Ore element grade random weight neural network soft measurement method based on XRF

The invention provides an XRF-based ore element grade random weight neural network soft measurement method, which comprises the following steps of: inputting characteristic peak area data of a to-be-measured element obtained in real time by adopting an XRF method and characteristic peak area data of m interference elements corresponding to the to-be-measured element into a constructed RVFL neural network, the grade prediction value of the corresponding to-be-detected element can be output. According to the XRF-based ore element content random weight neural network soft measurement method provided by the invention, the construction of an RVFL neural network comprises the step of calculating the optimal solution of the weight beta from an input layer and a hidden layer to an output layer, only the output weight of the random weight neural network needs to be trained, other weights are randomly set, and the optimal solution of the weight beta is calculated. Gradient information of the network is not needed in th