海洋数据集的质量检查验收抽样方法翻译.docx

1、海洋数据集的质量检查验收抽样方法翻译Acceptance sampling plan of quality inspection for ocean datasetCompared with the dataset of industrial products, ocean datasets have several distinct characteristics,such as large quantities and being multi-source,multi-dimension and multi-type.Based on the acceptance quality leve

2、l (AQL) and limit quality level (LQL), we designed an acceptance sampling plan of quality inspection for ocean datasets(ASP-OD),used this plan to inspect ocean dataset quality,and evaluated its advantage.ASP-OD has a consistent and stable discriminatory power in dependent of lot size which solves th

3、e problem of strictness for large lot size, toleration for small lot size in the percent sampling plan. ASP-OD establishes a relationship between lot size and sampling size,and provides a plan for a given lot size.This plan overcomes the deciency of ISO2859-based sampling plans,different lot size co

4、rresponding to the same sampling plan, in the quality inspection of ocean datasets. Collectively, this study suggests that ASP-OD is a suitable sampling plan for the inspection of ocean dataset quality.Keywords: ocean dataset;quality inspection;AQL;LQL;acceptance sampling plan1.IntroductionWith the

5、rapid development of ocean monitoring technology,huge amounts of ocean data have been collected from various sources,such as remote sensing images, buoys, cruise data and underwater observation data.Thus,ocean datasets have gradually become a classic example of multi-modal big data.However,the bigge

6、st obstacle for preparing an ocean atlas is how to control the quality of data.The quality control of ocean datasets is an important part of any ocean analysis/forecasting system.Using or accepting erroneous data could lead to an invalid conclusion or an incorrect analysis.By contrast,rejecting extr

7、eme but valid data sometimes could cause the missing of key events and anomalous features.To date,a growing number of scientists have begun to focus on the quality inspection of ocean data.An automated quality control system was proposed to inspect oceanic temperature and temperature-salinity profil

8、es. The Surface Ocean CO2 Atlas(SOCAT) project was performed to investigate the global dataset of marine surface CO2. During this project, all data were designed to be put in a uniform format following a strict protocol. Quality control was conducted according to clearly defined criteria.In addition

9、,the quality and consistency of NASA ocean colour data,including spectral water-leaving reflectance,chlorophyll- concentration, and diffuse attenuation,were examined using common algorithms and improved instrument calibration knowledge.These studies have put forward several quality inspection plans

10、for ocean data,especially for one or a few elements. Ocean datasets are usually composed of multi-element,multi-scale and multi-temporal geo-information elements.Moreover,there is a potential interplay between different elements in an ocean dataset.Thus,it is required to propose a novel acceptance s

11、ampling plan to inspect the quality of ocean data as a complete and indivisible dataset.The goal of quality inspection is to judge whether the data reach the required quality through a sampling plan.Currently, the optimisation of acceptance sampling plans has been conducted to satisfy the balance be

12、tween inspection risk and inspection cost for the quality inspection of industrial products. Some acceptance sampling plans have been designed based on inspection risk, which aimed to minimise either the producers risk or the consumers risk.Some other acceptance sampling plans were designed based on

13、 the inspection cost, which aimed to reduce the sampling number.These existing plans are mainly used to inspect the quality of industrial products.Generally,industrial products are produced in a controlled and consistent manner, and usually have certain items and uniform units.Compared with industri

14、al products,ocean data have some distinct characteristics, such as being multi-source,multi-dimensional multi-type,multi-time-state,with different accuracy and nonlinearity.Thus,these existing acceptance sampling plans are not suitable for the quality inspection of ocean datasets.In this paper, we d

15、esigned an acceptance sampling plan of quality inspection for an ocean dataset (ASP-OD). In section 2, the conceptual framework,derivation process and the formulas of ASP-OD are shown.In section 3,we apply the ASP-OD to inspect the quality of ocean data,and compare its advantages over existing accep

16、tance sampling plans.In section 4,we summarise this study,and propose that ASP-OD is a suitable acceptance sampling plan for the quality inspection of ocean datasets.2.Design of ASP-ODThe theory of ASP-ODThe acceptance sampling plan of quality inspection for ocean datasets was designed as S(N,n,c).H

17、ere,N is the lot size and comprises all inspected ocean data from which the sample is to be taken;n is the sample size and consists of a number of sampling units selected from the lot size,which is a compromise between the accuracy of product inspection and the cost of the inspection;c,the acceptanc

18、e number,is used to judge whether the inspected ocean data meet the requirement of the ocean data consumer.The process of quality inspection is shown as below:(1)n-sampled data are extracted from the lot size N;(2)the quality of extracted data is inspected one by one;(3)if the number of non-conformi

19、ng data (d) is larger than the acceptance number (c),the quality of inspected ocean data is considered to be non-conforming.Otherwise,the quality of inspected data is considered to be conforming.Based on the acceptance sampling plan S(N, n, c), the percent non-conforming (P ) is calculated by (1)whe

20、re D is the number of non-conforming ocean data in the total ocean dataset.Generally,it is difficult to obtain the values of D and P unless the total data are100 percent inspected. Sampled ocean data are used to estimate the parameters for lot size.Thus,P is usually estimated using the percent non-c

21、onforming estimator(p);p is calculated by (2)where d is the number of non-conforming ocean data in the sampled dataset.Based on the above-mentioned parameters,the acceptance quality probability(L(p)of the acceptance sampling plan S(N, n, c) can be calculated by (3)(0dn,dD,n-dN-Np）Operating character

22、istic curves (OC-curve)are powerful tool sin the field of quality control, as they display the discriminatory power of an acceptance sampling plan.Here,we considered the quality level as the horizontal axis and the corresponding acceptance probability as the vertical axis.The relationship between L(

23、p) and the proportion p of non-conforming items was represented as the OC-curve of sampling inspection in a rectangular coordinate system.Generally,considering the interests of both the producers and consumers, acceptance quality level(AQL)and limiting quality level (LQL)were adopted to design the a

24、cceptance sampling plan. LQL is a maximum quality level of defectives tolerated in the inspection data. When the quality level is worse than LQL, the consumers tend to reject the inspected data.AQL represents a mean quality level of defective samples tolerated in the inspection.If the quality level

25、of the inspected data is better than AQL,the producers tend to accept the inspected data.To meet the requirement of both producers and consumers,AQL and LQL were taken into account in the ASP-OD design, which was shown as two points in the OC-curve (Figure 1). The first point is denoted as Figure1.

26、OC-curve of the acceptance sampling plan(p0,1-)p0 , i.e. AQL, is the proportion of non-conforming items that can be tolerated to judge that the entire lot can be accepted.,the producers risk,is the probability of rejection of the inspected lot even though the quality level of the lot is equal to or

27、better than AQL.The second point is denoted as(p1,).p1,i.e.LQL,is the proportion of nonconforming items that can be tolerated to judge that the entire lot can be rejected.,the consumers risk,is the probability of acceptance of the inspected lot even though the quality level of the lot is equal to or

28、 worse than LQL.Under the condition of the two points on the OC-curve,the relationship between the lot size,the sample size and the acceptance number is calculated.The problem could be formulated as a nonlinear programming problem.The ASP-OD modelFrom the perspective of the producer, the acceptance

29、sampling plan should satisfy the following condition: (4)D1, a positive integer, is the number of non-conforming data elements in the inspected ocean dataset. When the proportion of non-conforming data is equal to AQL,the value of D1 is calculated byD1=round(Np1) (5)From the perspective of the consu

30、mer,the acceptance sampling plan should satisfy the following condition (6)D2 , a positive integer, is the number of nonconforming data elements in the inspected ocean dataset. When the proportion of nonconforming data is worse than the limiting quality level (LQL), the value of D2 is calculated byD

31、2=round(Np2) (7)The total residual error,means the sum of residual errors of the acceptance probability at both AQL and LQL.The role of is used for the calculation of the value of n and c in the acceptance sampling plan.Here,we chose the minimal to determine the optimal n and c for the acceptance sa

32、mpling plan at AQL and LQL.The optimal acceptance sampling plan is formulated as the following nonlinear optimisation problems.t. (8) (9) 1 is the residual error of the acceptance probability based on the producers risk.2 is the residual error of the acceptance probability based on the consumers risk. The nonlinear optimisation problem is solved based on the iterative algorithm.The iterative algorithm is implemented in Matlab software.3.Case studyIn this section, we employed ASP-OD,the percent sampling plan (PSP) and the ISO 2859-based sampling plan (I