Data processing: generic control systems or specific application – Generic control system – apparatus or process – Optimization or adaptive control
Reexamination Certificate
2001-03-21
2004-10-12
Voeltz, Emanuel Todd (Department: 2121)
Data processing: generic control systems or specific application
Generic control system, apparatus or process
Optimization or adaptive control
C700S108000, C700S109000, C700S121000, C702S179000
Reexamination Certificate
active
06804563
ABSTRACT:
The present invention relates to a method and any system using such a method of statistical process control based on multidimensional processing of data.
Its purpose is firstly to trigger a warning when the process departs from “normal” operation which ensures that its production is of the required quality, and secondly to make proposals for identifying the probable cause(s) of the anomaly.
Statistical process control (SPC) is presently in use in a very large number of businesses, in all countries (mainly industrialized countries), for all types of industrial production: engineering, electronics, chemistry, pharmaceuticals, agri-food, plastics materials, . . . .
Its purpose is to ensure product quality by inspecting the manufacturing process itself and not only by inspecting the characteristics of its products. SPC has become essential in achieving “zero defects” and when the business seeks to comply with international quality assurance standards (ISO 9000).
Its technical objective is to detect possible drift in the manufacturing process and to remedy it before non-compliant products are manufactured.
The use of this method has now extended beyond the context of manufacturing goods and covers producing services (banking, insurance, consultancy, . . . ).
When running a process (cf. FIG.
1
), various measurements (indicators) associated with the same process are tracked: input characteristics (raw materials); output characteristics (products); process operating parameters. Each unit of observation (measurement instant or element produced) is thus associated with a plurality of digital values obtained by the measurements, thus enabling it to be represented by a point in the multidimensional space of the measurements taken.
The usual practice in SPC consists in monitoring the process by tracking a plurality of control charts which are graphical representations of the way an observed magnitude varies and which present predefined control limits (see FIG.
2
), one per measurement. Each control chart is then interpreted independently of the others, triggering warnings independently.
Various types of control chart are available (known as “Shewart, CuSum, EWMA, MME”), with the last three being accepted as being better at detecting small amounts of “drift” than the first.
Generally, control charts are used on grouped data: by plotting the averages of a plurality of grouped-together measurements, small amounts of drift are detected better, and in addition the distribution of the values coincides better with the assumption of normality that underlies the method. By plotting the variances or the extents of each group, it is possible to detect an increase in measurement variability that has some special cause.
The usual practice which consists in simultaneously and independently monitoring a plurality of control charts constitutes a method that is clumsy and not very effective in multidimensional SPC:
it raises too many false warnings which can give rise to unnecessary corrections; these then need to be reassessed very quickly, and lead to the process being controlled in a manner that is chaotic and expensive, with multiple corrections;
it can detect real anomalies too late; and
it has difficulty in detecting the causes of anomalies when they are not directly associated with a measurement. This encourages taking a multiplicity of measurements which is expensive and leads to a multiplicity of control charts.
The method and system of the present invention seek to mitigate those drawbacks: the method is one of statistical process control on the basis of taking indicators or measurements on inputs, on outputs, and on control and operating parameters of said process, and which can be represented by observation points in frames of reference that associate their values with their sampling indices; according to the invention:
a) the observed values are transformed so that the resulting values are compatible with the multidimensional Gaussian distribution model, and constitute data corresponding to the observation points used in the remainder of the method;
b) said observation points are situated in a multidimensional space, in which each dimension is associated with a measured magnitude;
c) amongst the observation points, points that are said to be “under control” and that correspond to proper operation of the process are distinguished from points which are said to be “out of control”;
d) the distribution center of the points under control is calculated as being the center of gravity of the observation points under control;
e) out of control observation points that are concentrated in some particular direction from the distribution center of the points under control are identified;
f) this direction is associated with a common cause for drift of said process;
g) each observation point and anomaly direction pair is associated with indicators in order to propose zero, one, or more causes of anomaly that are liable to be associated with the observation that has been made; and
h) when an anomaly is analyzed in this way, a warning is triggered and the drift detected in this way in the industrial process is remedied.
Said center of gravity of the observation points being inspected corresponds to a point whose components are the means of the components of the observation points under inspection.
SPC inspection consists in conventional manner in regularly observing p continuous magnitudes y
1
, y
2
, . . . , y
p
either statistically or by sampling. These magnitudes can equally well represent characteristics of raw materials, characteristics of manufactured products, or operating parameters of the manufacturing process. The p-dimensional vector made up of these p measurements at a given “instant” is written y and is referred to as the observation vector of the process, with the endpoint of this vector being the observation point of the process and the origin of this vector being the original of the frame of reference in question.
It is clear that in this context the concept of “instant” goes beyond a strictly temporal interpretation: measurements associated with the same “instant” are, wherever possible, measurement of parameters relating to the production of the same manufactured unit or batch. Perfect traceability of the manufacturing process is then necessary in order to be able to define which measurements are associated with the same “instant”.
When the process is “under control”, the values of y at various successive instants t
0
, t
0
+1, t
0
+2, . . . vary “little” about a value y
0
, which is the desired target for ensuring that production is of satisfactory quality. This variation is due to random variations in the characteristics of the raw materials (material hardness, chemical composition of a component, supplier, . . . ), of the environment (temperature, humidity, . . . ), or of the process (setting of a machine, attention of an operator, . . . ). These characteristics have influence over one or more components of y and they are written z
1
, z
2
, . . . , z
m
and together they form a vector written z. The vector z is referred to herein as the explanatory vector of the process.
For a characteristic of the process to be considered as an “observation variable” of the process y
j
it must be evaluated at each “instant”.
For a characteristic of the process or of the inputs to be considered as a “cause variable” z
k
of the process, it must be modified by an agent external to the system proper: voluntary or involuntary human action, variation in the environment, wear, or aging. Generally, for reasons of expense or of feasibility, these variables are not measured at each “instant” (otherwise they would also appear as variable y
j
) and in this sense they constitute “hidden variables” that influence the behavior of the process. Evaluating them is often expensive, lengthy, imprecise, and is performed only in the event of an anomaly.
A variable can be quantitative if its possible values are numerical and belong to a known range of values (temperature, pressure, . . . ), or it can be quali
Cohen & Pontani, Lieberman & Pavane
Todd Voeltz Emanuel
LandOfFree
Multidimensional method and system for statistical process... does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Multidimensional method and system for statistical process..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Multidimensional method and system for statistical process... will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-3261662