Computer-implemented method and system for machine tool damage assessment, prediction, and planning in manufacturing shop floor
A self-aware machine platform is implemented through analyzing operational data of machining tools to achieve machine tool damage assessment, prediction and planning in manufacturing shop floor. Machining processes are first identified by matching similar processes through an ICP algorithm. Machining processes are further clustered by Hotelling's T-squared statistics. Degradation of the machining tool is detected through a trend of the operational data within a cluster of machining processes by a monotonicity test, and the remaining useful life of the machining tool is predicted through a particle filter by extrapolating the trend under a first-order Markov process. In addition, process anomalies across machines are detected through a combination of outlier detection methods including SOMs, multivariate regression, and robust Mahalanobis distance. Warnings and recommendations are flexibly provided to manufacturing shop floor based on policy choice.
1. A computer-implemented system for detecting machine tool wear, predicting machine tool failure, and manufacturing shop floor planning, comprising:
a computing device comprising a processor and configured to:
obtain a machine tool's operational data comprising positional parameters and movement parameters during a time window;
identify a plurality of machining processes of the machine tool based upon a match of the positional parameters and the movement parameters that defines the machining processes;
cluster the plurality of the machining processes into one or more process clusters, based upon a similarity of the machining processes;
detect the machine tool's wear by characterizing a trend of change in a parameter from the one or more clusters of machining processes performed by the machine tool; and
predict the machine tool's remaining useful life by extrapolating the trend under a first-order Markov process, wherein the machine tool is replaced based on the predicted remaining useful life.
2. A computer-implemented system according to claim 1 , wherein the operational data comprise of at least one of X-axis position, Y-axis position, Z-axis position, feed rate, spindle speed, spindle power, spindle load, and tool number.
3. A computer-implemented system according to claim 1 , the computing device further configured to:
determine a tool path shape in a 3-D space for the plurality of the machining processes by plotting X-, Y-, and Z- positions of the machine tool during associated with the processes;
determine a movement parameter shape by plotting the spindle speed, feed rate, and time of the machine tool in the parameter space for the plurality of the machining processes associated with the machine tool; and
determine the match between two machining processes from the plurality of the machining processes by matching both the tool path shapes in the 3-D space and the movement parameter shapes in the parameter space between two machining processes, via finding the smallest possible differences between the matched shapes using an ICP algorithm.
4. A computer-implemented system according to claim 1 , the computing device further configured to:
cluster the machining processes based on the similarity of the machining processes, further comprising:
obtain an initial cluster of machining processes;
calculate the difference between at least one of the plurality of the machining processes and the average of the initial cluster using a T2 method of:
T 2=( d a −{circumflex over (d)} a )* s −1 * ( d a −{circumflex over (d)} a )′
wherein d a is a measurement of the difference and calculated by d a =[d s ,d p ], d s is a difference between tool path shapes in shape spaces, d p is a difference between movement parameter shapes in parameter spaces, {circumflex over (d)} a is the mean value of the difference of the initial cluster of the processes, and s is the covariance;
determine a T2 limit by a formula of
T
2
limit
=
(
N
-
1
)
(
N
+
1
)
p
N
(
N
-
p
)
F
α
(
p
,
N
-
p
)
,
where F α (p,N−p) is the 100α% confidence level of F−distribution with p and N−p degrees of freedom;
keep the at least one process out of the cluster if the T2 statistics is above the T2 limit ; and
assign the at least one process into the initial cluster of the processes and update the centroid of the cluster if the T2 statistics is below the T2 limit .
5. A computer-implemented system according to claim 1 , the computing device further configured to:
characterize the trend of the parameter from one of the one or more clusters through detecting a unidirectional change of the parameter via a monotonicity test as described in for
Monotonicity
(
F
)
=
#
d
/
dF
>
0
n
-
1
-
#
d
/
dF
<
0
n
-
1
wherein F is the measurement of the parameter, d/dF is the derivative, and n is the number of measurement in a period of time;
set up a threshold of F ; and
provide a notification of a degradation when the threshold is exceeded by the calculated monotonicity,
wherein the parameter used for the trend detection is chosen from at least one of the following group: the machine tool's spindle power, the machine tool's torque, heat generated by the machine tool during the processes, and frictions of a bearing associated with the machine tool during the processes.
6. A computer-implemented system according to claim 1 , the computing device further configured to:
define the first-order Markov process model with a second order polynomial model of:
Xk=a k t k +b k t k 2 +c k =X k −1 +( a k +2bt k−1 ) Δt + b k Δt 2
wherein the Xk is the state of tool wear, t k is the time at step K, and a k , b k , and C k are parameters of the second order polynomial model.
7. A computer-implemented system according to claim 6 , the computing device further configured to:
notify of the remaining useful life,
wherein the machine tool is a cutting tool and the parameter is the values of the spindle powers of the processes.
8. A computer-implemented method for detecting machine tool wear, predicting machine tool failure, and manufacturing shop floor planning, comprising:
obtaining by a computing device comprising a processor a machine tool's operational data comprising positional parameters and movement parameters during a time window;
identifying by the computing device a plurality of machining processes of the machine tool based upon a match of the positional parameters and the movement parameters that defines the machining processes;
clustering by the computing device the plurality of the machining processes into one or more process clusters, based upon a similarity of the machining processes;
detecting by the computing device the machine tool's wear by characterizing a trend of change in a parameter from the one or more clusters of machining processes performed by the machine tool; and
predicting by the computing device the machine tool's remaining useful life by extrapolating the trend under a first-order Markov process, wherein the machine tool is replaced based on the predicted remaining useful life.
9. A method according to claim 8 , wherein the operational data comprise of at least one of X-axis position, Y-axis position, Z-axis position, feed rate, spindle speed, spindle power, spindle load, and tool number.
10. A method according to claim 8 , further comprising:
determining by the computing device the match of the positional parameters and the movement parameters, further comprising the steps of:
for the plurality of the machining processes associated with the machine tool, determining a tool path shape for each machining process in a 3-D space by plotting X-, Y-, and Z- positions of the machine tool during the each machining process;
for the plurality of the machining processes associated with the machine tool, determining a movement parameter shape by plotting the spindle speed, feed rate, and time of the machine tool in a parameter space for the each machining process; and
determining the match between two machining processes from the plurality of the machining processes by matching both the tool path shapes in the 3-D space and the movement parameter shapes in the parameter space between the two machining processes, via finding the smallest possible differences between the matched shapes using an ICP algorithm.
11. A method according to claim 8 , further comprising:
clustering by the computing device the machining processes based on the similarity of the machining processes, further comprising the steps of:
obtaining an initial cluster of machining processes;
for at least one of the plurality of the machining processes, calculating the difference between the at least one process and the average of the initial cluster using a T2 method of:
T 2=( d a −{circumflex over (d)} a )* s −1 *( d a −{circumflex over (d)} a )′,
wherein d a is a measurement of the difference and calculated by d a =[d s ,d p ], d s is a difference between tool path shapes in shape spaces, d p is a difference between movement parameter shapes in parameter spaces, {circumflex over (d)} a is the mean value of the difference of the initial cluster of the processes, and s is the covariance;
determining a T2 limit by a formula of
T
2
limit
=
(
N
-
1
)
(
N
+
1
)
p
N
(
N
-
p
)
F
α
(
p
,
N
-
p
)
,
where F α (p,N−p) is the 100α% confidence level of F−distribution with p and N−p degrees of freedom;
keeping the at least one process out of the cluster if the T2 statistics is above the T2 limit ; and
assigning the at least one process into the initial cluster of the processes and updating the centroid of the cluster if the T2 statistics is below the T2 limit .
12. A method according to claim 8 , further comprising:
for one of the one or more clusters, characterizing by the computing device the trend of the parameter through detecting a unidirectional change of the parameter via a monotonicity test as described in
Monotonicity
(
F
)
=
#
d
/
dF
>
0
n
-
1
-
#
d
/
dF
<
0
n
-
1
wherein F is the measurement of the parameter, d/dF is the derivative, and n is the number of measurement in a period of time;
setting up by the computing device a threshold of F ; and
notifying by the computing device of a degradation when the threshold is exceeded by the calculated monotonicity,
wherein the parameter used for the trend detection is chosen from at least one of the following group: the machine tool's spindle power, the machine tool's torque, heat generated by the machine tool during the processes, and frictions of a bearing associated with the machine tool during the processes.
13. A method according to claim 8 , further comprising:
defining by the computing device the first-order Markov process model with a second order polynomial model of:
Xk=a k t k +b k t k 2 +c k =X k−1 +( a k +2 bt k−1 ) Δt+b k Δt 2
wherein the Xk is the state of tool wear, t k is the time at step K, and a k , b k , and c k are parameters of the second order polynomial model.
14. A method according to claim 13 , further comprising:
notifying by the computing device of the remaining useful life,
wherein the machine tool is a cutting tool and the parameter is the values of the spindle powers of the processes.
15. A method according to claim 8 , further comprising:
providing by the computing device a recommendation on scheduling a machine maintenance by comparing a performance history of the machine to another machine performing comparable machining process, comprising:
comparing a histogram of a spindle speed measured by RPM weighted by spindle load at the speed for the one machine to the other; and
comparing the total traverse of across all feed axes on the one machine to the other.