数据分类算法+故障诊断马田系统.rar

864 IEEE TRANSACTIONS ON RELIABILITY, VOL. 60, NO. 4, DECEMBER 2011

Mahalanobis-Taguchi System as a Multi-Sensor

Based Decision Making Prognostics Tool for

Centrifugal Pump Failures

Ahmet Soylemezoglu, Sarangapani Jagannathan, Senior Member, IEEE, and Can Saygin

Abstract—A novel Mahalanobis Taguchi System (MTS) based

fault detection, isolation, and prognostics scheme is presented. The

proposed scheme fuses data from multiple sensors into a single

system level performance metric using Mahalanobis Distance

validated via experiments performed on a mono-block centrifugal

water pump testbed. The pump has been instrumented with

vibration, pressure, temperature, and flow sensors; and exper-

iments involving healthy and various types of faulty operating

conditions have been performed. The experiments show that the

proposed approach renders satisfactory results for centrifugal

water pump fault detection, isolation, and prognostics. Overall,

the proposed solution provides a reliable multivariate analysis and

real-time decision making tool that 1) fuses data from multiple

sensors into a single system level performance metric; 2) extends

1

, and deployed for real-time monitoring, diagnostics, and

prognostics in a wide variety of industrial environments.

Index Terms—Centrifugal pump, fault isolation and prognostics,

Mahalanobis distance-based fault detection, Mahalanobis Taguchi

system, real-time decision making.

ACRONYMS

aOFAT adaptive one-factor-at-a-time

CFD computational fluid dynamics

neering, Missouri University of Science and Technology, Rolla, MO 65409 USA

(e-mail: sarangap@mst.edu).

Texas San Antonio, San Antonio, TX 78249 USA (e-mail: can.saygin@utsa.

edu).

at http://ieeexplore.ieee.org.

observation

standard deviation of the normalized values

correlation matrix

inverse of the correlation matrix

Mahalanobis Distance for the th observation

number of characteristics

signal-to-noise ratio for the th run of the OA

number of abnormalities under consideration

SOYLEMEZOGLU et al.: MTS AS A MULTISENSOR BASED DECISION MAKING PROGNOSTICS TOOL 865

riety of process industries, such as chemical processing

plants, food service, electric utility, and automotive companies.

Centrifugal pumps are often critical components in the entire

production and process chain [1].

Vigorous Energy Efficiency program of the European Commis-

sion, pumps annually consume 117 Terawatt-hours of electricity

in Europe alone [2]. Unexpected failure of centrifugal pumps

can become costly due to the interruption in processes, and fur-

any defects or impending failures, while “abnormal operation”

refers to a pump running with defects that will eventually lead

to failures.”

can result in problems such as abnormal noise, leakage, drop

in hydraulic performance, structural vibration, and damage on

the pump components by corrosion and pitting. Therefore, a

variety of techniques that strive to maximize the availability

of pumps have been developed and presented in the literature.

Based on the literature survey regarding centrifugal pumps

or quantitative model based techniques, and 2) using process

history based models. Table I summarizes some of the various

types of failure modes, and the sensors that can be utilized to

identify these failures; whereas Table II shows the detection

techniques.

In their paper, Byskov et al. [3] investigate the flow field in a

centrifugal pump impeller at design load and quarter-load using

large eddy simulation (LES). They compare velocities predicted

by the LES, and steady-state Reynolds averaged Navier-Stokes

(RANS) simulations based on the Baldwin-Lomax, and Chien

sional curvature blade geometry in cavitating, and non-cav-

itating conditions using different experimental techniques,

and a three-dimensional numerical model. They compare

experimental and numerical results under various flow condi-

tions. Their results show the ability of the developed model

to simulate complex three dimensional cavitation in rotating

866 IEEE TRANSACTIONS ON RELIABILITY, VOL. 60, NO. 4, DECEMBER 2011

TABLE I

S

SOYLEMEZOGLU et al.: MTS AS A MULTISENSOR BASED DECISION MAKING PROGNOSTICS TOOL 867

rough set rules using experimental data. The performance of

the decision tree and the rough set are compared. They also

evaluate the classification accuracy of a principal component

analysis based decision tree-fuzzy system. The results show

that the performance of the decision tree-fuzzy hybrid system

performs better than the rough set-fuzzy hybrid system.

fault diagnosis of rotary machines. The fault diagnosis system is

based on a series of standard fault pattern pairings between fault

symptoms and faults. The fuzzy neural network model is trained

machines.

As it can be seen from the literature survey, most authors

report success in condition monitoring and fault diagnosis

of pumps using various quantitative model-based, qualitative

model-based, and process history based methods. However,

to the best of our knowledge, a method which can also suc-

cessfully predict the remaining useful life of a pump or its

components in addition to fault diagnosis is not presented.

In this paper, a Mahalanobis-Taguchi System (MTS) based

prognostics tool for centrifugal pumps is developed.

accuracy of predictions, and to optimize the system [16]. MD,

introduced by P.C. Mahalanobis in 1936 [17], is a multivariate

generalized measure used to determine the distance of a data

point to the mean of a group. MD is measured in terms of the

standard deviations from the mean of the samples, and provides

a statistical measure of how well the unknown data set matches

with the ideal one. The advantage of the MD is that it is sensitive

to the intervariable changes in the reference data. Therefore, it

has traditionally been used to classify observations into different

groups and diagnoses [18].

• Implementation of MTS requires limited knowledge of sta-

tistics.

• It relies typically on simple arithmetic, contextual knowl-

edge, and intuition.

• Its success has been demonstrated in various practical ap-

plications.

In their paper, Chinnam et al. [19] use ten features derived

from two degradation signals (thrust force, and torque) in a

drilling operation to determine the condition of the drilling tool.

The authors utilize MD to generate a measurement scale, and

to display the effectiveness of MTS in data classification. They

868 IEEE TRANSACTIONS ON RELIABILITY, VOL. 60, NO. 4, DECEMBER 2011

using the MTS, determining the key parameters. The authors

are able to show that the MTS methodology can be used to eval-

uate consumer satisfaction, and manufacturers can utilize this

knowledge to make the appropriate design decisions regarding

attributes that are delivered in a product.

Cudney et al. [22] further investigate the performance

data and customer satisfaction relationship using the Ma-

halanobis-Taguchi Gram-Schmidt (MTGS) technique, and

showed improved results. Foster et al. [23] develop an alterna-

tive search procedure to be used within the MTS. The adaptive

One-Factor-AT-a-Time (aOFAT) procedure replaces the or-

thogonal arrays traditionally utilized in MTS, and features are

MTS has also been used as a process control and improve-

ment tool to aid manufacturing systems. Asada [24] use MTS

to forecast the yield of a wafer production system, which is af-

fected by variability of the electrical properties, and environ-

mental dust. MDs are used to compare the relationship between

wafer yield and distance. Hayashi et al. [25] develop an MD

based method to maximize productivity in a semi-conductor

manufacturing facility. Their method consists of real-time mon-

itoring of the ratio of work-in-process inventory to number of

in-process wafers, which they call flow factor , using MD values

various sensors. Sensor data are analysed using MTS, and the

quality of the current operation is communicated back to the

user via a wireless network. The authors were able to obtain

100% data collection on each fastener, as opposed to other statis-

tical process control techniques which rely on sampling. In [27]

is presented an MTS-based methodology which detects grip

length of bolted joints in real-time during fastening. Grip length

is the length of the unthreaded portion of a bolt shaft. When the

total thickness of joining members is greater than the grip length

of the bolt, it is called under-grip, which compromises the struc-

dard deviation of various factors, including torque-over-angle

ratio, angle-over-torque ratio, total angle turned, and work done.

Then, the signature is processed using the MTS-based approach

to detect the grip-length, all of which occurs in real-time as

the fastening process is completed. The proposed approach is

also applied to detect the presence of re-used fasteners, which

is another quality concern because some material properties and

physical conditions of bolts and nuts can change if they are

reused several times.

Dasgupta [28] develop a unified framework for achieving

lean processes. Their methodology assists contemporary lean

assessment tools through providing a measure of leanness by

tions for improvements based on cost considerations.

MTS is a data driven method that relies on multivariate

statistical analysis. Therefore, it is important to compare MTS

to well-established multivariate statistical techniques such as

principal component analysis, discriminant analysis, cluster

analysis, canonical component analysis, and multidimensional

scaling, as well as artificial neural networks. For an in-depth

comparison of these techniques to MTS, please refer to Kim et

a new technique. However, the main contribution of this study

is the combination of MTS-based clustering with prognostics

and its application to pump failures, which has not been studied

in the literature. By combining these methods, the benefits of

MTS have been expanded over to the prognostics field. More

specifically, this paper 1) uses MTS to fuse multi-sensor infor-

mation into a single system performance metric; 2) builds upon

MTS by introducing the use of MD-based fault clusters for fault

isolation; 3) utilizes the progression of MD values over time to

estimate the remaining useful life of components (i.e., time to