工業(yè)焊接機(jī)器人機(jī)械，電氣系統(tǒng)復(fù)雜，工作區(qū)域大，運(yùn)行速度快，因而無(wú)法準(zhǔn)確預(yù)測(cè)在不同工況下有可能出現(xiàn)的所有危險(xiǎn)，尤其在人工示教編程或者維護(hù)時(shí)，任何操作失誤和未知的系統(tǒng)缺陷都有可能造成設(shè)備損壞甚引發(fā)重大事故。那么如何在焊接機(jī)器人進(jìn)行設(shè)備預(yù)測(cè)性維護(hù)？山東數(shù)控焊接設(shè)備廠家為您分析：

The mechanical and electrical systems of industrial welding robots are complex, with large working areas and fast operating speeds, making it difficult to accurately predict all the hazards that may occur under different working conditions. Especially during manual teaching programming or maintenance, any operational errors and unknown system defects may cause equipment damage or even major safety accidents. So how to perform predictive maintenance on welding robots? Shandong CNC welding equipment manufacturer analyzes for you:

預(yù)測(cè)性維護(hù)的分類(lèi)

Classification of Predictive Maintenance

預(yù)測(cè)性維護(hù)可以分為基于設(shè)備機(jī)理和基于數(shù)據(jù)驅(qū)動(dòng)預(yù)測(cè)兩種類(lèi)型。基于機(jī)理模型的預(yù)測(cè)是建立設(shè)備故障與機(jī)械動(dòng)力學(xué)、熱力學(xué)和計(jì)量學(xué)等數(shù)學(xué)模型的關(guān)聯(lián)關(guān)系預(yù)測(cè)設(shè)備故障，而數(shù)據(jù)驅(qū)動(dòng)模型則是通過(guò)大量數(shù)據(jù)的學(xué)習(xí)和訓(xùn)練，形成智能化的決策模型。

Predictive maintenance can be divided into two types: device mechanism based and data-driven prediction based. The prediction based on mechanism model is to establish the relationship between equipment failure and mathematical models such as mechanical dynamics, thermodynamics and metrology to predict equipment failure, while the data-driven model is to form an intelligent decision-making model through learning and training a large amount of data.

前者更適用于旋轉(zhuǎn)類(lèi)設(shè)備，數(shù)據(jù)驅(qū)動(dòng)模型更適用于復(fù)雜不確定系統(tǒng)和黑箱過(guò)程的預(yù)測(cè)和控制，數(shù)據(jù)驅(qū)動(dòng)模型是基于經(jīng)驗(yàn)數(shù)據(jù)統(tǒng)計(jì)關(guān)系或統(tǒng)計(jì)特征的預(yù)測(cè)和控制方法，其效果依賴(lài)于輸入數(shù)據(jù)的準(zhǔn)確性和響應(yīng)頻率。

The former is more suitable for rotating equipment, while data-driven models are more suitable for prediction and control of complex uncertain systems and black box processes. Data-driven models are prediction and control methods based on empirical data statistical relationships or statistical features, and their effectiveness depends on the accuracy and response frequency of input data.

預(yù)測(cè)性維護(hù)的實(shí)施流程

Implementation process of predictive maintenance

01

01

數(shù)據(jù)獲取

Data acquisition

通過(guò)模擬仿真和傳感器測(cè)量獲得目標(biāo)設(shè)備或系統(tǒng)的全壽命數(shù)據(jù)。

Obtain full life data of the target equipment or system through simulation and sensor measurement.

02

02

數(shù)據(jù)處理

data processing

包括數(shù)據(jù)預(yù)處理和特征提取，對(duì)數(shù)據(jù)進(jìn)行過(guò)濾和整理，識(shí)別數(shù)據(jù)中工況信息，剔除非重要變量，通過(guò)特征提取的方法得到衰退特征，供模型訓(xùn)練使用。

This includes data preprocessing and feature extraction, filtering and organizing the data, identifying working condition information in the data, removing non important variables, and obtaining decay features through feature extraction methods for model training.

03

03

特征提取

feature extraction 

刪除對(duì)任務(wù)無(wú)有用信息的屬性，對(duì)傳感器數(shù)據(jù)特征提取方法進(jìn)行設(shè)計(jì)，建立基于傳感數(shù)據(jù)特征提取的計(jì)算機(jī)預(yù)測(cè)性維護(hù)模型,并進(jìn)行對(duì)比實(shí)驗(yàn)。

Delete attributes that have no useful information for the task, design feature extraction methods for sensor data, establish a computer predictive maintenance model based on sensor data feature extraction, and conduct comparative experiments.

04

04

模型訓(xùn)練

model training

選擇適當(dāng)機(jī)器學(xué)習(xí)模型，利用經(jīng)處理后的全壽命數(shù)據(jù)進(jìn)行訓(xùn)練，獲得在不同工況下可以對(duì)設(shè)備的故障進(jìn)行準(zhǔn)確預(yù)測(cè)或系統(tǒng)剩余壽命進(jìn)行準(zhǔn)確預(yù)測(cè)的模型。

Select appropriate machine learning models and train them using processed full life data to obtain models that can accurately predict equipment failures or system remaining life under different operating conditions.

05

05

模型驗(yàn)證

Model validation

根據(jù)系統(tǒng)故障預(yù)測(cè)的仿真，可以驗(yàn)證維護(hù)和維修策略的可行性，并將論證結(jié)果導(dǎo)入策略庫(kù)中作為方案。

Based on the simulation of system fault prediction, the feasibility of maintenance and repair strategies can be verified, and the demonstration results can be imported into the expert strategy library as a solution.

06

06

模型部署

Model deployment

部署預(yù)測(cè)性維護(hù)算法模型，根據(jù)工況識(shí)別數(shù)據(jù)的反饋信息進(jìn)行故障診斷，決定設(shè)備或系統(tǒng)的維修策略；根據(jù)現(xiàn)場(chǎng)工況的數(shù)據(jù)進(jìn)行多維度分析進(jìn)行壽命預(yù)測(cè)，決定設(shè)備或系統(tǒng)的維護(hù)和保養(yǎng)策略。

Deploy predictive maintenance algorithm models, diagnose faults based on feedback information from condition identification data, and determine maintenance strategies for equipment or systems; Perform multi-dimensional analysis based on on-site working conditions data to predict service life and determine maintenance and upkeep strategies for equipment or systems.

為解決焊接機(jī)器人規(guī)模化應(yīng)用過(guò)程中操作與維護(hù)規(guī)范化問(wèn)題，通過(guò)分析焊接機(jī)器人應(yīng)用現(xiàn)狀，應(yīng)用意義及發(fā)展前景，展現(xiàn)焊接機(jī)器人操作與維護(hù)規(guī)程必要性，同時(shí)分析焊接機(jī)器人在日常應(yīng)用中存在的不足及問(wèn)題，突出焊接機(jī)器人操作及維護(hù)規(guī)程的重要性。更多相關(guān)事項(xiàng)就來(lái)我們網(wǎng)站http://m.oimhc.com咨詢(xún)！

To address the standardization of operation and maintenance in the large-scale application process of welding robots, the necessity of welding robot operation and maintenance regulations is demonstrated by analyzing the current application status, significance, and development prospects of welding robots. At the same time, the shortcomings and problems of welding robots in daily applications are analyzed, highlighting the importance of welding robot operation and maintenance regulations. For more related matters, come to our website http://m.oimhc.com consulting service