3 challenges in operations – and how AI can help solve them

You’ve heard it a million times – machine-learning solutions are proven, supported by available technology, and already transforming the way companies operate. This is because they are simply faster, more consistent and more accurate than human employees when it comes to repetitive, laborious and observational tasks.

Operational settings come with their own set of obstacles, issues and inefficiencies that are recognizeable across businesses and even sectors. In this blog post, we describe three familiar challenges and present high-impact solutions involving AI technologies.

Challenge 1: inconsistent or inefficient quality inspection

Outdated automatic quality control inspection systems aren’t very good at identifying subtle differences in quality, and human workers may inconsistently label products. The result: too many rejections and inferior products reaching the market.

Computer vision is one of the three core technologies we apply at delaware.ai. Together with our partner Robovision, we created a deep-learning model that analyzes images of goods on the production line on a pixel-by-pixel basis, using training data contributed by expert labelers. Such a system is capable of discerning the difference between superficial defects such as scratches and more serious problems that affect a product’s performance – and continues to learn as it operates.

Challenge 2: huge quantities of unstructured text

While this activity doesn’t fall under the "typical" operations umbrella (production environments, quality control, etc.), it is indeed an operational process when a business has to sift through paper-based and digital text to find pieces of information necessary to offer services to customers. Common search technologies are often rudimentary and not particularly helpful when employees need to find very specific documents in multiple languages.

Enter natural language processing, our second core technology. A natural language processing algorithm is capable of recognizing and understanding the semantics of textual data, tagging, grammatically analyzing and extracting terminology from it as accurately as a trained employee.

Challenge 3: the wasteful production line

Probably one of the most universal challenges we encounter, production waste is not only expensive – it can also lead to lower product quality and higher costs, damage brand reputation and impact corporate sustainability targets.

Driver analysis enables process engineers to zero in on what, how and when to tweak elements of the production line to achieve the perfect balance of maximal quality and minimal defects. In this approach, a model predicts the outcome of the production process based on diverse production parameters. Those parameters with the strongest effects on product outcomes are ‘drivers’, which can be configured to achieve optimal products.

Why is driver analysis so powerful? It transforms employees’ gut feelings into a statistically sound model that reconstructs the entire production process, leading to less scrap, cost savings and more.

Partnerships that put proven AI models to work for you

delaware.ai joins forces with Robovision and DataStories to bring the benefits of machine learning to your processes quickly, efficiently and cost effectively. Our experienced data scientists work closely with your team to develop models with the biggest impacts, supported by technology and platforms that enable your business to configure them without the need for data science expertise.

Interested in learning more about the operational challenges that we help real companies tackle through AI? Download our exclusive e-book: 'AI & operations: Adding intelligence to your operational processes.'

Need some help identifying the best applications of Industry 4.0 technologies in your company? Get in touch with one of our experts.

Our expert

Wouter Labeeuw

Wouter Labeeuw

Wouter Labeeuw works as data science and machine learning consultant, adding intelligence to applications. He is a computer scientist with a PhD in engineering, where the research was focused on applying machine learning in the context of electrical demand response. In January 2016, he started at Delaware, still focusing on machine learning but in a broader context. Within his current role, he is responsible for the data science team within delaware.

Contact us