TWI663555B - Reliability analysis method based on market feedback data - Google Patents

Reliability analysis method based on market feedback data Download PDF

Info

Publication number
TWI663555B
TWI663555B TW105123631A TW105123631A TWI663555B TW I663555 B TWI663555 B TW I663555B TW 105123631 A TW105123631 A TW 105123631A TW 105123631 A TW105123631 A TW 105123631A TW I663555 B TWI663555 B TW I663555B
Authority
TW
Taiwan
Prior art keywords
data
mathematical
interval
reliability
reliability analysis
Prior art date
Application number
TW105123631A
Other languages
Chinese (zh)
Other versions
TW201804378A (en
Inventor
王福琨
朱道鵬
呂亦宸
Original Assignee
國立臺灣科技大學
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 國立臺灣科技大學 filed Critical 國立臺灣科技大學
Priority to TW105123631A priority Critical patent/TWI663555B/en
Publication of TW201804378A publication Critical patent/TW201804378A/en
Application granted granted Critical
Publication of TWI663555B publication Critical patent/TWI663555B/en

Links

Abstract

本發明實施例提供一種基於市場回饋資料之可靠度分析方法。基於市場回饋資料之可靠度分析方法包括以下步驟:蒐集市場產品回饋資料;整理並建立產品失效存活數據;將區間數據資料輸入至分析軟體以進行數學模型之配適;使用衡量指標值進行數學模型配適度之選擇;以及將最配適模型進行可靠度概率推估。 Embodiments of the present invention provide a reliability analysis method based on market feedback data. The reliability analysis method based on market feedback data includes the following steps: collecting market product feedback data; collating and establishing product failure survival data; inputting interval data data to analysis software for mathematical model adaptation; using measurement index value for mathematical model The choice of appropriateness; and the most appropriate model for reliability probability estimation.

Description

基於市場回饋資料之可靠度分析方法 Reliability analysis method based on market feedback data
本發明係關於一種可靠度分析方式,特別是關於一種是針對大數據基礎下產品市場回饋資料運用多種機率模型,估算產品可靠度及失效率的方法,進而透過人性化介面開發成可靠度分析軟體。 The invention relates to a reliability analysis method, in particular to a method for estimating product reliability and failure rate by using a plurality of probability models for product market feedback data under the big data basis, and then developing a reliability analysis software through a humanized interface. .
預估電子資訊產品壽命一直皆為電子業界定義新產品保固之重要指標,無論軍需品還是民需品都必須有明確之可靠度指標,並且需於產品說明書或包裝上將其明確地標示。故,瞭解並掌握可靠度之計算及預測係市場之需要與趨勢,也係生產廠商之責任與核心技術。 It is estimated that the life of electronic information products has always been an important indicator for the definition of new products in the electronics industry. Whether it is munitions or civilian products, there must be clear reliability indicators, and they must be clearly marked on the product specification or packaging. Therefore, understanding and mastering the calculation and prediction of reliability is the market's needs and trends, as well as the responsibility and core technology of the manufacturer.
目前市面上企業較常使用的可靠度評估方法,主要為壽命預估及壽命實測兩種。此兩種方法既耗時耗成本,又不能代表產品的真實使用狀況且無法呈現產品的真實市場使用資訊。而最能展現產品的真實市場資訊則是產品的市場回饋數據,然而,當業主得到市場回饋資料時,通常不會使用有效利用數據,導致在進行產品可靠度分析,仍使用上述兩種方法,既昂貴耗時且失真。 At present, the reliability evaluation methods commonly used by enterprises in the market are mainly life expectancy and life measurement. These two methods are time consuming and costly, and do not represent the actual use of the product and can not present the real market usage information of the product. The real market information that best shows the product is the market feedback data of the product. However, when the owner receives the market feedback data, the effective use data is usually not used, resulting in the product reliability analysis, and still use the above two methods. It is expensive, time consuming and distorted.
有鑑於此,本揭露內容提供一種基於市場回饋資料之可靠度分析方法,使得企業能有效使用產品的市場回饋資料,進而進行可靠度解析,以供產品設計改良開發。 In view of this, the disclosure provides a reliability analysis method based on market feedback data, so that the enterprise can effectively use the market feedback data of the product, and then perform reliability analysis for product design improvement development.
本發明實施例提供一種基於市場回饋資料之可靠度分析方法。基於市場回饋資料之可靠度分析方法包括以下步驟:蒐集市場產品回饋資料;整理並建立產品失效存活數據;將區間數據資料輸入至分析軟體以進行數學模型之配適;使用衡量指標值進行數學模型配適度之選擇;以及將最配適模型進行可靠度概率推估。區間數據資料為界定時間區間內的產品失效存活數據。 Embodiments of the present invention provide a reliability analysis method based on market feedback data. The reliability analysis method based on market feedback data includes the following steps: collecting market product feedback data; collating and establishing product failure survival data; inputting interval data data to analysis software for mathematical model adaptation; using measurement index value for mathematical model The choice of appropriateness; and the most appropriate model for reliability probability estimation. The interval data is the product failure survival data within the defined time interval.
在本發明其中一個實施例中,其中該分析軟體更包括以下步驟:以最大概似估計法建立一概似估計函數。 In one embodiment of the invention, wherein the analysis software further comprises the step of establishing an approximate estimation function in a most approximate likelihood estimation method.
在本發明其中一個實施例中,其中該分析軟體更包括以下步驟:以差分演算法對該概似估計函數進行參數估計,以找到最適合市場回饋數據之適當模型。 In one embodiment of the present invention, wherein the analysis software further comprises the step of parameter estimating the approximate estimation function by a differential algorithm to find an appropriate model that is most suitable for market feedback data.
在本發明其中一個實施例中,其中差分演算法為差分進化演算法(Differential Evolution Algorithm,DEA)。 In one embodiment of the present invention, the differential algorithm is a Differential Evolution Algorithm (DEA).
在本發明其中一個實施例中,其中透過赤井信息法則(Akaike Information Criterion)來進行該數學模型的配適選擇。 In one embodiment of the invention, the adaptive selection of the mathematical model is performed by Akaike Information Criterion.
在本發明其中一個實施例中,其中數學模型為韋伯分配模型,韋伯分配模型在區間資料型態下為概似估計函數。 In one embodiment of the present invention, wherein the mathematical model is a Weber distribution model, the Weber distribution model is an approximate estimation function in the interval data type.
在本發明其中一個實施例中,其中數學模型為Lognormal分配模型,Lognormal分配模型在區間資料型態下為概 似估計函數。 In one embodiment of the present invention, wherein the mathematical model is a Lognormal allocation model, and the Lognormal distribution model is an interval data type. Like estimation function.
在本發明其中一個實施例中,其中數學模型為Exponential分配模型,Exponential分配模型在區間資料型態下為概似估計函數。 In one embodiment of the present invention, wherein the mathematical model is an Exponential allocation model, the Exponential allocation model is an approximate estimation function in the interval data type.
在本發明其中一個實施例中,其中數學模型為Log-logistic分配模型,Log-logistic分配模型在區間資料型態下為概似估計函數。 In one embodiment of the present invention, wherein the mathematical model is a Log-logistic distribution model, and the Log-logistic distribution model is an approximate estimation function in the interval data type.
在本發明其中一個實施例中,其中數學模型為Burr XII分配模型,Burr XII分配模型在區間資料型態下為概似估計函數。 In one embodiment of the present invention, wherein the mathematical model is a Burr XII distribution model, the Burr XII distribution model is an approximate estimation function in the interval data type.
綜上所述,本發明所提供之基於市場回饋資料之可靠度分析方法至少具有以下優點:使得企業能有效使用產品的市場回饋資料,進而進行可靠度解析,以供產品設計改良開發;以及具有相當人性化的介面,使得企業在獲得產品投入市場所積累的相關資料後,能快速且容易地進行產品可靠度分析。 In summary, the reliability analysis method based on the market feedback data provided by the present invention has at least the following advantages: enabling the enterprise to effectively use the market feedback data of the product, and then performing reliability analysis for product design improvement and development; The user-friendly interface enables companies to quickly and easily analyze product reliability after obtaining relevant data accumulated by the product.
為使任何熟習相關技藝者了解本發明之技術內容並據以實施,且根據本說明書所揭露之內容、申請專利範圍及圖式,任何熟習相關技藝者可輕易地理解本發明相關之目的及優點,因此將在實施方式中詳細敘述本發明之詳細特徵以及優點。 In order to make the technical content of the present invention known to those skilled in the art and to implement the present invention, and in accordance with the disclosure, the scope of the application, and the drawings, the related objects and advantages of the present invention can be easily understood by those skilled in the art. The detailed features and advantages of the present invention will be described in detail in the embodiments.
S110、S120、S130、S140、S150‧‧‧步驟 S110, S120, S130, S140, S150‧‧ steps
S132、S134‧‧‧步驟 S132, S134‧‧‧ steps
X1、X2、X3、X4、X5‧‧‧產品 X1, X2, X3, X4, X5‧‧‧ products
圖1為根據本發明例示性實施例所繪示之可靠度分析方法的流程圖。 FIG. 1 is a flow chart of a reliability analysis method according to an exemplary embodiment of the present invention.
圖2為根據本發明例示性實施例所繪示之可靠度分析方法之另一流程圖。 2 is another flow chart of a reliability analysis method according to an exemplary embodiment of the present invention.
圖3為根據本發明例示性實施例所繪示之產品失效之時間區間之示意圖。 FIG. 3 is a schematic diagram showing a time interval of product failure according to an exemplary embodiment of the present invention.
圖4為根據本發明例示性實施例所繪示之區間數據資料輸入介面之示意圖。 4 is a schematic diagram of an interval data data input interface according to an exemplary embodiment of the present invention.
圖5為根據本發明例示性實施例所繪示之數學模型進行模型配適之示意圖。 FIG. 5 is a schematic diagram of a mathematical model for model adaptation according to an exemplary embodiment of the present invention.
圖6為根據本發明例示性實施例所繪示之計算出產品的可靠度與失效率之曲線圖。 FIG. 6 is a graph of calculating reliability and failure rate of a product according to an exemplary embodiment of the present invention.
在下文將參看隨附圖式更充分地描述各種例示性實施例,在隨附圖式中展示一些例示性實施例。然而,本發明概念可能以許多不同形式來體現,且不應解釋為限於本文中所闡述之例示性實施例。確切而言,提供此等例示性實施例使得本發明將為詳盡且完整,且將向熟習此項技術者充分傳達本發明概念的範疇。在諸圖式中,可為了清楚而誇示層及區之大小及相對大小。類似數字始終指示類似元件。 Various illustrative embodiments are described more fully hereinafter with reference to the accompanying drawings. However, the inventive concept may be embodied in many different forms and should not be construed as being limited to the illustrative embodiments set forth herein. Rather, these exemplary embodiments are provided so that this invention will be in the In the drawings, the size and relative sizes of layers and regions may be exaggerated for clarity. Similar numbers always indicate similar components.
應理解,雖然本文中可能使用術語第一、第二、第三等來描述各種元件,但此等元件不應受此等術語限制。此等術語乃用以區分一元件與另一元件。因此,下文論述之第一元件可稱為第二元件而不偏離本發明概念之教示。如本文中所使用,術語「及/或」包括相關聯之列出項目中之任一者及一或多者之所有組合。 It will be understood that, although the terms first, second, third, etc. may be used herein to describe various elements, such elements are not limited by the terms. These terms are used to distinguish one element from another. Thus, a first element discussed below could be termed a second element without departing from the teachings of the inventive concept. As used herein, the term "and/or" includes any of the associated listed items and all combinations of one or more.
以下將以至少一實施例說明本發明之基於市場回饋資料之可靠度分析方法。本發明在大數據資料下針對產品市場回饋資料進行可靠度壽命分析,其發明主要利用多種機率分配模型及運用參數估計方法,找出最配適數據之機率模型,進而推估產品可靠度資訊,與目前市面上之壽命評估方法有很大的差異。 The reliability analysis method based on the market feedback data of the present invention will be described below with reference to at least one embodiment. The invention performs reliability life analysis on product market feedback data under big data data, and the invention mainly utilizes multiple probability distribution models and uses parameter estimation methods to find the probability model of the most suitable data, and then estimates product reliability information. There is a big difference between the current life assessment methods on the market.
〔基於市場回饋資料之可靠度分析方法的一實施例〕 [An embodiment of a reliability analysis method based on market feedback data]
本揭露內容透過建構市場回饋資料可靠度分析方法並開發成軟體,此分析軟體將可以提供業主有效運用市場回饋資料並用之進行可靠度分析,其分析結果將能提供企業進行相關產品設計開發改良。本揭露內容著重在對產品市場回饋資料進行有效地處理,並運用參數估計方法進行機率模型配適,找出最適模型,進而推估產品可靠度資訊,其研究成果可應用於多種產業界,並及時提供分析結果。 Through the construction of market feedback data reliability analysis method and development into software, the analysis software will provide the owner with effective use of market feedback data and use it for reliability analysis. The analysis results will provide enterprises with relevant product design and development and improvement. The content of this disclosure focuses on the effective processing of product market feedback data, and uses the parameter estimation method to match the probability model, find the optimal model, and then estimate the product reliability information. The research results can be applied to various industries, and Provide analysis results in a timely manner.
請參照圖1與圖2,圖1為根據本發明例示性實施例所繪示之可靠度分析方法之流程圖。圖2為根據本發明例示性實施例所繪示之可靠度分析方法之另一流程圖。如圖1與圖2所示,基於市場回饋資料之可靠度分析方法100包括以下步驟:蒐集市場產品回饋資料(步驟S110)。整理並建立產品失效存活數據(步驟S120)。將區間數據資料輸入至分析軟體以進行數學模型之配適(步驟S130)。使用衡量指標值進行數學模型配適度之選擇(步驟S140)。將最配適模型進行可靠度概率推估,其中區間數據資料為界定時間區間內的產品失效存活數據(步驟S150)。步驟S130包括以下步驟:以最大概似估計法建立概似估計函數(步驟S132)與以差分演算法對概似估計函數進行參數估計,以找到最適合市場回 饋數據之適當模型(步驟S134)。 Please refer to FIG. 1 and FIG. 2. FIG. 1 is a flowchart of a reliability analysis method according to an exemplary embodiment of the present invention. 2 is another flow chart of a reliability analysis method according to an exemplary embodiment of the present invention. As shown in FIG. 1 and FIG. 2, the reliability analysis method 100 based on market feedback data includes the following steps: collecting market product feedback data (step S110). The product failure survival data is collated and established (step S120). The section data data is input to the analysis software to perform the mathematical model adaptation (step S130). The selection of the mathematical model fit is performed using the metric value (step S140). The most suitable model is subjected to reliability probability estimation, wherein the interval data data is product failure survival data within a defined time interval (step S150). Step S130 includes the steps of: establishing a generalized estimation function by a most approximate likelihood estimation method (step S132) and parameter estimation of the approximate estimation function by a differential algorithm to find the most suitable market return. An appropriate model of the data is fed (step S134).
本發明提供一種大數據下市場回饋資料可靠度分析模式,使得企業能有效使用產品的市場回饋資料,進而進行可靠度解析,以供產品設計改良開發。 The invention provides a reliability analysis mode of market feedback data under big data, so that the enterprise can effectively use the market feedback data of the product, and then perform reliability analysis for product design improvement development.
本發明研究開發出一套市場回饋資料分析軟體,且具有相當人性化的介面,使得企業在獲得產品投入市場所積累的相關資料後,能快速且容易地進行產品可靠度分析。 The invention researches and develops a set of market feedback data analysis software, and has a quite humanized interface, so that the enterprise can quickly and easily perform product reliability analysis after obtaining relevant data accumulated by the product into the market.
承上,當產品投入市場一段時間後,產品經過消費者使用,會發生故障或損耗的現象。此時,透過收集市場上產品的存活數量及故障數量的相關數據,運用本發明之可靠度分析方法,將會得到產品之故障機率,並用以預估未來產品可能發生故障之數量,用以輔助業主評估未來需要準備多少維修零件或產品庫存,進行市場售後服務行為。 According to the above, when the product is put into the market for a period of time, the product will be used by consumers, and failure or loss will occur. At this time, by collecting data on the number of surviving products and the number of faults in the market, using the reliability analysis method of the present invention, the probability of failure of the product will be obtained, and the number of possible failures of the product in the future can be estimated to assist The owner evaluates how many repair parts or product inventories need to be prepared in the future to conduct market after-sales activities.
請同時參照圖1至圖3,圖3為根據本發明例示性實施例所繪示之產品失效之時間區間之示意圖。當產品投入市場時,經過消費者使用,會出現產品發生故障或磨損之現象。此時就必須透過維修或更換的行為來保持產品/系統的運作。然而,在某些情況下,導致產品發生故障的事件是無法確定的,因此,事件的發生是可能發生在任何時間區間內的。從圖3可知,因為產品X1至X5發生失效的精確時間無法得知,但是可以獲得的資訊是產品發生故障的時間會在哪一個時間區間內,通常這種數據會發生在不需要對產品進行時時刻刻監控的對象上,這也比較符合實務上之作法。 Please refer to FIG. 1 to FIG. 3 simultaneously. FIG. 3 is a schematic diagram showing a time interval of product failure according to an exemplary embodiment of the present invention. When the product is put on the market, after the consumer uses it, the product may malfunction or wear out. At this point, the operation of the product/system must be maintained through repair or replacement. However, in some cases, the event that caused the product to fail is undetermined, so the occurrence of the event may occur in any time interval. As can be seen from Figure 3, because the exact time of failure of products X1 to X5 is unknown, the information available is in which time interval the product fails. Usually, this data will occur without the need for the product. On the subject of constant monitoring, this is also in line with practical practices.
由於在進行產品可靠度預估前,必須透過建立數學 模型,才可進行估算,且此數據是基於區間資料(Interval censored data)的型態,故將會透過建立區間資料型態的數學模型,透過最大概似估計法(Maximum Likelihood Method,MLE)建構概似估計函數(Likelihood Function),並運用演算法(Evolution Algorithms)的方式估算出概似函數之參數,進而推估產品的可靠度資訊。 Because mathematics must be established before product reliability predictions are made The model can only be estimated, and this data is based on the type of Interval censored data, so it will be constructed by the mathematical model of the interval data type, through the Maximum Likelihood Method (MLE). It approximates the Likelihood Function and uses the Evolution Algorithms to estimate the parameters of the approximate function and then evaluate the reliability information of the product.
本發明之分析模式是基於區間資料(Interval censored data)的型態下所進行的,跟傳統上區間資料型態下的分析模式有很大的差異,其差異點在於傳統的Interval censored data研究皆在於非常小樣本的資料下所進行的分析,然而,本發明是處於大量數據的基礎下所進行的分析,在模型上因數據量的大小差異,會產生不同類型的分析模式。 The analysis mode of the present invention is based on the type of Interval censored data, which is quite different from the traditional analysis mode of the interval data type, and the difference lies in the traditional Interval censored data research. The analysis is carried out under very small sample data. However, the present invention is an analysis performed on the basis of a large amount of data, and different types of analysis modes are generated on the model due to the difference in the amount of data.
在實務上,產品投入市場的數量通常很大,導致傳統的分析模式不適用於大量數據的基礎下,故本發明提供一個大數據基礎下的可靠度分析模式,也比較符合實務上之應用。 In practice, the quantity of products put into the market is usually large, which leads to the traditional analysis mode not applicable to a large amount of data. Therefore, the present invention provides a reliability analysis mode under the big data basis, and is also more practical in practical applications.
據此,本發明針對產品市場回饋資料,開發分析軟體,透過此分析軟體,將可以提供業主如何並有效得使用產品的市場回饋資料進行可靠度分析,進而獲得產品的市場真實使用資訊,以供產品設計開發改良。 Accordingly, the present invention develops analysis software for product market feedback data, and through the analysis software, can provide the market feedback information of how the owner can effectively use the product for reliability analysis, thereby obtaining real market use information of the product for providing Product design development and improvement.
接下來要教示的,是進一步說明基於市場回饋資料之可靠度分析方法的運作機制。 What is to be taught next is to further explain the operation mechanism of the reliability analysis method based on market feedback data.
以下為基本統計模型,在此以韋伯(Weibull)分配舉例說明。 The following is a basic statistical model, which is illustrated here by the Weibull allocation.
韋伯分配(Weibull Distribution),它是在可靠度工程中最為廣泛使用的概率分配模型,在工業應用領域中,Weibull分 配常常在產品壽命試驗中使用,並且可以很靈活地的透過參數的適當選擇,來模擬出許多不同類型故障機率的行為。此外,它也提供了一個相當準確的分析和預測結果。其Weibull分配的可靠度函數與概率密度函數和下: Weibull Distribution, which is the most widely used probability distribution model in reliability engineering. In industrial applications, Weibull distribution is often used in product life testing and can be flexibly adapted through appropriate parameters. To simulate the behavior of many different types of failures. In addition, it provides a fairly accurate analysis and prediction. The reliability function and probability density function of its Weibull distribution are as follows:
其中,f(.)為概率密度函數,R(.)為可靠度函數,α為尺度參數,β為形狀參數。 Where f(.) is the probability density function, R(.) is the reliability function, α is the scale parameter, and β is the shape parameter.
然而,由於數據型態為Interval censored data,以下為區間資料(Interval censored data)形態下的概似估計函數說明: However, since the data type is Interval censored data, the following is an approximate estimated function description in the form of Interval censored data:
其中,r為產品故障數量,m為產品存活數量,l為時間區間,f(.)為概率密度函數,R(.)為可靠度函數。 Where r is the number of product failures, m is the number of product survivors, l is the time interval, f(.) is the probability density function, and R(.) is the reliability function.
承上,將Weibull分配模型建構成區間資料型式,其Weibull分配在區間資料形態下的概似估計函數模型如下: In the paper, the Weibull distribution model is constructed into an interval data type, and the approximate estimation function model of the Weibull distribution in the interval data form is as follows:
Lognormal分配模型建構成區間資料型式,其Lognormal分配在區間資料形態下的概似估計函數模型如下: The Lognormal distribution model is constructed into an interval data type, and the approximate estimation function model of Lognormal distribution in the interval data form is as follows:
Exponential分配模型建構成區間資料型式,其Exponential分配在區間資料形態下的概似估計函數模型如下: The Exponential distribution model is constructed into an interval data type. The approximate estimation function model of the Exponential distribution in the interval data form is as follows:
Log-logistic分配模型建構成區間資料型式,其Log-logistic分配在區間資料形態下的概似估計函數模型如下: The Log-logistic allocation model is constructed into an interval data type. The approximate estimation function model of Log-logistic distribution in the interval data form is as follows:
Burr XII分配模型建構成區間資料型式,其Burr XII分配在區間資料形態下的概似估計函數模型如下: The Burr XII distribution model is constructed into an interval data type. The approximate estimation function model of Burr XII distribution in the interval data form is as follows:
承上,當概似函數模型建構完畢時,接下來必須對函數進行參數估計,並找到最適合此市場回饋數據之適當模型, 本發明將透過演算法(Evolution Algorithms)的方式估算出概似函數之參數,在此以差分進化演算法(Differential Evolution,DE)舉例說明。 In conclusion, when the approximate function model is constructed, it is necessary to estimate the parameters of the function and find the appropriate model that is most suitable for the feedback data of this market. The present invention estimates the parameters of the approximate function by means of Evolution Algorithms, which is exemplified by a differential evolution algorithm (DE).
差分進化演算法(DE)類似遺傳演算法(Genetic Algorithm,GA),包含變異,交叉操作,淘汰機制,而差分進化算法與遺傳算法不同之處,在於變異的部分是隨選兩個解成員變數的差異,經過伸縮後加入當前解成員的變數上,因此差分進化算法無須使用機率分布產生下一代解成員。演算步驟及說明如下:初始化:隨機初始化參數,並計算適應值。 The differential evolution algorithm (DE) is similar to the Genetic Algorithm (GA), which includes mutation, crossover, and elimination mechanisms. The difference between the differential evolution algorithm and the genetic algorithm is that the part of the mutation is the two solution member variables. The difference is added to the variable of the current solution member after scaling, so the differential evolution algorithm does not need to use the probability distribution to generate the next-generation solution member. The calculation steps and explanations are as follows: Initialization: Randomly initialize parameters and calculate the fitness value.
突變:進行突變之運算,隨機選取數個變數向量,進行突變,產生合成向量。 Mutation: Perform the mutation operation, randomly select several variable vectors, and perform mutation to generate a synthetic vector.
交叉:以合成向量及目標向量進行重組,並在重組之後產生試驗向量,並將目標向量和試驗向量比較,並計算適應值。 Crossover: Recombination is performed with the composite vector and the target vector, and after the recombination, a test vector is generated, and the target vector is compared with the test vector, and the fitness value is calculated.
選擇:適應值較佳的一個會被當成下一個迭代的目標向量。 Choice: The one with the better fitness value will be treated as the target vector for the next iteration.
最後若未能達到停止條件則回到突變的步驟,若達滿足條件則輸出最佳解。 Finally, if the stop condition is not reached, the step of returning to the mutation is returned, and if the condition is satisfied, the optimal solution is output.
當透過差分進化演算法估算出各個模型的最適參數時,接下來透過赤井信息法則(Akaike Information Criterion,AIC)進行數學模型配適選擇,AIC值通常用來衡量數學模型之間哪個模型較適合用於此研究數據。其模型說明如下:AIC=-2log(L)+2.k When the optimal parameters of each model are estimated by differential evolution algorithm, the mathematical model is selected through Akaike Information Criterion (AIC). The AIC value is usually used to measure which model is more suitable between mathematical models. Study the data here. The model is described as follows: AIC = -2log( L ) + 2. k
其中log(L)為概似估計值,k為模型中的參數數量, AIC最小值的數學模型表示為模型最適配此數據,表示應該選擇此模型進行可靠度分析。 Where log(L) is the approximate estimate and k is the number of parameters in the model. The mathematical model of the AIC minimum is expressed as the model best fits this data, indicating that this model should be selected for reliability analysis.
接下來,將以一例子來詳細說明本揭露內容。 Next, the disclosure will be described in detail by way of an example.
請同時參照圖4、圖5與圖6,圖4為根據本發明例示性實施例所繪示之區間數據資料輸入介面之示意圖。圖5為根據本發明例示性實施例所繪示之數學模型進行模型配適之示意圖。圖6為根據本發明例示性實施例所繪示之計算出產品的可靠度與失效率之曲線圖。 Please refer to FIG. 4, FIG. 5 and FIG. 6. FIG. 4 is a schematic diagram of an interval data data input interface according to an exemplary embodiment of the present invention. FIG. 5 is a schematic diagram of a mathematical model for model adaptation according to an exemplary embodiment of the present invention. FIG. 6 is a graph of calculating reliability and failure rate of a product according to an exemplary embodiment of the present invention.
首先:透過蒐集特定產品市場回饋數據(表一)、以鴻海科技股份有限公司之某產品為例,某產品從2013生產至2015年9月共生產了4109件產品,其中2013年1-3月間生產了150件,而這150件分別在出貨第270-360天內故障回收1件、在360-450天內故障2件…。 First of all, by collecting specific product market feedback data (Table 1) and taking a product of Hon Hai Technology Co., Ltd. as an example, a product produced 4109 products from 2013 production to September 2015, including January-March 2013. 150 pieces were produced, and these 150 pieces were recovered one piece in the 270-360 days of shipment and two pieces in the 360-450 days.
接下來:整理並建立失效存活數量(表二),其中產品在0-90天內共存活295件,失效故障9件…。 Next: Finish and establish the number of failed survivors (Table 2), in which the product survived a total of 295 pieces in 0-90 days, 9 failure failures...
接下來:將區間數據資料輸入至已自行研發軟體中(如圖4所示),透過Exponential、Loglogistic、Lognormal、Weibull、Burr XII等數學模型進行模型配適(如圖5所示)。 Next: Input the interval data into the self-developed software (as shown in Figure 4), and adapt the model through mathematical models such as Exponential, Loglogistic, Lognormal, Weibull, Burr XII (as shown in Figure 5).
接下來:選取衡量指標(AIC)值最小之模型,此模型表示數據最配適之模型。 Next: Select the model with the lowest AIC value, which represents the model with the most appropriate data.
接下來:最後透過研製之軟體將可以計算出產品的可靠度與失效率,並繪製圖形(如圖6所示),此時將可以透過失效率預估此產品未來可能會故障的數目。 Next: Finally, through the development of the software, you can calculate the reliability and failure rate of the product, and draw a graph (as shown in Figure 6). At this time, the number of possible failures of the product in the future can be estimated through the failure rate.
綜上所述,本發明所提供之基於市場回饋資料之可靠度分析方法至少具有以下優點:使得企業能有效使用產品的市場回饋資料,進而進行可靠度解析,以供產品設計改良開發;以及具有相當人性化的介面,使得企業在獲得產品投入市場所積累的相關資料後,能快速且容易地進行產品可靠度分析。 In summary, the reliability analysis method based on the market feedback data provided by the present invention has at least the following advantages: enabling the enterprise to effectively use the market feedback data of the product, and then performing reliability analysis for product design improvement and development; The user-friendly interface enables companies to quickly and easily analyze product reliability after obtaining relevant data accumulated by the product.
惟上述各實施例係用以說明本發明之特點,其目的在使熟習該技術者能瞭解本發明之內容並據以實施,而非限定本發明之專利範圍,故凡其他未脫離本發明所揭示之精神而完成之等效修飾或修改,仍應包含在以下所述之申請專利範圍中。 The embodiments are described to illustrate the features of the present invention, and the purpose of the present invention is to enable those skilled in the art to understand the present invention and to implement the present invention without limiting the scope of the present invention. Equivalent modifications or modifications made by the spirit of the disclosure should still be included in the scope of the claims described below.

Claims (10)

  1. 一種基於市場回饋資料之可靠度分析方法,適用於一處理器與一顯示器,包括:藉由該處理器接收一產品失效存活數據的一區間數據資料(interval censored data),並對該區間數據資料進行一數學模型之配適,其中該產品失效存活數據關聯於一市場產品回饋資料;藉由該處理器基於一衡量指標值從多個特定數學模型中選擇配適於該區間數據資料的該數學模型;藉由該處理器基於最配適於該區間數據資料的該數學模型推估一產品的可靠度概率;以及藉由該顯示器顯示該產品的可靠度概率,其中該區間數據資料為所界定的一時間區間內的該產品失效存活數據。 A reliability analysis method based on market feedback data, which is applicable to a processor and a display, comprising: receiving, by the processor, an interval censored data of a product failure survival data, and the interval data data Performing a mathematical model adaptation, wherein the product failure survival data is associated with a market product feedback data; and the processor selects the mathematics suitable for the interval data from a plurality of specific mathematical models based on a metric value a model; estimating, by the processor, a reliability probability of a product based on the mathematical model most suitable for the data of the interval; and displaying, by the display, a reliability probability of the product, wherein the interval data is defined The product fails survival data for a time interval.
  2. 如請求項第1項所述之可靠度分析方法,更包括以下步驟:以最大概似估計法建立一概似估計函數。 The reliability analysis method described in Item 1 of the claim further includes the following steps: establishing a generalized estimation function by using the most approximate likelihood estimation method.
  3. 如請求項第2項所述之可靠度分析方法,更包括以下步驟:以一差分演算法對該概似估計函數進行參數估計,以找到最適合市場回饋數據之適當模型。 The reliability analysis method according to Item 2 of the claim further includes the following steps: parameter estimation of the approximate estimation function by a differential algorithm to find an appropriate model that is most suitable for market feedback data.
  4. 如請求項第3項所述之可靠度分析方法,其中該差分演算法為一差分進化演算法(Differential Evolution Algorithm,DEA)。 The reliability analysis method according to claim 3, wherein the difference algorithm is a differential evolution algorithm (DEA).
  5. 如請求項第1項所述之可靠度分析方法,其中透過赤井信息法則(Akaike Information Criterion)來進行該數學模型的配適選擇。 The reliability analysis method according to Item 1 of the claim, wherein the mathematical model is adaptively selected by Akaike Information Criterion.
  6. 如請求項第2項所述之可靠度分析方法,其中該數學模型為一韋伯分配模型,該韋伯分配模型在區間資料型態下為該概似估計函數。 The reliability analysis method according to claim 2, wherein the mathematical model is a Weber distribution model, and the Weber distribution model is the approximate estimation function in the interval data type.
  7. 如請求項第2項所述之可靠度分析方法,其中該數學模型為一Lognormal分配模型,該Lognormal分配模型在區間資料型態下為該概似估計函數。 The reliability analysis method according to claim 2, wherein the mathematical model is a Lognormal allocation model, and the Lognormal distribution model is the approximate estimation function in the interval data type.
  8. 如請求項第2項所述之可靠度分析方法,其中該數學模型為一Exponential分配模型,該Exponential分配模型在區間資料型態下為該概似估計函數。 The reliability analysis method according to claim 2, wherein the mathematical model is an Exponential allocation model, and the Exponential allocation model is the approximate estimation function in the interval data type.
  9. 如請求項第2項所述之可靠度分析方法,其中該數學模型為一Log-logistic分配模型,該Log-logistic分配模型在區間資料型態下為該概似估計函數。 The reliability analysis method according to claim 2, wherein the mathematical model is a Log-logistic distribution model, and the Log-logistic distribution model is the approximate estimation function in the interval data type.
  10. 如請求項第2項所述之可靠度分析方法,其中該數學模型為一Burr XII分配模型,該Burr XII分配模型在區間資料型態下為該概似估計函數。 The reliability analysis method according to claim 2, wherein the mathematical model is a Burr XII distribution model, and the Burr XII distribution model is the approximate estimation function in the interval data type.
TW105123631A 2016-07-26 2016-07-26 Reliability analysis method based on market feedback data TWI663555B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
TW105123631A TWI663555B (en) 2016-07-26 2016-07-26 Reliability analysis method based on market feedback data

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
TW105123631A TWI663555B (en) 2016-07-26 2016-07-26 Reliability analysis method based on market feedback data

Publications (2)

Publication Number Publication Date
TW201804378A TW201804378A (en) 2018-02-01
TWI663555B true TWI663555B (en) 2019-06-21

Family

ID=62014303

Family Applications (1)

Application Number Title Priority Date Filing Date
TW105123631A TWI663555B (en) 2016-07-26 2016-07-26 Reliability analysis method based on market feedback data

Country Status (1)

Country Link
TW (1) TWI663555B (en)

Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103995963A (en) * 2014-05-09 2014-08-20 卢申林 Calculation method for product reliability
CN104459560A (en) * 2014-12-29 2015-03-25 合肥工业大学 Method for predicting remaining service life of lithium battery based on wavelet denoising and relevant vector machine

Patent Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103995963A (en) * 2014-05-09 2014-08-20 卢申林 Calculation method for product reliability
CN104459560A (en) * 2014-12-29 2015-03-25 合肥工业大学 Method for predicting remaining service life of lithium battery based on wavelet denoising and relevant vector machine

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
Tieling Zhang, Min Xie. Failure Data Analysis with Extended Weibull Distribution, http://dx.doi.org/10.1080/03610910701236081,2007年5月7日 *
Tieling Zhang, Min Xie. Failure Data Analysis with Extended Weibull Distribution, http://dx.doi.org/10.1080/03610910701236081,2007年5月7日。

Also Published As

Publication number Publication date
TW201804378A (en) 2018-02-01

Similar Documents

Publication Publication Date Title
RU2522037C2 (en) Aircraft engine fault identification
Ghasemi et al. Parameter estimation methods for condition-based maintenance with indirect observations
US8560279B2 (en) Method of determining the influence of a variable in a phenomenon
Li et al. Reliability modeling of a series system with correlated or dependent component degradation processes
Gitzel Data Quality in Time Series Data: An Experience Report.
Cardoso et al. A kernel density estimate-based approach to component goodness modeling
Renganathan et al. An observer based approach for achieving fault diagnosis and fault tolerant control of systems modeled as hybrid Petri nets
Wang et al. New control charts for monitoring the Weibull percentiles under complete data and Type‐II censoring
Nguyen et al. Evolving reusable operation-based due-date assignment models for job shop scheduling with genetic programming
TWI663555B (en) Reliability analysis method based on market feedback data
US11003518B2 (en) Component failure prediction
Jain et al. Software reliability growth model (SRGM) with imperfect debugging, fault reduction factor and multiple change-point
Kaneishi et al. Software reliability modeling and evaluation under incomplete knowledge on fault distribution
Huynh et al. Degradation modeling and RUL estimation of deteriorating systems in s-plane
JP6711323B2 (en) Abnormal state diagnosis method and abnormal state diagnosis device
CN110197288A (en) The remaining life prediction technique of equipment under the influence of failure
JP6702297B2 (en) Abnormal state diagnosis method and abnormal state diagnosis device
John et al. Application of multistage process control methodology for software quality management
Lee Inferences on the lifetime performance index for Weibull distribution based on censored observations using the max p-value method
Daneshkhah et al. Prior elicitation and evaluation of imprecise judgements for Bayesian analysis of system reliability
Umar Software testing defect prediction model-a practical approach
CN111373426A (en) Vendor selection
Asraful Haque et al. A Logistic Growth Model for Software Reliability Estimation Considering Uncertain Factors
Rana et al. Comparing between maximum likelihood estimator and non-linear regression estimation procedures for NHPP software reliability growth modelling
WO2020230324A1 (en) Learning device, learning method, and learning program