Hybrid Manufacturing, Automation and Smart Factories

Introduction and Context

Modern production increasingly combines additive manufacturing (AM), subtractive manufacturing, robotics, sensing, industrial software and data analytics. This integration is commonly described as hybrid manufacturing when additive and subtractive processes are combined within one machine or coordinated production system. It becomes part of a smart factory when machines, people, products and information systems are connected so that production can be monitored, analysed and improved continuously.

A conventional CNC machine removes material from a solid workpiece. An additive machine creates material layer by layer. Each process has different strengths. Subtractive machining provides dimensional accuracy, surface finish and mature process control, whereas additive manufacturing enables complex internal geometry, lightweight structures, customization and material deposition only where required. A hybrid process can deposit material to near-net shape and then machine critical surfaces, or repair a damaged component and restore it to final tolerance.

The objective of this chapter is to explain how additive and subtractive processes are integrated with automation, robotics, sensors, industrial connectivity, digital twins, manufacturing execution systems (MES), enterprise resource planning (ERP), cloud–edge computing and artificial intelligence (AI). The emphasis is not merely on equipment, but on the flow of information required to make manufacturing adaptive, traceable and economically useful.

Learning Objectives

After studying this chapter, the reader should be able to:

1. Explain the principles and industrial value of hybrid additive–subtractive manufacturing.
2. Distinguish automation, connectivity, digitization, digitalization and smart manufacturing.
3. Describe the roles of robots, sensors, PLCs, CNC controllers, MES, ERP and digital twins.
4. Select appropriate edge, cloud and industrial communication functions.
5. Interpret manufacturing data for monitoring, anomaly detection and predictive maintenance.
6. Explain the basic operation and limitations of AI-based parameter optimization.
7. Analyse a hybrid manufacturing implementation from technical, economic, safety and governance perspectives.

1. Hybrid Manufacturing Principles

1.1 Additive and subtractive capabilities

Additive processes include laser powder-bed fusion, directed energy deposition, wire-arc additive manufacturing, material extrusion and binder jetting. Subtractive processes include milling, turning, drilling, grinding and electrical discharge machining. A hybrid system may combine these operations in three ways:

• Sequential integration: additive and subtractive operations are performed on separate machines and coordinated through a shared digital process plan.
• Cell-level integration: several machines, robots, inspection devices and material-handling systems operate as an automated manufacturing cell.
• Machine-level integration: deposition and machining tools are mounted on, or integrated with, one machine platform.

In machine-level integration, the workpiece may remain clamped while material is deposited and subsequently machined. Avoiding reclamping reduces datum-transfer error and setup time. However, the machine must manage thermal distortion, tool access, deposition strategy, chip removal, contamination and process qualification.

1.2 Why combine the processes?

The main technical motivations are:

• Near-net-shape production: additive deposition reduces the volume that must later be removed.
• Repair and remanufacture: worn or damaged regions can be rebuilt before machining.
• Feature integration: channels, bosses, wear-resistant layers and graded materials may be added locally.
• Reduced setup error: deposition and machining can use a common coordinate system.
• Improved surface quality: machining corrects the roughness and geometric variation of additive surfaces.
• Material efficiency: material is deposited selectively rather than cut from a large billet.

Hybrid manufacturing is not automatically superior. Additive deposition may introduce porosity, residual stress, anisotropy and dimensional variation. Machining still requires tool access, fixturing and chip management. The optimal process is therefore determined by the complete lifecycle cost and performance, not by the presence of advanced equipment alone.

1.3 Process-chain model

A typical hybrid process chain is:

1. Capture or design the three-dimensional geometry.
2. Define material, build orientation, deposition path and machining allowances.
3. Generate additive and subtractive toolpaths.
4. Verify collision, reachability, thermal constraints and material compatibility.
5. Load material and establish workpiece and tool references.
6. Deposit material while recording process data.
7. Inspect intermediate geometry and update the process plan if necessary.
8. Machine reference surfaces and functional features.
9. Perform final inspection and release the digital production record.

The important principle is closed-loop manufacturing. In an open-loop process, a program is executed without using measured results to modify the operation. In a closed-loop process, sensing and inspection influence subsequent decisions, such as changing deposition speed, adjusting energy input, adding material, or applying a corrective machining allowance.

2. Automation Architecture

2.1 Levels of control

A manufacturing system can be viewed as a hierarchy:

A reliable architecture separates safety-critical control from noncritical analytics. A cloud service may calculate a maintenance forecast, but it should not be the sole mechanism preventing a robot collision. Local safety controllers, emergency stops, guarding and interlocks must remain functional if the network is unavailable.

2.2 Robotics in hybrid production

Industrial robots are used for loading and unloading, tool changing, deposition, inspection, cleaning, pallet transfer and collaborative assistance. Robot selection requires consideration of payload, reach, repeatability, stiffness, singularities, workspace, environmental exposure and end-effector compatibility.

Robot repeatability describes the ability to return to a position under similar conditions. It is not identical to absolute accuracy, which describes closeness to the commanded coordinate. Calibration, thermal expansion, joint compliance and payload deflection can cause an accurate-looking program to produce inaccurate parts. External metrology, such as laser trackers or structured-light scanners, may therefore be required.

For a robot pose, the forward-kinematic relationship can be represented conceptually as:

Here, T₀ⁿ is the homogeneous transformation from the base frame to the tool frame, A_i is the transformation associated with joint or link i, q_i is the joint variable, and n is the number of joints. The equation applies to a defined robot geometry and coordinate convention. It does not by itself account for deflection, backlash or calibration error.

Robotic additive deposition often requires coordinated motion of the robot, deposition head and substrate positioner. The process planner must maintain suitable torch or nozzle orientation, stand-off distance, travel speed and material-feed condition. A robot can reach a point geometrically but still be unable to reach it with the required orientation or without collision.

3. Sensors and Manufacturing Data

3.1 Sensor categories

Important sensors include:

• Thermocouples and infrared cameras for temperature measurement.
• Accelerometers and acoustic sensors for cutting and spindle-condition monitoring.
• Dynamometers and motor-current sensors for cutting-force estimation.
• Melt-pool cameras, photodiodes and pyrometers for additive-process monitoring.
• Laser scanners and structured-light systems for geometry measurement.
• Encoders for position and velocity.
• Pressure, flow and humidity sensors for gas, coolant and powder systems.
• Vision systems for part identification, surface inspection and robot guidance.

Sensor data are useful only when the measurement is linked to a physical decision. For example, a high vibration value may indicate chatter, imbalance, bearing damage or a loose fixture. The signal must therefore be interpreted in relation to spindle speed, tool engagement, workpiece material and machine state.

3.2 Sampling and data quality

A sensor signal is a time-dependent sequence. If the highest frequency of interest is f_max, a sampling frequency greater than twice that value is required under the ideal Nyquist condition:

Here, f_s is the sampling frequency in hertz. This condition assumes appropriate anti-alias filtering and a band-limited signal. In industrial practice, higher sampling rates may be selected to provide margin and to capture transients.

Data quality problems include missing values, sensor drift, timestamp errors, inconsistent units, communication loss and changes in machine configuration. A data pipeline should record sensor identity, calibration status, sampling rate, units, machine state, program number, tool identity and time synchronization. Without this context, large data volumes do not necessarily create useful knowledge.

3.3 Manufacturing performance measures

Useful measures include cycle time, first-pass yield, scrap rate, overall equipment effectiveness (OEE), energy per part, tool life, dimensional capability and unplanned downtime. OEE is commonly expressed as:

Availability is the proportion of planned production time during which the equipment is operating. Performance compares actual production rate with an ideal rate while operating. Quality is the proportion of good units among total units produced. The three terms are dimensionless fractions between 0 and 1, or percentages when multiplied by 100. The measure is meaningful only when planned time, ideal cycle time and quality definitions are consistently established.

4. Industrial Connectivity and Information Systems

4.1 Connectivity versus interoperability

Connectivity means that devices can exchange signals. Interoperability means that the exchanged information has a shared meaning and can be used correctly by another system. A machine may transmit a value labelled “temperature,” but the receiving system also needs to know whether it is spindle temperature, melt-pool temperature or ambient temperature, and whether the unit is degrees Celsius or degrees Fahrenheit.

Common industrial communication technologies include Ethernet-based networks, OPC UA, MQTT, fieldbus systems and vendor-specific protocols. OPC UA supports structured information models and secure communication. MQTT uses a publish–subscribe pattern that is useful for lightweight telemetry. Neither protocol automatically guarantees correct data semantics, real-time performance or cybersecurity; these depend on architecture and configuration.

4.2 MES and ERP

ERP systems manage business resources such as orders, purchasing, inventory, finance and workforce planning. MES systems manage the execution and traceability of production. Typical MES functions include dispatching work orders, recording material batches, assigning programs, tracking work-in-progress, collecting inspection results and maintaining electronic device histories.

The relationship can be summarised as follows: ERP determines what the organization needs and when; MES determines how approved work is executed and recorded; controllers and machines perform the physical operation. Poor integration can lead to duplicate data entry, obsolete programs, incorrect material selection and weak traceability.

4.3 Digital thread and digital twin

A digital thread is the connected flow of information across design, process planning, production, inspection, maintenance and service. A digital twin is a digital representation of a physical asset, process or product that is updated using relevant data and can support analysis or decision-making.

A CAD model alone is not a complete digital twin. A manufacturing twin may include geometry, machine configuration, toolpath, material batch, sensor data, thermal history, inspection results and maintenance state. The level of fidelity should match the decision being supported. A simple energy model may be adequate for scheduling, while a thermal-mechanical model may be required to predict distortion.

Digital twins have limitations. Models may be uncertain, sensors may be incomplete, and a simulation calibrated for one material or machine may not transfer to another. A twin should therefore expose assumptions, confidence levels and model version rather than presenting every prediction as fact.

5. Cloud–Edge Manufacturing Systems

5.1 Functional distinction

Edge computing places computation near the machine. It is appropriate for low-latency decisions, local buffering, machine protection, data reduction and operation during network interruption. Cloud computing provides scalable storage, fleet-level comparison, model training, enterprise integration and collaboration.

A practical architecture assigns functions according to latency, reliability, bandwidth, confidentiality and computational demand:

This is a distributed system rather than a choice between “cloud” and “no cloud.” Edge nodes can filter data, calculate features and transmit exceptions, reducing bandwidth while preserving important events.

5.2 Cybersecurity and resilience

Connected production increases the attack surface. Threats include unauthorized remote access, ransomware, manipulated programs, stolen intellectual property, sensor spoofing and denial of service. Controls should include network segmentation, least-privilege access, multifactor authentication, secure backups, patch management, asset inventories, logging and incident-response procedures.

Safety and cybersecurity are connected. A compromised toolpath or robot command can create physical danger. Access control must therefore be coordinated with functional safety, change approval and machine validation. Operators should be able to place equipment in a safe state without depending on a remote service.

6. Monitoring, Anomaly Detection and Predictive Maintenance

6.1 Monitoring and diagnosis

Monitoring compares current process behaviour with expected behaviour. Diagnosis identifies the likely cause. For example, monitoring may detect increased spindle vibration; diagnosis may distinguish tool wear from bearing damage by combining vibration frequency, cutting load, tool age and operating conditions.

A simple standardized residual is:

Here, x_t is the current observation, μ is the reference mean, σ is the reference standard deviation, and z_t is the standardized residual. The method assumes that the reference distribution is reasonably stable and that σ is nonzero. A large absolute value may indicate an anomaly, but it is not proof of a fault; operating-state changes can also produce it.

Moving averages reduce short-term noise:

Here, w is the window length in samples and m_t is the average at time t. Larger windows smooth more strongly but delay detection. The method is unsuitable when rapid transients are the primary fault signature.

6.2 Machine-learning approaches

Manufacturing anomaly detection may use:

• Threshold methods: simple, interpretable limits based on engineering knowledge.
• Statistical process control: control charts and multivariate monitoring.
• Supervised learning: classification or regression trained with labelled faults or quality outcomes.
• Unsupervised learning: clustering or autoencoders trained primarily on normal data.
• Physics-informed models: combine process equations with measured data.
• Hybrid models: use physical constraints and machine learning together.

Supervised models can be highly effective when fault labels are reliable. Unfortunately, failures are rare, labels may be delayed, and machine conditions change. A model trained on one tool, material and cutting regime may fail under another. Validation should therefore use time-separated data, unseen jobs and, where possible, different machines.

6.3 Predictive maintenance

Predictive maintenance estimates the future condition or remaining useful life of equipment. It should support a maintenance decision, not merely produce a probability. A useful workflow is:

1. Define the failure mode and business consequence.
2. Identify measurable precursors.
3. Establish a healthy baseline.
4. Extract meaningful features.
5. Validate detection and false-alarm rates.
6. Link predictions to inspection, spare parts and work orders.
7. Review outcomes and update the model.

False alarms can be expensive because they interrupt production and encourage operators to ignore warnings. Missed detections can cause safety incidents, scrap or catastrophic damage. Performance must therefore be evaluated using the costs of both error types, not accuracy alone.

6.4 Parameter optimization

AI can recommend cutting speed, feed, layer height, laser power, travel speed or deposition rate. Optimization must respect constraints such as machine limits, thermal conditions, surface integrity, dimensional tolerance, tool life, energy and safety.

For a generic objective function, the selected process vector p may be written conceptually as:

Here, p* is the selected parameter vector, P is the feasible set of safe and qualified parameters, and J(p) is the objective, such as cost, cycle time or defect risk. The notation does not specify an optimization algorithm. In practice, J may combine several normalized objectives, and constraints must be experimentally validated.

A model should not recommend conditions outside its training domain without uncertainty assessment and engineering approval. Bayesian optimization, designed experiments, response-surface methods and reinforcement learning are possible approaches, but each requires careful definition of the objective, constraints and reward. A statistically optimal setting is not necessarily manufacturable or certifiable.

7. Worked Example: Edge-Based Tool-Wear Detection

Consider a milling cell that records spindle power once per second during a stable cutting segment. For a new segment, the reference mean power is 4.0 kW and the reference standard deviation is 0.25 kW. The measured power is 4.8 kW.

Using the standardized residual:

The units cancel, so the result is dimensionless. If the engineering rule is to investigate values with |z| ≥ 3, the segment is flagged. The result does not establish that the tool is worn. The operator or diagnostic system should check material, tool engagement, coolant, fixture condition and program identity. If several consecutive segments are flagged under comparable conditions, a tool inspection or controlled replacement may be justified.

This example illustrates three principles: a threshold must be defined before evaluation, units must be consistent, and an anomaly flag is a decision aid rather than an automatic diagnosis.

8. Applied Case: Illustrative Hybrid Repair Cell

Background

The following is an illustrative industrial case. A manufacturer repairs high-value steel forming dies. Conventional repair involves manual welding, repeated setup, extensive machining and final inspection. Lead time is long, and repair quality varies with operator technique.

Problem

The organization experiences three problems: inconsistent deposited volume, distortion after welding and weak traceability of repair material and process conditions. The die must meet tight dimensional requirements on functional surfaces.

Analysis

An engineering team maps the process and identifies critical variables: substrate temperature, deposition current, wire-feed rate, travel speed, torch orientation, interpass temperature, machining allowance and final geometry. A robot-mounted deposition head is selected, while the die remains on a positioner. A five-axis machining centre is used after deposition. A laser scanner measures the damaged region before and after repair.

A local edge computer collects robot position, current, voltage, wire-feed rate, temperature and scanner data. The MES issues the repair order, confirms material-batch identity and records the approved program. ERP supplies the customer order, estimated cost and material availability.

Approach and solution

The process is divided into controlled stages:

1. Scan the damaged die and align the measured geometry with the nominal CAD model.
2. Generate a deposition path with a qualified material and a planned machining allowance.
3. Calibrate the robot, positioner and scanner coordinate frames.
4. Deposit material while monitoring current, voltage, temperature and travel speed.
5. Use an edge rule to flag excessive thermal or geometric deviation.
6. Scan the deposited region and update the machining model.
7. Machine datum and functional surfaces without removing the workpiece from the integrated fixture when possible.
8. Inspect dimensions and surface finish, and archive the complete digital record.

The system does not allow the AI model to alter deposition parameters without limits. It may recommend a revised travel speed within a qualified process window, but an approved engineer authorizes changes. Safety functions remain local to the robot and machine controllers.

Outcome

The expected outcome is reduced setup error, improved traceability, more consistent repair geometry and reduced unnecessary machining. The business case must still include robot integration, scanner calibration, software validation, operator training, maintenance and cybersecurity. A faster cycle time alone is insufficient if qualification costs or false alarms offset the benefit.

Learning points

The case demonstrates that smart manufacturing is a socio-technical system. Equipment, data models, employees, quality procedures and business systems must work together. The digital twin is valuable because it connects measured damage, deposition history, machining allowance and final inspection. The AI component is useful only within a qualified process window and under human and procedural control.

9. Implementation Framework and Best Practices

A successful implementation should begin with a clear production problem rather than with a technology purchase. The following sequence is effective:

1. Define the business and engineering objective.
2. Map the current process and identify losses, risks and constraints.
3. Establish data ownership, naming conventions, units and time synchronization.
4. Select a pilot with measurable scope and manageable safety risk.
5. Integrate sensors and systems using a documented architecture.
6. Establish baseline performance before applying AI.
7. Validate models using representative and independent data.
8. Conduct machine, process, software and cybersecurity validation.
9. Train operators, maintenance staff, programmers and engineers.
10. Scale only after benefits, failure modes and governance are understood.

Important best practices include version-controlled NC programs, calibrated sensors, controlled material batches, traceable tool identities, standardized coordinate systems, automated data validation and clear ownership of alarms. Human-machine interfaces should display actionable information rather than overwhelming operators with raw signals.

10. Challenges, Limitations and Governance

Technical limitations

Hybrid processes can suffer from thermal distortion, residual stress, porosity, poor bonding, tool accessibility problems and coordinate-registration errors. Sensor placement may be difficult, and data from different vendors may lack common semantics. Models can drift as tools, materials, software and machine conditions change.

Economic limitations

Capital cost includes machines, robots, sensors, integration, software, training and qualification. Savings may arise from reduced setup, scrap, labour, material use and downtime, but these benefits must be measured against maintenance and data-infrastructure costs. A small production volume may not justify a highly automated cell.

Human factors

Automation changes job roles rather than eliminating the need for expertise. Skilled personnel are needed to interpret process physics, validate programs, maintain equipment, investigate anomalies and manage exceptions. Poorly designed automation can increase cognitive load if alarms are excessive or unexplained.

Ethical and regulatory issues

Manufacturing data may contain employee information, customer designs and supplier intellectual property. Access and retention must be governed. AI recommendations should be explainable enough for responsible engineering review, especially where product safety is involved. Audit trails should record model versions, input data, recommendations, approvals and overrides.

Safety standards and regulatory requirements vary by jurisdiction and machine type. Risk assessment should address robot motion, laser radiation, high voltage, hot material, fumes, powder exposure, compressed gas, chips and unexpected automatic restart. Conformance to relevant machinery, electrical, functional-safety, additive-process and quality-management requirements must be addressed by the responsible organization; connectivity does not replace formal safety validation.

11. Contemporary Developments

Current development is moving toward autonomous process planning, multimodal inspection, interoperable digital twins, low-latency industrial wireless networks, energy-aware scheduling and foundation models for engineering information. More machines are exposing structured data, enabling fleet-level learning and cross-process comparison.

However, technological novelty should not be confused with industrial maturity. Autonomous parameter adjustment requires reliable sensing, causal understanding, bounded operating regions and a safe recovery strategy. Future systems are likely to be human-supervised adaptive systems, in which AI performs detection, prediction and recommendation while engineers define constraints and approve high-consequence changes.

Chapter Summary

Hybrid manufacturing combines additive deposition with subtractive machining to exploit the geometric flexibility of additive processes and the accuracy and surface quality of machining. Integration may occur within one machine, across an automated cell or through a coordinated production network. Robots provide motion and material handling, while sensors supply information about geometry, force, temperature, vibration and process condition.

A smart factory connects field devices, controllers, MES, ERP, digital twins and cloud–edge systems. Edge computing supports rapid local decisions and resilience; cloud systems support historical analysis, model training and enterprise collaboration. Manufacturing data must be time-synchronized, calibrated, semantically defined and connected to a practical decision.

Monitoring detects deviations, diagnosis investigates causes, predictive maintenance estimates future condition and AI-based optimization recommends process parameters within qualified constraints. Statistical thresholds and machine-learning models are useful but can produce false alarms, drift and unsafe extrapolation. Cybersecurity, functional safety, human expertise, traceability and governance are therefore essential parts of the engineering solution.

Key Terms

Hybrid manufacturing: Integration of additive and subtractive processes in one coordinated production system.

Additive manufacturing: Production by adding material, usually layer by layer, according to digital geometry.

Subtractive manufacturing: Production by removing material using processes such as milling, turning or grinding.

Closed-loop control: Control in which measured output influences subsequent process action.

Robot repeatability: Ability of a robot to return to a position under specified conditions.

Digital thread: Connected flow of product and process information across the lifecycle.

Digital twin: Data-connected digital representation of a physical asset, process or product.

MES: Manufacturing execution system that manages and records production operations.

ERP: Enterprise resource planning system for business resources, orders and supply chains.

Edge computing: Processing performed near the machine or production cell.

Cloud computing: Network-accessible computing and storage that supports scalable analysis and collaboration.

Anomaly detection: Identification of observations that differ significantly from expected behaviour.

Predictive maintenance: Maintenance planning based on predicted equipment condition or failure risk.

Process window: Qualified range of operating conditions that produces acceptable results.

Residual stress: Internal stress remaining in a material after processing or loading.

Interoperability: Ability of systems to exchange and correctly interpret information.

OEE: Product of availability, performance and quality factors.

Digital model drift: Reduction in model validity caused by changing equipment, data or operating conditions.

Review Questions

Short Answer Questions

1. What are the principal advantages and limitations of combining additive and subtractive manufacturing?
2. Distinguish robot repeatability from robot accuracy.
3. What functions are normally performed by an MES that are not normally performed by an ERP system?
4. Why are edge computers useful in manufacturing cells?
5. Why does an anomaly flag not necessarily prove that a machine fault has occurred?

Long Answer Questions

1. Explain the architecture of a smart hybrid manufacturing cell from sensors to ERP, including the function of each level.
2. Discuss how a digital thread and digital twin support traceability and process improvement.
3. Compare threshold monitoring, supervised learning and unsupervised learning for manufacturing condition monitoring.
4. Explain the technical, economic, human and cybersecurity challenges of implementing connected hybrid manufacturing.
5. Describe a complete implementation framework for introducing AI-based parameter optimization into a qualified production process.

Analytical / Application Questions

1. A milling process has a reference power mean of 6.0 kW and standard deviation of 0.4 kW. A new observation is 7.3 kW. Calculate the standardized residual and determine whether it exceeds a three-standard-deviation investigation rule. State one reason why the result should not be treated as conclusive proof of tool failure.
2. Design an edge–cloud architecture for a robotic directed-energy-deposition cell. Identify which functions must remain local and which functions can be placed in the cloud, and justify your decisions using latency, safety and data-volume considerations.
3. A factory reports excellent AI classification accuracy but frequent false alarms after changing material suppliers. Analyse possible causes and propose a validation and governance plan before returning the model to production.

Multiple Choice Questions with Answers

Question 1

Which statement best describes hybrid manufacturing?

A. Using only additive manufacturing for every feature

B. Combining additive and subtractive processes in a coordinated production system

C. Replacing all human operators with robots

D. Connecting an ERP system directly to a household network

Correct Answer: B. Combining additive and subtractive processes in a coordinated production system

Question 2

Which function should normally remain local to a machine or safety controller?

A. Fleet-wide historical benchmarking

B. Long-term production-cost analysis

C. Emergency-stop and safety interlock operation

D. Cloud model training

Correct Answer: C. Emergency-stop and safety interlock operation

Question 3

What is the primary role of an MES?

A. Managing corporate finance only

B. Executing and recording production operations

C. Designing consumer marketing campaigns

D. Replacing all machine controllers

Correct Answer: B. Executing and recording production operations

Question 4

In the equation z_t = (x_t − μ) ÷ σ, what does σ represent?

A. The current machine speed

B. The reference standard deviation

C. The sampling interval

D. The robot payload

Correct Answer: B. The reference standard deviation

Question 5

A process observation is 4.8 kW, with reference mean 4.0 kW and standard deviation 0.25 kW. What is its standardized residual?

A. 0.8

B. 2.0

C. 3.2

D. 4.8

Correct Answer: C. 3.2

Question 6

Which statement about a digital twin is most accurate?

A. It is always identical to the physical system in every detail

B. It is only a static CAD drawing

C. It is a data-connected representation used for analysis or decision-making

D. It is a replacement for all physical inspection

Correct Answer: C. It is a data-connected representation used for analysis or decision-making

Question 7

Why is a cloud-only control architecture unsuitable for an emergency stop?

A. Cloud systems cannot store historical data

B. Emergency stopping requires dependable local and deterministic action

C. Cloud systems cannot communicate with ERP

D. Emergency stops require additive manufacturing

Correct Answer: B. Emergency stopping requires dependable local and deterministic action

Question 8

Which factor is most important when applying AI outside its training domain?

A. Ignoring uncertainty if accuracy was high during training

B. Allowing unrestricted parameter changes

C. Assessing uncertainty and maintaining qualified process constraints

D. Removing all engineering approval

Correct Answer: C. Assessing uncertainty and maintaining qualified process constraints

Question 9

OEE is commonly calculated as:

A. Availability × Performance × Quality

B. Speed × Voltage × Temperature

C. Cost × Inventory × Revenue

D. Accuracy × Repeatability × Payload

Correct Answer: A. Availability × Performance × Quality

Question 10

What is interoperability?

A. The ability of a robot to lift any payload

B. The ability of systems to exchange and correctly interpret information

C. The speed of a spindle at maximum load

D. The removal of all network security controls

Correct Answer: B. The ability of systems to exchange and correctly interpret information