A leading pharmaceutical manufacturer with more than 140 years of business was using a medicine-dissolving machine that relied on human observation to determine errors. Monitoring the machine for errors takes time, which slowed down the progress of other work and increased the chance for human error. The client needed a way to accurately determine if something is wrong, in order to guarantee precise tests.
• Implemented an AI solution to eliminate the majority of human observation
• Collected video data of the client’s experiments by running them and viewing stored footage of the machine’s normal operation
• Developed a program in 3 months that understands when the machine’s state is correct or incorrect via deep learning
• Used computer vision to “visualize” process frame by frame, calculating deviation between correct and incorrect states automatically.
• Created a user-friendly dashboard for alerts, telling users if there’s anything out of the norm
• This solution saves time by notifying when the client’s machine isn’t in the correct state, instead of having to guess
• Allows employees to avoid bottlenecks in their work and finish tasks in a shorter amount of time.
• Mitigates the amount of human mistakes
• A step toward predictive maintenance, which we hope to implement for the client in the future
Full case study (77 kB)