significant influencing variables. A statistical analysis (Table
7) of the regression coefficients shows a similar selection
as the CND plot of Figure 5. The significant model
parameters were: â4, â3, â2, (â22), â33, and â44. The final
model is graphically represented as a response surface
projection plot given in Figure 6.
Three optimisation experiments, based on Figure 6, were
carried out in the pilot plant reactors. Varying the amount
of catalyst Cu(II) as graphed in Figure 7 shows an improved
yield of vanillin.
Using the same type of raw material, the optimised
oxidation procedure gives a vanillin yield of 5.7% against a
yield of 4.1% with the current procedure which corresponds
to a relative improvement of 37%. The yield obtained in the
optimisation experiments is however 0.5%, corresponding
to 9.6% relative higher yield than predicted by the response
surface model represented in Figure 6. This observation is
most probably due to the natural variation in the LS750 raw
material. Similar variation in the vanillin yield is frequently
observed also in plant scale reactors.
(B) Oxidation of LS from High-Temperature Digestion
Process (LS750 and LS850) with Copper Oxide as Catalyst.
The copper (II) catalyst, which is used in the vanillin plant,
is charged to the hydrolysis-oxidation reactors as a water
solution of copper sulphate. When the oxidation reaction is
completed, the copper is recovered as sludge of inorganic
and organic substances, from which the copper is extracted
by using ammonia and precipitated as copper (II) oxide.