2results about How to "Highly replaceable" patented technology

This invention discloses a spring impact fatigue testing device, comprising a base mechanism, a height adjustment mechanism, and an impact mechanism. The base mechanism includes a base plate, an adjusting seat plate, and a fixed mounting base, the adjusting seat plate and the fixed mounting base being horizontally adjustable along a relatively vertical path. The height adjustment mechanism includes a lifting bracket, a support base, and a lifting slide, the support base being fixedly connected to the base plate. The impact mechanism includes a swing arm, a swing arm connecting seat, a swing mounting base, and a mounting frame, the swing arm connecting seat being fixed to the side of the lifting slide, the swing arm being pivotally connected to the swing arm connecting seat, the mounting frame being slidably and lockably mounted on the swing arm, the swing mounting base being located on the lower side of the mounting frame, and the upper end face of the fixed mounting base and the lower end face of the swing mounting base respectively having a retaining structure adapted to the shape of the spring end. The height adjustment mechanism is equipped with a lifting control component for adjusting the height of the lifting slide. The impact mechanism is equipped with a drive control component for driving the mounting frame to perform up and down impact actions.

The application provides a civil aviation vertical field language model training method and reasoning method and device, and relates to the field of artificial intelligence. The model comprises a plurality of task agents, a plurality of low-rank matrices, a knowledge database and a language model as a base model; the method is based on a task to construct an external knowledge database; a structured prompt word text is generated by using a task agent module, a standardized reasoning result is matched for the generated prompt word text; the training data of each task is constructed by using the structured prompt word text and the standardized reasoning result; the low-rank matrix parameters are trained by using the training data of the task, and the trained low-rank matrix parameters corresponding to the task are obtained; the trained low-rank matrix parameters corresponding to the task and the language model as the base model are integrated as the trained fine-tuning language model corresponding to the task; the scheme combines the existing general language model to construct the civil aviation vertical field language model which can quickly adapt to market changes and can effectively serve the whole ecological field.