6results about How to "Guaranteed distillation effect" patented technology

This utility model discloses a purification device and an electrolyte preparation system for 2-methyl-adipic acid. The purification device first purifies the crude electrolyte with high viscosity through a thin-film evaporator, separates the light components, and then performs distillation. This combines the advantages of high evaporation efficiency of the thin-film evaporator with the distillation effect of the distillation column, resulting in a high-purity purified product, which promotes the industrial production of high-quality aluminum electrolytic capacitor electrolyte.

The utility model discloses a rectification device for producing triallyl isocyanurate, which belongs to the technical field of triallyl isocyanurate processing and comprises a heating platform and a rectification column. A plurality of heating pipes for heating the rectifying column are arranged in the heating platform, a groove for accommodating the rectifying column is formed in the surface of the heating platform, and the heating pipes are arranged along the edge of the groove; the rectifying column is located in the groove and is horizontally arranged, a plurality of rectifying outlets are formed in the rectifying column, and sealing blocks are arranged at the positions of the rectifying outlets and are used for sealing the rectifying outlets. According to the rectification device disclosed by the utility model, effective rectification treatment can be carried out on triallyl isocyanurate, the rectification column is horizontally arranged, the structural stability is stronger, the rectification column does not need to be fixed by adopting more fixing pieces, the component cost of the device is saved, and the rectification effect on the triallyl isocyanurate is ensured.

The utility model relates to a rectifying device for biodiesel, including esterificator, rectifying tower and condenser, rectifying tower includes reboiler and the rectifying column of solid connection in reboiler top, first heating pipe is solidly established in reboiler, the rectifying column includes the first rectifying section, secondary heating section and second rectifying section of solid connection from below to above in proper order, second heating pipe is solidly established in secondary heating section, the outlet of esterificator communicates with the import of reboiler through first pipeline, and the top outlet of second rectifying section communicates with the import of condenser through second pipeline, and the outlet of condenser communicates with the recovery mouth of esterificator, the outlet of first heating pipe communicates with the import of second heating pipe through third pipeline, the utility model improves the rectification effect and utilization ratio to methyl alcohol, and then improves methyl alcohol purity, and the methyl alcohol purity is conveniently detected, and it is more convenient to use.

The present application relates to the technical field of artificial intelligence and distributed privacy computing, and particularly relates to a global model correction method for federated learning based on adaptive screening of generated data. The method is applied to a server, the server and a plurality of clients constitute a federated learning system, and the method comprises: performing federated learning iteration with the plurality of clients to obtain a target classification model; wherein the tth iteration in the federated learning iteration is: broadcasting a (t-1)th round master classification model and a target generator to the clients, freezing the generator by the clients, reporting after training the classification model and the discriminator, aggregating the classification model parameters by the server, training the generator with the multi-discriminator parameters reported by the clients, outputting a pseudo sample set by the generator, performing weighted knowledge distillation with the aggregated model as a teacher and the master model of the last round as a student to obtain a master classification model of the present round. The present application can improve the model application effect of the global model finally output by the federated learning system.

The utility model relates to the technical field of ethanol processing, and provides an ethanol production rectification device which comprises a rectification tank and a condensation tank, the condensation tank is fixed outside the rectification tank and communicated with the top of the rectification tank, a driving motor is installed at the bottom of the rectification tank, and a condensation separation assembly is arranged in the condensation tank. The heating and stirring assembly is arranged in the rectification tank and comprises a smashing frame, the smashing frame is rotationally connected to the bottom of the rectification tank, the output end of the driving motor is connected with the smashing frame, and a pair of fixing frames are fixedly connected to the inner wall of the rectification tank. According to the technical scheme, the problems that the working efficiency is reduced due to the fact that no stirring mechanism is arranged in equipment and the rectification time is prolonged in the related technology, and the rectification effect is affected due to the fact that particles are precipitated when ethyl alcohol is directly poured into the rectification equipment and are accumulated at the bottom are solved.

The present invention discloses a knowledge distillation method based on parameter efficient module and multi-teacher knowledge distillation, the method comprising: obtaining a downstream task data set, generating multiple sub-teacher models according to a pre-constructed parameter efficient multi-teacher model, fine-tuning each sub-teacher model to obtain multiple fine-tuned sub-teacher models; training a student model to be trained according to the downstream task data set, generating a supervisory signal for the training according to the downstream task data set and multiple fine-tuned sub-teacher models, updating parameters according to the supervisory signal and the training result obtained by the student model during the training process, and obtaining and outputting the trained student model when the preset training conditions are met; obtaining target processing data, inputting the target processing data into the trained student model, and outputting the target result. The present invention can take into account both efficiency and accuracy when training the student model, so as to generate accurate target results through the trained student model.