Integration of AI & ML in Mechanical Engineering

Unit-3: AI in Advanced Manufacturing

Introduction:

Today’s manufacturing systems are becoming increasingly complex, dynamic and connected.
The factory operations face challenges due to countless uncertainties and interdependencies.
Manufacturing is entering a period of innovation and change with increased integration of
sensors, Internet of things (IOT), advances in robotics and automation, leads to smart
manufacturing and Industry 4.0. AI plays a vital role in transforming the manufacturing domain
through advanced analytical tools for processing vast amounts of manufacturing data generated.
Recent advances in computational hardware, sensing technology for the collection of critical
process/ machine data have made the application of these AI techniques feasible in a practical
sense. AI techniques provide solutions for significant improvement in productivity, quality,
flexibility, safety and cost.

Evolution of AI:

AI defined as “ability of computers to perform cognitive functions associated with human minds,
such as perceiving, reasoning, learning and problem solving. The perceptron invented in 1956, a
neural network structure designed to simulate a human neural system by utilizing a weighted
sum of inputs. The inability of perceptron to process the simplest logic and high computational
complexity led to reduced enthusiasm for supporting AI related research, the beginning of first
AI winter. The recovery from first winter was enabled by expert system built based on experts
knowledge and ‘if-then rules’, made many to believe that expert knowledge was best means to
create AI. The knowledge based AI is inoperative in the design and maintenance of inherently
complex logic, leading to second Ai winter. ML/DL contributed to recovery of AI from second
winter along with the advances in sensing and computational infra structure.

Opportunities for AI in manufacturing:

The main driving factors and requirements in most daily manufacturing operations across
industries are meeting demands, quality and cost while ensure a safe working environment for
all. It is difficult to remain profitable with growing product and process complexity, higher
variability in customer demand and preferences, high competition in markets. The tough
competition drives the manufacturers to adopt AI over conventional tools and approaches. Now-
a-days, huge amounts of continuously generated data are produced by machines, sensors,
controllers and labour records etc.,
The data is categorized as :

- Environmental data collected from ambient sensors, eg: room temperature and humidity.
- Process data collected from sensors on process machines, eg: machining and grinding
coolant temperature, power and heat treatment temperature.
- Production operation data recorded in controller system, eg: elapsed time of each part in
each operation station, machine downtime, idle time and shift scheduling.
- Measurement or check data from product quality inspections, eg: product diameter, form.

This data is useful for pattern discovery that can contain information to solve tough problems
with understanding of physical characteristics of a system or process. Ai provides the ability to
transforms large amounts of complex manufacturing data in today’s factory into actionable
information.

Hierarchical approach to manufacturing system:

The Ai in manufacturing has attracted much attention in scientific community with number of
publications steadily growing.

The publications covers:

- The key elements in smart manufacturing system, government initiatives.

- Constituent technologies such as IOT, cyber-physical systems (CPS), cloud computing,
big data analytics and information and communications technology (ICT).
- ML/DL techniques with their applications in manufacturing.
- AI for condition monitoring and fault diagnosis.
- AI in human-robot collaboration.
The hierarchical 3 level decomposition of manufacturing system (system-process-material)
consists of

a) Low level physical process (for which modeling of material property plays an important
role).
b) Mid level sensing/monitoring process (foundation of maintenance, prognostics and health
management (PHM).
c) Top level managing manufacturing operation (system level modeling and performance
analysis).

AI for manufacturing system optimization:

The performance of manufacturing system consists of optimum material flow through work
cells.

a) Quality Analysis: Classical statistics based approaches like control charts and run charts
mainly focus on process level quality assurance.
- With auto encoder and RNN, quality control of work in progress products in powder
metallurgy process is developed.
- A Combination of cluster analysis and SVM is introduced to achieve improved quality
monitoring by early detection of quality defects with time series data.
- Bayesian networks are used to conduct root cause diagnosis of quality issues to infer fault
causes from observed quality deviations in automotive body assembly process.
 b) Manufacturing system decision making and control:
- Manufacturing system control is concerned with decision-making and controlling of the
physical activities in a factory in order to improve productivity, quality and overall
efficiency.
- In flexible manufacturing system, multiple product orders awaiting processing, it is
critical to determine which order should be prioritized and which machine should
undertake the job. The ML based methods like supervised and RL are extensively used
such as Decision Tree, Neural Networks, SVM and ensemble learning methods.
- With rapid advancement of RL in recent times, the deep Q network (DQN) is applied to
solve a large scale scheduling problem in a semi conductor manufacturing system.
- Maintenance activities aim to keep machines in desrirable reliability levels or to quickly
recover them from random failures. RL has been used to obtain near optimal maintenance
policies for manufacturing system.
- In a manufacturing system, resource allocation is the process of assigning manufacturing
resources like robots, human operators, and machines for specific time periods to the set
of manufacturing processes.
- Resource allocation problems may arise in both design & daily operation. NN is trained
to quickly predict system throughput given system design parameters.
- With the advancement of industrial IOT, AI and ML, there is increasing utilization
penetrating the entire manufacturing system. ML algorithms potentially provide powerful
tools to reduce cycle time and scrap; improve quality and improve resource utilization in
certain hard manufacturing problems.

AI for manufacturing applications of human-robot collaboration:

Human robot collaboration (HRC) is ‘ a state in which a purposely designed robot system and an
operator work on simultaneous tasks within a collaborative work space ie., whiere the robot
system and a human can perform tasks concurrently or even jointly.

- The objective of applying robotics in manufacturing has been to leverage the advantages
robots have over humans such as repeatability, endurance, strength, ability to operate in
hazardous environments.
- With the recent developments in robot capabilities such as sensing, perception, control,
teaching and learning and very nature of human-robot relationship is being examined
with a focus toward attaining greater productivity, flexibility and adaptability.
- However, advances in AI technologies may enable and accelerate the development and
implementation of HRC into the industry.
- Above Fig. shows an organizing frame work to map AI/ML technologies to existing and potential
industrial HRC applications.
- Plan comprises 3 elements, like system modeling, setting and meeting objectives and
formulating strategies.
- Teach, involves communicating commands to robots execute specific actions. The system is built
on a semantic network as the basis for a natural language processing system comprised of
automatic speech recognition, visual simulation environment and reasoning. Human voice
commands are inputs to reasoning system which takes inputs from the natural language
processor to search for a reasonable set of robotic actions.
- Operate, concerns the required active steps, processes, tasks that are completed independently,
sequentially or simultaneously by the human(s) and robots in HRC systems to achieve planned
objectives.
- Monitor, requires the human operator to continuously observe and evaluate the execution of
robots acting and progress toward completion of all tasks and respond to any anomalous events
that may arise during operation.
- Internet of robot things (IORT) describes intelligent devices that can monitor events, fuse sensor
data from a variety of sources, use local and distributed intelligence to determine the best
course of action.
- Intervene, relates to the intervention of operations and considers what alternative steps or
corrective actions must be pursued if, in the course of monitoring, the human or robot observes
the occurrence of a fault or anomaly. An important aspect of intervention in the control of HRC
 system is the capability for real time decision making. In complex and ambiguous situations,
human have a relative advantage compared to robots owing to human cognitive abilities by
virtue of experience.
- Learn, human cognitive activity of gaining knowledge from observing the results of the systems
performance. A learning approach is developed based on symbolic AI (inductive logic
programming- ILP) for task execution in cognitive robots.

AI for process monitoring, diagnostics, prognostics:

The manufacturing is closely associated with improving the reliability of machine equipment to
reduce unexpected down time. Research on machine condition monitoring, defect root cause
diagnosis and remaining useful life (RUL) prognosis establishes the core competence for
maintenance. The goal is to enable timely detection and isolation of the incipient faults of
components, predict their progression and support rational decision making. The significance of
maintenance in manufacturing lies in revealing the health state of individual machines or production
system in real time, providing a diagnosis of anomalies root cause and preventing the occurrence of
a failure, to ultimately achieve near zero down time.

The shift from manual to automated production in 3 rd industrial revolution witnessed the expansion
of data collection and adoption of AI methods. The ML methods can learn useful data patterns and
relate to processing condition and performance, with the underlying concept that data are the
reflection of the physical machine characteristics.

AI for condition based maintenance:

Manufacturing system are subject to faults caused by a broad range of causes from heat generation
to corrosion and fatigue. Fault diagnosis is intended to pin point the source of fault generation and
provide an estimate of the severity of the fault, which forms the basis for condition based
maintenance. The idea of AI based diagnosis is to formulate the task as classification and associate
the information embedded in data to fault types and severity levels. RF is used for induction motor
fault diagnosis with vibration, current, voltage and flux as input and sensor signal as output. A novel
method for detecting filament breakage and nozzle clogging in fused deposition modeling has been
developed.

As sensing technology has continually evolved and adopted on the factory floors, image data has
been increasingly gathered and analyzed for process monitoring and fault diagnosis. A vision system
based on deep CNN has been built for process fault detection is selective laser melting (SLM).

AI based prognosis for predictive maintenance:

Prognosis aims at predicting the temporal progression of machine performance degradation, from
its current state to final functional failure. AI based prognosis is part of the data driven method that
relies on establishing a machine performance evolution model to predict future machine
performance based on its current and past status.
From above fig. the machine performance is difficult to measure directly, an Artificial health
index (HI) is often created from sensor data to represent the machine performance. SVM has
been the most commonly used for machine prognosis. RNN and its variant LSTM (long short
term memory) offer Advantages for machine prognosis.

AI for manufacturing process control:

AI techniques can provide understanding of material to process relationship for enhanced process
control, efficiency, quality and productivity. This understanding helps in optimization of process
and designing new materials. In factory, the increasing amount of instrumentation and ability to
collect large amounts of data along with ability to extract necessary process signals and features
will continue to facilitate control of machining process using AI techniques such as SVM, RF,
DNN and GA. The objective of implementing an AI tool in manufacturing process control is to
produce high quality products cost effectively. For example, the quality of laser cutting process
is increased by optimization of parameters such as laser power, cutting speed and pulse
frequency using ANN. During milling process, ANN predictive models are beneficial in two
ways. Firstly, prediction of surface roughness based on the real time milling parameters, the
roughness deviations exceeds threshold indicates change of worn cutting inserts. Secondly,
prediction of wear time based on the different combinations of tooling parameters. The process
parameters are adjusted to increase tool life span. The grinding is highly non-linear and its multi-
modal relationships are difficult to be defined accurately via mathematical equations since output
variables are not only dependent on input variables but also dependent on other intermediate
variables. An intelligent approach to online control of grinding process is developed. AI based
controllers on grinding forces or other in process data such as tooling deflections. The challenges
faced in manufacturing process is determining the manufacturing conditions on material
property, that influence the quality of final manufactured parts. AI techniques are used for
predicting the properties such as hardness, melting point etc., based on micro structure, lattice
constant, electron affinity etc.,
Challenges and opportunities for future research:

Factories are getting smarter as companies are increasingly able to leverage AI to transform
information from various aspects of the manufacturing system into actionable insights. ML is
used in specific applications of process monitoring, optimization, but its utilization is limited at
the system level of decision making. It is due to non-dynamical nature of manufacturing system
and the complex multi stage processes and dependencies among vast amounts of heterogeneous
data generated there in.

However, the appropriate selection of techniques and algorithms remain challenging. An in depth
understanding of a problem, its causes, consequences, and desired solution state must be known
or well investigated to improve the likelihood of effective AI tool selection and subsequent
model building, data analysis and interpretation. The quality of manufacturing data have strong
influence on performance and suitability of AI algorithm relative to expected results. The
increasing availability of heterogeneous data in manufacturing system, high degree of irrelevant
and redundant information present new challenges. The data collection from different points,
management approaches for different applications by dynamically linking different data,
information and models is challenging.

In real applications, obtaining data from different operating conditions in a systematic way is
challenging financially and operationally. Two approaches namely transfer learning and data
synthesis is used. Transfer learning includes a series of techniques to adapt a model that has been
well established in a source domain in which training data are sufficient, to a related, target
domain in which data is scarce. For condition monitoring or fault diagnosis, the transfer learning
allow experimental data obtained from machines without faulty conditions for model
construction on real production lines when faulty conditions may occur. Data synthesis refer to
generating synthetic data that highly resembles the real data by sufficiently learning data
characteristics, which allow us to bypass the limitation of data availability. Eg: Generative
adversarial networks are used to generate synthetic vibration signals to establish the gear box
fault diagnostic model, captured the fault related signal features such as characteristic
frequencies effectively.

The understanding and interpreting the output of AI tools and related technical details become
increasingly challenging to data scientists and professionals with specialized skills in this
domain. This creates a knowledge gap leaving plant managers and production engineers lack this
back ground with the ability to comprehend and appropriately interpret the meaning of results
from AI models in a manner consistent with the context of problem area to take decisions due to
lack of expertise in AI. This necessitates research into explainable AI models, known as XAI,
which facilitates mental support models to assist users of AI technologies more readily adopt
them as powerful tools enable smart manufacturing.