Abstract: The forging temperature of a molded aging oven is difficult to measure directly. Thus, we build a temperature soft measurement model based on mixed kernel partial least squares algorithm (KPLS). The model estimates the actual forging temperature by collecting the furnace wall temperature, which is easy to obtain. To improve the accuracy of the model, we apply a local weighting algorithm to determine the weights of the training samples. Experimental results show that the soft measurement model of local weighted mixed-kernel partial least squares (LWKPLS) has better adaptability to data and meets the requirements of actual temperature prediction accuracy. It solves the quality problems of aluminum alloy products under oven burning temperature, thereby providing the basis for optimization control of the production process.