Survival analysis method for predicting machine damage time

Abstract

The invention relates to a survival analysis method for predicting machine damage time, which decomposes a survival analysis problem for predicting the machine damage time into sub-problems of a timeslice, and greatly reduces the difficulty of using a neural network to model a long-time sequence prediction problem after decomposing a time sequence prediction problem on the whole time length; therisk probability of each time slice is modeled by using the same neural network, and the final survival probability is obtained through a conditional probability rule. On the premise of not carrying out any assumption on the time distribution of the damage time of the machine, a prediction model can be trained by combining big data. The invention not only can be used for predicting the survival probability of discrete time slices, but also can play a role in predicting the survival probability of continuous time. Experiments prove that the prediction accuracy of the survival analysis model trained through the deep neural network is far higher than that of a traditional method. And through parallel calculation, the algorithm can perform long-distance survival probability prediction under the condition of not increasing the operation time.

Description

technical field [0001] The invention relates to the modeling of the damage time of machinery and equipment in the engineering field, especially the modeling and research of the problem by using the survival analysis method. Background technique [0002] In engineering, survival analysis is often used to predict when a machine will fail. Survival analysis is a discipline that studies survival phenomena and response event data and their statistical laws. The subject is a branch of statistics that is widely used in fields such as medicine, biology, and finance. [0003] Traditional survival analysis methods often require a very strong assumption about the data distribution. For example, the commonly used parameter regression, when using this method, we must first select a distribution, and then use the data to fit the parameters in the distribution equation. There is also a semi-parametric method of the Cox method. This method assumes that the data is distributed with equal ...

AI Technical Summary

Problems solved by technology

[0005] 1. The Chinese invention patent "A Time Series Deep Survival Analysis System Combined with Active Learning" with the application number CN111312393A proposes a survival analysis system combined with active learning. This method is mainly aimed at discrete time domains and is not suitable for continuous time domains.

[0006] 2. The Chinese invention patent "Construction of Survival Analysis Model, Survival Rate Prediction Method, Device and Equipment" with application number CN111243738A proposes a survival analysis model, but this method cannot be modeled for the unique characteristics of a single patient. Can only do coarse-grained modeling

[0008] 1. Kan Ren et al. published Deep Recurrent Survival Analysis "Deep Recurrent Survival Analysis Neural Network" at the 33rd session of the Association for the Advancement of Artificial Intelligence (Association for the Advancement of Artificial Intelligence). This paper uses a recurrent neural network for the survival analysis problem to deal with, its shortcoming is that the output delay of the result is too long

[0009] 2. DeepHit: A Deep Learning Approach to Survival Analysis With Competing Risks published by Changhee Lee et al. at the 32nd session of the Association for the Advancement of Artificial Intelligence (Association for the Advancement of Artificial Intelligence), the article Modeling the survival analysis problem with a neural network has the disadvantage of not being able to fully use the information in the data

[0010] The following conclusions can be drawn from the analysis of relevant domestic and foreign patents and related research: At present, in the field of survival analysis and prediction, there is no fine-grained survival analysis and prediction model that is applicable to both continuous time domain and discrete time domain

Images