ABDELMALEK ESSAADI

ENSA TANGEIR

Machine Learning in predictive maintenance

Realized by:
Supervised by:
EL YEMLAHI
Ayoub
M. Elyahyaoui Abdelouahid

EL MOUSTAMSIK
Samira

EL IDRISSI
Khaoula

2021/2022
H

Table des matières :

Table des matières :...............................................................................................................................2

Table des figures :...............................................................................................................................3
I. Work plan.......................................................................................................................................4
I. Work plan:.....................................................................................................................................5
II. Selection of articles:....................................................................................................................5
II. Article 1: Machine learning for predictive maintenance of industrial machines using IoT sensor
data. 6
I. Authors:.........................................................................................................................................7
 Ameeth Kanawaday: University of Southern California..........................................................7
 Aditya Sane: College of Engineering, Pune.............................................................................7
II. Introduction:..................................................................................................................................7
III. Internet of things:.......................................................................................................................7
Definition:...........................................................................................................................................7
industrial IoT (IIOT):............................................................................................................................7
IV. Data collection:...........................................................................................................................7
1. Machine details:.........................................................................................................................8
2. System Setup:.............................................................................................................................8
3. Data Description:........................................................................................................................8
V. Approach:.......................................................................................................................................8
1. Exploratory Data Analysis:..........................................................................................................8
2. Preprocessing in Data Mining:....................................................................................................9
a. Data Cleaning:........................................................................................................................9
b. Data Transformation:...........................................................................................................10
c. Data Reduction:....................................................................................................................10
3. Predictive Analysis:...................................................................................................................11
4. Data Preprocessing:..................................................................................................................11
5. ARIMA:.....................................................................................................................................11
VI. Result:.......................................................................................................................................12
VII. Future scope:............................................................................................................................12
III. Article 2: Machine learning approach for predictive maintenance in industry 4.0....................13
I. Resume:......................................................................................................................................14
II. Predictive maintenance:............................................................................................................14
1. Aspects of predictive maintenance:.........................................................................................14

2
 2. Approaches of PDM:.................................................................................................................14
III. The data analysis:..................................................................................................................15
IV. Machine Learning algorithms:..................................................................................................18

Table of figures:

Figure 1: Gant.........................................................................................................................................5
Figure 2:The decomposition of the time series data............................................................................11
Figure 3:ARIMA model.........................................................................................................................12
Figure 4: The machine..........................................................................................................................15
Figure 5:Features Dataset....................................................................................................................16

3
 I. Work plan

I. Work plan:

4
 Figure 1: Gant

II. Selection of articles:

 Article1: Machine learning for predictive maintenance of industrial machines using IoT
sensor data

 Article2: Machine Learning approach for Predictive Maintenance in Industry 4.0

5
II. Article 1: Machine learning for
predictive maintenance of
industrial machines using IoT
sensor data.

6
 I. Authors:
 Ameeth Kanawaday: University of Southern California
 Aditya Sane: College of Engineering, Pune

II. Introduction:
This paper explores the use of Autoregressive Integrated Moving Average (ARIMA)
forecasting on the time series data collected from various sensors from a Slitting Machine, to
predict the possible failures and quality defects, thus improving the overall manufacturing
process.

III. Internet of things:

Definition:
The internet of things: describes physical objects (or groups of such objects) that are
embedded with sensors, processing ability, software, and other technologies that connect
and exchange data with other devices and systems over the Internet or other
communications networks.

industrial IoT (IIOT):

The industrial internet of things (IIOT) refers to the extension and use of the internet of things
(IoT) in industrial sectors and applications. With a strong focus on machine-to-machine
(M2M) communication, big data, and machine learning, the IIOT enables industries and
enterprises to have better efficiency and reliability in their operations. The IIOT encompasses
industrial applications, including robotics, medical devices, and software-defined production
processes.

(Machine to machine (M2M) is direct communication between devices using any

communications channel, including wired and wireless)

IV. Data collection:

The main aim of this work is to provide prognostics for industrial machines to increase
productivity and prevent quality failures.

1. Machine details:
The data have been collected from a slitting machine with 14 arms (7 on each side) which
produce variable sized packaging rolls from a huge wound roll which is used to rewind and
cut packaging films for various customers.

After collecting data and studying the process we’ve found that:

• Tension and pressure were of higher importance in degradation of the packaging

roll.

7
 • As the roll diameter increases, tension and pressure need to be reduced to
maintain the preset rate.

2. System Setup:
PLC: A programmable logic controller or programmable controller is an industrial computer
that has been ruggedized and adapted for the control of manufacturing processes, such as
assembly lines, machines, robotic devices, or any activity that requires high reliability, ease
of programming, and process fault diagnosis.

IPC: An industrial PC is a computer intended for industrial purposes (production of goods

and services), with a form factor between a nettop and a server rack. Industrial PCs have
higher dependability and precision standards, and are generally more expensive than
consumer electronics.

For our machine:

• The PLC on the machine stores data for various parameters which need to be
monitored.
• the data is sent to the adaptor which then converts the data into TCP form which
is fed to the Industrial Personal Computer (IPC).
• The IPC is connected to the internet and pushes the data to the cloud using
MQTT protocol in the form of data packets.

3. Data Description:
• Data generated from the Slitting machine was collected using sensors and
pushed to the cloud.
• This data was sampled per second and collected for a period of one month.
• The data is stored as CSV in the system, consisting of 5 columns namely: Time
Stamp, Tension, Pressure, Width, and Diameter.

V. Approach:
In this step, we will incorporate data analysis, clustering, and supervised learning methods to
gain insights from the data.

1. Exploratory Data Analysis:

• The sensors send the data when the machine shows a change in state.
• The data received consists of many Null values.
• Start preprocessing:
• Replacing the Null values with the sliding mean values.
• A thorough analysis was carried out on the outliers as well as the failure points
which were marked in the incoming labeled data to find the differences from the
good cycles of production.

8
 • A deeper analysis was done to understand the effects of the parameters on the
product quality.

2. Preprocessing in Data Mining:

Data preprocessing is a data mining technique which is used to transform the raw data in a
useful and efficient format.

Steps Involved in Data Preprocessing:

VI.Data Cleaning:
The data can have many irrelevant and missing parts. To handle this part, data cleaning is
done. It involves handling of missing data, noisy data etc.

Missing Data:
This situation arises when some data is missing in the data. It can be handled in various
ways.

Some of them are:

 Ignore the tuples:

This approach is suitable only when the dataset we have is quite large and multiple values
are missing within a tuple.

 Fill the Missing values:

There are various ways to do this task. You can choose to fill the missing values manually,
by attribute mean or the most probable value.

Noisy Data:
Noisy data is a meaningless data that can’t be interpreted by machines. It can be generated
due to faulty data collection, data entry errors etc. It can be handled in following ways:

 Binning Method:

This method works on sorted data in order to smooth it. The whole data is divided into
segments of equal size and then various methods are performed to complete the task. Each
segmented is handled separately. One can replace all data in a segment by its mean or
boundary values can be used to complete the task.

 Regression:

Here data can be made smooth by fitting it to a regression function. The regression used
may be linear (having one independent variable) or multiple (having multiple independent
variables).

 Clustering:

This approach groups the similar data in a cluster. The outliers may be undetected or it will
fall outside the clusters.

9
 VII.Data Transformation:
This step is taken in order to transform the data in appropriate forms suitable for mining
process. This involves following ways: /.

 Normalization:

It is done in order to scale the data values in a specified range (-1.0 to 1.0 or 0.0 to 1.0)

 Attribute Selection:

In this strategy, new attributes are constructed from the given set of attributes to help the
mining process.

 Discretization:

This is done to replace the raw values of numeric attribute by interval levels or conceptual
levels.

 Concept Hierarchy Generation:

Here attributes are converted from lower level to higher level in hierarchy. For Example-The
attribute “city” can be converted to “country”.

VIII.Data Reduction:
Since data mining is a technique that is used to handle huge amount of data. While working
with huge volume of data, analysis became harder in such cases. In order to get rid of this,
we use data reduction technique. It aims to increase the storage efficiency and reduce data
storage and analysis costs.

The various steps to data reduction are:

 Data Cube Aggregation:

Aggregation operation is applied to data for the construction of the data cube.

 Attribute Subset Selection:

The highly relevant attributes should be used, rest all can be discarded. For performing
attribute selection, one can use level of significance and p- value of the attribute. The
attribute having p-value greater than significance level can be discarded.

 Numerosity Reduction:

This enables to store the model of data instead of whole data, for example: Regression
Models.

 Dimensionality Reduction:

This reduces the size of data by encoding mechanisms. It can be lossy or lossless. If after
reconstruction from compressed data, original data can be retrieved, such reduction are
called lossless reduction else it is called lossy reduction. The two effective methods of
dimensionality reduction are: Wavelet transforms and PCA (Principal Component Analysis).

3. Predictive Analysis:

10
The use of predictive analysis proves to be a viable design solution for industrial machine
prognostics.

The application:

We have used Autoregressive Integrated Moving Average (ARIMA) for forecasting the
machine parameters to map the future states of the machine.

4. Data Preprocessing:
• Receiving data from sensors.
• The data is essentially discrete-time data sampled per second of time.
• The decomposition of the time series data.
• They found an increasing trend in the residues.
• The time series was stationaries using differencing.
• They used Logarithmic transformation to reduce the variance of the time series
data.
• As a result, the rolling mean and standard deviation were made independent of
time.

Figure 2:The decomposition of the time series data

5. ARIMA:
The autoregressive integrated moving average (ARIMA) model was used to predict future
points in the data series.

Since the data showed non-stationarity, we have used ARIMA.

The model is fitted in the dataset and further points are predicted to find if the future states
which might lead to a failure.

11
 Figure 3:ARIMA model

IX. Result:
It can be inferred that:

• The deep neural network model was more efficient in modeling the data.
• The occurrences of the bad quality cycles are low as compared to the good
quality cycles.
• The model keeps on actively learning with the newly arriving data and keeps on
updating the weights.

X. Future scope:

• The system can be further trained to predict the remaining useful life of the
machine before it requires maintenance or replacement.
• The use of stacked architectures of various models can be used to increase the
confidence in classification.
• The use of proactive anomaly detection can be used to send signals to the
machine controller to control the machine parameters to prevent bad quality
production cycles and hence increase the overall productivity.

12
III. Article 2: Machine learning
approach for predictive
maintenance in industry 4.0

13
 I. Resume:
This article describes a machine learning architecture for predictive maintenance, based on
random forest approach. And this is by developing the data collection and data system
analysis, applying the Machine Learning approach and comparing it to the simulation tool
analysis.

II. Predictive maintenance:

Predictive maintenance refers to the intelligent monitoring of equipment to avoid future
failure. It allows to avoid unnecessary equipment replacement, reduce machine downtime,
find the root cause the fault, and in this way save costs and improve efficiency.

1. Aspects of predictive maintenance:

Accordingly, predictive maintenance mainly focuses on two aspects: energy efficiency
improvement (key of energy saving) and unscheduled downtime reduction.

The algorithms developed around these two can also be generally divided into two
categories:

• energy and efficiency, there have been many evaluation methods and devices
developed, as in,
• system condition monitoring, including the detection of motor faults, with various
fault- detection techniques developed, as for instance in.

2. Approaches of PDM:
In PDM three kinds of approach can be distinguished:

• Data-driven approach: Uses historical data to learn a model of system behavior

• Model-based approach: has the ability to incorporate physical understanding of
the target product, relying on the analytical model to represent the behavior of the
system.
• Hybrid approach.
• Machine Learning-based PDM can be divided into the following two main classes:

• Supervised: where information on the occurrence of failures is present in the

modelling dataset.
• Unsupervised: where logistic and/or process information is available, but no
maintenance data exist.

In this resume we’ll be describing a machine learning approach that was applied on a cutting
machine.

14
 Figure 4: The machine

III. The data analysis: 

One of the most important things to understand when analyzing data is the type of
relationship between the measured quantities. To do this, graphic visualization with scattered
plots can be viably used to identify how the data are dependent.
Three essential data sources must be found:
1- Fault history: Even though error events are very rare in predictive maintenance
applications. when compiling predictive models that estimate failures, it is necessary
for the algorithm to learn the normal operating scheme, in addition to the failure
scheme through the training process. As a result, a sufficient number of examples in
both categories is required.
2- Maintenance/repair history: the detailed asset of maintenance history, which contains
information about replaced components, preventive maintenance tasks performed,
and so on.
3- Machine conditions: to estimate how many days (or hours, kilometers, etc.) a
machine lasts before a failure occurs.
In order to prepare the data in the format required to create the functions to be included in
the Machine Learning algorithm, it is necessary to perform some pre-processing steps. The
first is to divide the duration of data collection into units of time in which each record is a unit
of time for an asset. The generic data schema could be:
Maintenance records: These are the records of the maintenance actions performed. The
unprocessed maintenance data is usually associated with an asset ID and a timestamp with
information on maintenance activities performed at that time. In the case of unprocessed

15
data, maintenance activities must be transformed into category columns, where each
category corresponds to a type of maintenance action. The base data schema for
maintenance records will include columns for asset IDs, time and maintenance actions.
• Fault records: these are records that belong to the target of the estimate, i.e.
failures or reason for the fault. They can be specific error codes or error events
defined by specific operating conditions. In some cases the data includes multiple
error codes, some of which correspond to faults of interest. Not all faults are the
goal of an estimate, so others are typically used to create functions that can be
related to failures. The base data schema for fault records will include columns for
asset ID, time and fault, or reason for failure if a pattern is available.
• Machine conditions: preferably real-time monitoring data related to the operating
conditions of the data.
• Machine and operator data: these data can be combined in a scheme to identify
the assets managed by a given operator together with the properties of the assets
and the operator.

Figure 5:Features Dataset

16
In Table: a list of data is reported with the statoRot status variable that will be the machine
condition variable in the result section.

The first step in modeling is the design of the functions. The idea of generating functions
consists conceptually in describing and abstracting the integrity condition of a machine at a
given time using historical data collected up to that instant. In the following, delay (lag)
functions need to be created from data sources including timestamps, and also static
functions created from static data sources providing examples from the illustrated use cases.

In the preventive maintenance the chronological data generally include timestamps that
indicate the time of collection of each single data. For each asset record, a window with
dimension “W” is selected, corresponding to the number of units of time for which we want to
calculate the chronological aggregations. The aggregate sequencing functions are then
calculated using the W periods before the date of that record. Examples of sequential
aggregations are: incremental counts averages, standard deviations, outliers based on
standard deviations, cumulative sum measures, minimum and maximum window values, or
trend changes, peaks and level changes using algorithms that detect anomalies, etc.

IV. Machine Learning algorithms:

The binary classification is used in predictive maintenance to estimate the probability that the
equipment will fail over a future period of time.

To use the binary classification, it is necessary to identify two types of examples, which are
called positive and negative. Each example is a record of a unit of time for an asset that
conceptually describes the operating conditions by designing the functions by using historical
and other data sources. In the context of the binary classification for predictive maintenance,
the positive types denote the errors, and the negative ones the normal operations. The aim is
to find a model that identifies the probability that each new example may fail or work normally
within the next unit of time.

Multi-class classification for predictive maintenance can be used to estimate two future
results. The first is to assign an asset to one of the different time periods in order to allocate
a time interval for the failure of each asset. The second consists in identifying the probability
of failure in a future period due to one of the multiple root causes. This allows the
maintenance personnel to handle the problem in advance. Another multiclass modeling
technique focuses on determining the most likely root cause of a given failure. In predictive
maintenance, similar to any other area of the solution containing time-stamp data, the typical
training and test routine must consider the variable aspects over time for a better
generalization of future unseen data.

Many Machine Learning algorithms depend on a number of hyper parameters that can
significantly modify the performance of the model. The optimal values of these hyper
parameters are not automatically calculated during the training of the model, but must be
specified by the data scientist. Various ways are available to find the optimal values of the
hyper parameters. The most common one is the “cross-validation of k sections”, which for
each set of values of the hyper parameters, the learning algorithm is executed k times. At
each iteration, the examples in the current section are used as a validation set, while the rest
of the examples are used as a training set. The algorithm performs training on training

17
examples and performance metrics are calculated on validation examples. At the end of this
cycle, for each set of values of the hyper parameters the values of the performance metrics k
are calculated and the values of the hyper parameters with the best average performances
are selected.

In the result section a Decision Forest (DF) classifier is used. The DF Classifier is an
ensemble learning method used to improve classification performance. It uses bagging by
combining the output of several classifiers.

18