical response. The correlation between actual and experimental results is usually strong. 

Many installed anaerobic digesters are flow mixed covered lagoons. This is a rudimentary example of a continuous flow batch reactor. Hydraulic retention time (HRT) is the most significant scaling characteristic for sizing an anaerobic digestor system. Typically a lagoon style anaerobic digester has an HRT on the order of 25 to 30 days. The lagoon is sized to be able to handle the conversion of the volatile organic solids to methane based upon anticipated inlet flows. Temperature varies with the environment and economics have been marginal.  To improve yield, anaerobic digesters have adopted temperature-controlled stirred tank batch reactor configurations. In a batch reactor with an agitator, the HRT reduces due to the higher surface area of the solid particles with the substrates and the microbial action on the surface area driving the reaction kinetics. HRT for a well mixed batch reactor can be reduced in half. This enables the reactor itself to have a smaller volume and makes the temperature control and the economics of heating more advantageous due to higher heat transfer of mixing. With the volatile organic solids now uniform from the agitation and at a higher temperature these combinationally drive the reaction kinetics further reducing the HRT and improving the economics. 

The critical few factors and the geometric, kinematic, and dynamic scale ratios of a stirred tank batch reactor are shown in accompanying diagram.

Columns are continuous flow catalytic bed reactors used for conversion of biomass feedstocks into various products. The use of catalytic columns is becoming common in the emerging field of renewable diesel. Hydrogen is added to a catalytic column having been developed to selectively react the feedstock carbon chains to a blend of alkanes meeting the ASTM 975 diesel specification. Catalytic bed reactors are usually continuous plug flow reactors since the catalytic activity is a critical factor in the conversion efficiency. The plug flow regime must be carefully designed so that back mixing does not inhibit the developing progression of the reacting plug flow mass.  

The reactor scale-up is based upon space velocity with the alkane product concentration as the characteristic variable. Space velocity is a term from catalytic column reactor development terminology. Two common terms are liquid hourly space velocity or the weight hourly space velocity, one being volume and the other mass basis. This biorefining application has a long established history in the oleochemical refining industry and the technology has proven transferrable from the hydrogenation of fats, oils and grease. These prior industrial learnings can be applied as a basis for the scale-up using renewable feedstocks to supply the emerging renewable diesel markets.

When scaling any reactor, do a reality check against the common reactor rules of experience. Do not be discouraged if new data and valid results are questioned by previous practitioners that may not support your findings. If you have done a rigorous scale-up and are confident in your results, then the new economical method will reap you the rewards.

Authors: Marc Privitera, Christina Borgese
Founding Engineers, PreProcess Inc.
cborgese@gmail.com