Powder Blending From Art to Science
Predictive models, PAT, and continuous processing promise to reduce variability and improve product quality.
By Angelo De Palma, Ph.D., Contributing Editor
Dry powder blending may be the most widely recognized unit operation in pharmaceutical manufacturing, but it’s also one of the least understood. The uniqueness of each individual drug formulation assures that no two blending processes can ever be identical.
Powder blending’s unpredictability has challenged engineers to explain in quantitative terms a phenomenon typically described empirically. One team at Rutgers University (New Brunswick, N.J.), led by engineering professors Benjamin Glasser and Troy Shinbrot, is moving closer to this elusive goal.
Their research, so far, has yielded surprising results. For one thing, it suggests that agitating a mixture longer and faster will not always result in a homogeneous blend. Blending that appears uniform may degrade into turbulence, causing ingredients to separate into layers.
In one experiment, researchers found that fine glass beads of different sizes blended uniformly at low mixing speeds, but formed distinct layers as mixing speed increased. The researchers were able to identify patterns of granular motion that promoted layer formation and interfered with uniform mixing.
|Spectral Dimensions' Blend Monitor, show here mounted on the blender base, and accompanying software optimize blending through in-process chemical imaging. Courtesy of Spectral Dimensions.|
Similar phenomena have been observed and predicted in pollution control and meteorology, but, at this point, engineers don’t fully understand their implications in powder or granular mixing. “We do not have predictable equations of transport or mixing that span the range of granular behaviors from load-bearing and solid-like to flowing and fluid-like,” Professor Shinbrot explains. Such equations are readily available for fluids, allowing for prediction of transport and mixing in those systems. Engineers can also simulate solid-like systems using modified elasticity models, and fluid-like systems with modified viscous models.
With powder blenders, modeling requires a combination of heuristics, rules of thumb, and experimentation. If particle sizes change even slightly, mixing modes shift between smoothly varying, regular regimes and erratic, chaotic ones. “If we could predict which regime would occur, we could design equipment and control blending processes with much more robustness and much less variability in outcome due to environmental factors like humidity, and material properties like particle size and cohesivity,” Shinbrot adds.
Almost any agitation protocol such as tumbling, ribbon blending or vibration can alternately produce mixed or segregated outcomes with variations in particle size, density or shape. In practical terms, this means that a blender can give poor blending or even segregation with bone-dry powders, but a well-mixed blend if a very small amount of water vapor is allowed into the process.
At this point, unpredictability is the only thing that can be predicted for powder blending, although research promises to change the picture in the future. Such change could vastly improve efficiency. “We currently do not have predictive capability,” admits Shinbrot, “so the [blending] analytics consist of performing trial-and-error experiments followed by statistical sampling—an inefficient, risky way of doing concentration-sensitive pharmaceutical processing.”PAT drives continuous blending
As researchers work on predictive systems, vendors are developing blenders that can be operated continuously, featuring automated weighing and feeding features and advanced analytics. The advantages of continuous blending over batch are obvious, including a smaller footprint, lower capital investment, fewer worker-hours required per unit produced, and fewer discontinuities related to stopping a process, recovering the product, cleaning up, and recharging. Continuous blenders are also easier to clean since they are straight-flow systems without the valves, bells and whistles of contained blenders.
Regulatory uncertainty and lack of availability of analytics impeded adoption of continuous technology in the past. Feeder mechanisms simply were not accurate enough to assure that the right blend would emerge from the downstream end, says Steve Knauth, national sales manager at Munson Machinery (Utica, N.Y.). “Now, with gravimetric feeders, any manufacturer in any industry can get measurable consistency in every production run provided appropriate care is taken to monitor the feed stream,” he says.
This has implications for improved control. “With all the new analytical devices emerging, it’s become easy to install feedback loops midway through and at the end of a blending machine,” observes Dr. Thomas Chirkot, laboratory manager at Patterson-Kelley (East Stroudsburg, Pa.). “You could say that process analytical technology (PAT) is driving continuous blending, and vice versa.”
Adapting blenders for continuous processing requires only minor modifications, Chirkot says. For one thing, batch blenders are typically closed, while continuous models require up- and downstream ports and accurate feeders. In some cases, they may also need additional pumping systems to remove blended product.
Size is another issue. Given lower volume operations and a greater need for lot control, Knauth says, “The smallest machine we make today is likely larger than what would be necessary for pharmaceutical processing. While the blender design is critical, steady, consistent product metering through feeding mechanisms is equally critical,” he adds. “The blender will homogenize whatever it sees from the feeders, whether it’s introduced in the proper proportions or not.”