LSTM-based hot-rolling roll-bending force predicting method

A Long Short Term Memory (LSTM)-based hot-rolling roll-bending force predicting method, comprising the following steps:

• 1) Acquiring final rolling data of a stand of a stainless steel rolling mill when performing a hot rolling process, and collecting roll-bending force data for experiment;
• 2) Dividing the roll-bending force data into two parts:
• a training set traindata and a test set testdata according to a specified ratio in time sequence;
• 3) Normalizing the traindata to obtain a normalized vector A;
• 4) building a matrix P by using the vector A in step 3);
• 5) taking first m rows of the matrix P in step 4) as an input and sending the first m rows to an LSTM network;
• 6) using a last row of the matrix P as a label, namely a true value, of the training set, performing calculation on an output value and the true value of the LSTM network by using a formula to obtain an error, and updating a weight and a bias of the LSTM network by a gradient descent method;
• 7) after the LSTM network training is completed, taking last m output data of the LSTM network as an input at a next moment, and then obtaining the output of the LSTM network at the next moment, wherein the output is a predicted value of a roll-bending force at the next moment;
• 8) repeating step 7) until a number of prediction data is obtained;
• 9) performing an inverse normalization on the obtained predicted value of the roll-bending force, and comparing the inverse normalized predicted value of the roll-bending force with the true value in the testdata to check the validity of the LSTM network;
• and
• 10) controlling the roll-bending force of the stainless steel rolling mill based on the comparing in step 9) at a next hot rolling process.