Pilot-scale anaerobic digesters were operated on ice-cream wastewater for over three years. The performance of four reactor designs, an anaerobic filter, contact process and UASB of capacity 5 m(3), and a 0.5 m(3) fluidised bed, was compared. The anaerobic filter, with a 3.3 m(3) Pall ring bed, operated stably at organic loading rates (B-v) around 6 kg COD m(-3) d(-1), giving total COD removals around 67%. The contact process gave consistently good total COD removals of 80%, but was limited by the poor performance of the settling compartment. Mixed liquor suspended solids did not rise above 3 kg m(-3), and this reactor operated at B-v of 1 kg COD m(-3) d(-1). The fluidised bed reactor, operated on sand or granular activated carbon support media at B-v of 4 and 2 kg COD m(-3) d(-1) respectively, gave about 60% total COD removal, but difficulty was experienced with GAC break-up. The UASB reactor gave the poorest performance, (approximately 50% total COD removal at a B-v of 2 kg COD m(-3) d(-1)) since successful granulation was not achieved. The anaerobic filter was also operated for 9 months after removing half of its packing material, but the former loading rate could not be achieved without instability. Poor biomass retention in all reactors, possibly related to the fat content of the wastewater, was the limiting factor in performance. Using experience gained at pilot scale, a full-scale upflow anaerobic filter was subsequently installed on an ice-cream factory site. This reactor showed similar performance to the pilot-scale filter.
