Use Cases

TEAMING.AI framework will be tested in three industrial use case scenarios selected to represent different levels and aspects of human involvement: quality control (Use Case 1) and machine and process diagnostics (Use Case 2). The third one (Use Case 3) focusses on harm prevention in dynamic production environments.

 
 
Objectives Use Case 1Use Case 2Use Case 3
human autonomy●●●●●●
Auditable Ethicsprevention of harm○○●●●
 explicability●●●●●●●●
Agile Developmentcross-functional teamwork efficiency●●●●●●●●
rate of setups●●●●●●●
 product diversity●●●●●●●
Operational Performanceautomation degree●●●●●●
 process dynamics●●●●
 dependency on human●●●●●●●

Index: no relevance ○ low relevance ● medium relevance ●● high relevance ●●●

 

Use Case 1

Transfer learning based robust quality inspection (for plastic injection sector)

Use Case 2

Machine diagnostics for plastic injection sector to improve quality and reduce waste

Use Case 3

Ergonomics and risk prevention in large part manufacturing

Location
FAR (Turkey) is a company which produces automobile plastic parts using injection moulding technique with more than 1000 product variants.

Location
IAL (Spain) is a supplier to the automotive industry, expert in the manufacturing of decorative and non- decorative automotive thermoplastic plastic injection parts for the interior of the vehicle.

Location
GOI (Spain) is specialised in high-precision machining of large-sized parts by either milling or grinding on the basis of cast materials or machine-welded structures.

Use Case Scenario 1

Transfer learning based robust quality inspection (for plastic injection sector)

AI/ML systems in plastic industry usually rely on machine vision techniques based on smart cameras and neuronal networks as classifiers to detect mentioned common faults. Stability problems during quality control process increases setup and maintenance time and “out of tolerance” products always have a risk to be used by customer which negatively affect production efficiency.

The scope

The scope of the demonstrator is to create an intelligent system, which will be able to communicate with the quality inspection unit and with the operator’s inputs through Human in the Loop (HITL) Dashboards, to automatically adapt the TEAMING.AI Engine to self-adjust the quality inspection model when a deviation from nominal is detected in the case of sufficient trust assurance evidence and otherwise to request human assistance. The goal is to increase the Overall Labor Effectivenss (OLE).

The goal

The goal is to use computer vision and advanced deep learning techniques with AI applications for detecting faults on product better/faster than human eye. Through the result of correct classification between non-fault and fault products and the data from production lines, Farplas will have a perfect data set for using AI at setting parameters for injection machines, more reliable and fastest quality control. FAR is a company which produces automobile plastic parts using injection moulding technique with more than 1000 product variants and will deploy the use-case on the production lines for collecting the real data and benchmarking.

Use Case Scenario 2

Machine diagnostics for plastic injection sector to improve quality and reduce waste

This use case focuses on machine and process diagnostics rather than checking the quality at the end of the production of an injection moulding process starting with the pre-processing (insert preheating, testing of raw materials, dyeing and dry), the actual process of injection (temperature, pressure, moulding cycle time) and the post-processing (annealing, humidity).

The goal

The use AI for early identification and automatic or semi-automatic correction of the manufacturing process parameters by implementing Machine Learning (ML) techniques aims to avoid quality defects in the plastic parts during the thermoplastic injection process. AI and ML will be used for real-time alarm generation system with prescriptive maintenance indications according to the analysis of quality defects. Implementing a Machine Learning (ML) and Human in the Loop (HITL) techniques will allow the interaction of the operators in charge of the production lines with the ML Predictive models.

The scope

The scope of the demonstrator is to create an intelligent system, which will be able to communicate with the injection machine’s control unit and with the operator’s inputs through Human in the Loop (HITL) Dashboards, to automatically adapt the TEAMING.AI Engine to self-adjust the quality control model parameters when a deviation from nominal is detected in the case of sufficient trust assurance evidence and otherwise to request human assistance based on preliminary process diagnostics. The goal is to increase the Overall Equipment Effectiveness (OEE).

IAL will develop and implement a predictive analysis system by using ML boosted by HITL techniques. In AL, the data will be taken, trained, tuned, tested and more data will be fed back into the algorithm to make it smarter, more confident, and more accurate.

Use Case Scenario 3

Ergonomics and risk prevention in large part manufacturing

Workers have to manipulate and manually clamp large-sized and heavy parts in highprecision manufacturing machines for grinding or milling operations with high quality. This process takes an important part of the total cycle time of a working order and workers are exposed to occupational risks.

The scope

The scope of the demonstrator is to create an intelligent system able to communicate with both the Tracking and scene analysis System (TSAS) and with the operator’s inputs through Human in the Loop (HITL) Wearable to compensate for each other’s limited visual perception to maximize awareness about dangerous situations and conditions that negatively affect physical or mental ergonomics.
Based on human feedback, TSAS learns to predict which sequences of actions are ergonomically favorable. As TSAS cannot be sure to have the complete information about the situation due to occlusions and non-visible areas, it expresses its self-trust to the human operator when displaying e.g. safety warnings. The goal is to increase the Overall Labor Effectiveness (OLE) and operator satisfaction.