Abstract: The present disclosure discloses a method for training an industrial question-answering model based on reinforcement learning and knowledge base matching, comprising: S1, training a reward model, and for an industrial knowledge question-answering, matching and comparing an output of an industrial question-answering model with a content of an industrial knowledge base, and generating a reward value based on a similarity between the output of the industrial question-answering model and the content of the industrial knowledge base; S2, ranking a plurality of reward values corresponding to a plurality of outputs of the industrial question-answering model and training and updating network parameters of the reward model based on a ranking loss function; and S3, training the industrial question-answering model, adding the plurality of reward values to a penalty term, and obtaining an optimal policy after performing a plurality of times of reinforcement learning on the industrial question-answering model using a rein

Abstract: The present disclosure discloses a method for training an industrial question-answering model based on reinforcement learning and knowledge base matching, comprising: S1, training a reward model, and for an industrial knowledge question-answering, matching and comparing an output of an industrial question-answering model with a content of an industrial knowledge base, and generating a reward value based on a similarity between the output of the industrial question-answering model and the content of the industrial knowledge base; S2, ranking a plurality of reward values corresponding to a plurality of outputs of the industrial question-answering model and training and updating network parameters of the reward model based on a ranking loss function; and S3, training the industrial question-answering model, adding the plurality of reward values to a penalty term, and obtaining an optimal policy after performing a plurality of times of reinforcement learning on the industrial question-answering model using a rein

Abstract: The present invention of damping segmental ring structure for subway tunnels built in grim environments of deformable ground can mitigate the stress-concentration of the tunnel lining structures. The deformable ground can be caused by differential settlement or high-intensity earthquakes. Embodiments of the invention have self-adjustment features and forms for deformation and rotation, which comprise one adapter in the middle, two transitional grooved segmental structures, and an internal steel tube. All three forms comprised 3 or 4 pieces with the same features so they can be easily installed, transported and erected on sites and bolts are used to bolt them together to form an integrity structure with damping characteristics. The damper placed in the middle comprises two loading plates that form the shell of the damper, the internal core of the damper which includes interbedded installed rubber pads and steel plates within the loading plates and spring systems that compress the internal core.

Abstract: Embodiments of the present foundation for onshore wind turbines comprise one solid cap structure, one tubelike upright cylindrical structure and a concrete corbel structure. All are constructed of high-strength cast-in-place reinforced concrete. The tubelike cylindrical structure has a purposely enlarged inner diameter than the wind turbine tower. The tubelike cylindrical structure has a wall thickness of 2 to 4 feet and embeds to the ground from 25 to 60 feet. The cap structure is constructed within the encompassed space of the tubelike cylindrical structure and seals the top of the tubelike structure. The anchor bolting system comprises anchor bolts, nuts, washers and embedment ring and is embedded in the reinforced concrete of the solid cap structure at the lower part and bolts the wind turbine tower flange to the solid cap structure.