2results about How to "Realize quantitative identification" patented technology

The present application belongs to the field of automobile electronics, and provides a kind of automobile wire harness fuse wire current carrying capacity matching test method and system, comprising: establishing standardization pulse loading profile, constant current test is extended to cover the composite dynamic test sequence of peak thermal shock and average thermal accumulation;Collect fuse thermal accumulation integral value and wire insulation layer multi-point temperature, introduce three-dimensional evaluation index, identify thermal response asynchronous condition;Pulse peak and duty cycle are gradually increased, respectively, approach the limit of wire transient temperature rise and fuse thermal accumulation melting boundary, determine the safe matching interval by cross comparison;According to the safe matching interval, divide the matching level, select the optimization for asynchronous risk and overheating priority failure combination respectively, and feed back the optimization result to the selection specification.The test distortion problem caused by the asynchronous thermal response of fuse and wire under dynamic pulse current is solved, and the matching reliability and test accuracy of automobile electrical system are improved.

The application discloses a soybean bollworm prediction and forecasting method based on machine learning, relates to the technical field of pest situation prediction, and comprises the following steps: collecting trapping count and state information, summarizing the total amount of trapping on the current day, and adopting S-shaped attenuation punishment to calculate an observation reliability index; aligning the phase of the total amount of trapping historical data in a year to determine a key occurrence window, and calculating a phenology continuity index of a future prediction window based on the sowing date and daily air temperature; establishing a mechanism constraint machine learning model based on a feedforward fully connected neural network, and outputting continuous time trapping intensity prediction values and change trend information; integrating the prediction values to obtain future pest situation accumulations, combining environmental suitability and phenological suitability to calculate pest situation risk quantities; generating maintenance instructions according to the observation reliability index, and outputting early warning and patrol frequency adjustment instructions according to the pest situation risk quantities. The application realizes quantitative suppression of low-quality trapping data, mechanism constraint stabilization of trapping intensity prediction, and comparable and reviewable risk early warning.