A novel sol-gel synthesis, in which a surfactant acts to make the pore size of the gel network more coarse and uniform, is shown to provide an effective alternative for the consolidation of stone. The new mesoporous silica avoids the main inconvenience of current commercial consolidants, which is their tendency to crack inside the pores of the stone. Since the cracking of xerogels is a well-known drawback of the sol-gel process, the synthesis presented here can be extended to other applications. Finally, preliminary studies of the effectiveness of the novel surfactant-templated sol in consolidating a typical biocalcareous stone are also discussed.

No abstract

No abstract

1 Bioprocess Biosyst Eng (2003) 25:229-233 In 3.3 Learning process, Eq. (2) should read:Furthermore, the second paragraph should read: Y, X, and U are, respectively, output, state and input vectors with dimensions l, n, m. A 1 = block-diag(A 1i ) and A 2 = block-diag(A 2i ) are (n·n) and (l·l)-local feedback block-diagonal weight matrices. (A 1i and A 2i are blocks of A 1 and A 2 with (1·1) dimensions. Equation (3) represents the local stability conditions, imposed on all blocks. A 1i , A 2i of A 1 , A 2 and B and C are (n·m) and (l·n)-weight matrices; D is a (n·l)-global output closed loop matrix; while S(x) is a vector-valued sigmoid activation function. The saturation function could be used as approximation of the sigmoid function to improve the RTNN architecture, [7]. The stability of the RTNN model is assured by the activation functions S and by the local stability conditions of (3).Equations 5 to 11 have been renumbered and the accompanying text corrected.• For the output layer:

This paper proposes using a new recurrent neural network model (RNNM) to predict and control fed batch fermentations of Bacillus thuringiensis. The control variables are the limiting substrate and the feeding conditions. The multi-input multi-output RNNM proposed has twelve inputs, seven outputs, nineteen neurons in the hidden layer, and global and local feedbacks. The weight update learning algorithm designed is a version of the well known backpropagation through time algorithm directed to the RNNM learning. The error approximation for the last epoch of learning is 2% and the total learning time is 51 epochs, where the size of an epoch is 162 iterations. The RNNM generalization was carried out reproducing a B. thuringiensis fermentation not included in the learning process. It attains an error approximation of 1.8%.