System for predicting UAV parts supply demand

Abstract

This disclosure proposes a system for predicting the resupply demand of unmanned aerial vehicle (UAV) parts. The system includes a storage unit and a processing unit. The storage unit stores a learned classification model and a regression model, along with a UAV mission record corresponding to a UAV to be analyzed. The classification model is used to estimate the probability of a resupply demand occurrence. The regression model is used to estimate a conditional demand expectation. The processing unit is coupled to the storage unit and configured to acquire UAV feature data based on the UAV mission record; obtain the resupply demand probability using the classification model based on the UAV feature data; obtain the conditional demand expectation using the regression model based on the UAV feature data; and fuse the resupply demand probability and the conditional demand expectation to obtain a parts demand quantity.

Claims

1. A system for predicting the supply demand of unmanned aerial vehicle (UAV) parts, comprising: A storage unit stores a learned classification model and a regression model, as well as a UAV mission record corresponding to a UAV to be analyzed. The classification model is used to estimate the probability of a resupply requirement, and the regression model is used to estimate a conditional demand expectation. A processing unit is coupled to the storage unit and configured to obtain UAV feature data based on the UAV mission record, obtain the resupply requirement probability based on the UAV feature data using the classification model, obtain the conditional demand expectation based on the UAV feature data using the regression model, and fuse the resupply requirement probability and the conditional demand expectation to obtain a part requirement quantity.

2. The UAV parts replenishment demand prediction system as described in claim 1, wherein the UAV characteristic data includes a replenishment timeliness, a replenishment frequency, a cumulative replenishment amount, a flight time, an average replenishment intensity, and a parts replacement rate, wherein the replenishment timeliness is the number of days between an observation deadline and the date of the most recent replenishment event, the replenishment frequency is the number of non-repeating events that trigger replenishment demand during an observation period, the cumulative replenishment amount is the total number of parts required for all replenishment events during the observation period, the flight time is the cumulative flight time from the first mission to the observation deadline, the average replenishment intensity is the cumulative replenishment amount divided by the replenishment frequency, and the parts replacement rate is the number of replaced parts divided by the number of replenishment events.

3. The UAV parts supply demand prediction system as described in claim 1, wherein the classification model employs a gradient boosting decision tree algorithm.

4. The UAV parts replenishment demand prediction system as described in claim 3, wherein the classification model is an XGBoost classifier and the regression model is a LightGBM regressor.

5. The UAV parts supply demand prediction system as described in claim 1, wherein the output of the regression model is a skewed scaling transformation result of the conditional demand expectation value, and the skewed scaling transformation result is inversely transformed to form the conditional demand expectation value.

6. The UAV parts replenishment demand prediction system as described in claim 1, wherein the classification model is learned using a first sample set, the first sample set including complex feature vectors and corresponding complex classification labels, and the regression model is learned using a second sample set, the second sample set including the feature vectors in the first sample set that correspond to the positive classification label and the corresponding complex target variables.

7. The UAV parts replenishment demand prediction system as described in claim 6, wherein the feature vectors included in the first sample set and the second sample set are generated based on mission damage records in the UAV mission records of the complex sample aircraft.

8. The UAV parts supply demand prediction system as described in claim 1, wherein the processing unit further performs a multi-dimensional supply demand statistical analysis based on the parts demand of the UAV to be analyzed for various models.

9. The UAV parts replenishment demand forecasting system as described in claim 8, wherein the multi-dimensional replenishment demand statistical analysis includes parts demand distribution analysis by UAV type, replenishment intensity analysis by mission type, and inventory safety level and replenishment recommendations.

10. The UAV parts replenishment demand prediction system as described in claim 1, wherein after obtaining the probability of the replenishment demand occurrence, the processing unit compares the probability of the replenishment demand occurrence with a threshold, and in response to the probability of the replenishment demand occurrence exceeding the threshold, estimates the expected value of the conditional demand using the regression model, and in response to the probability of the replenishment demand occurrence being lower than the threshold, skips the estimation of the expected value of the conditional demand and directly determines the demand for the parts as zero.

Images