bg

Transform Your Supply Chain Planning and Marketing Strategies with Google Cloud and SAP Integration

Cloud Marketplace

Explore Pi Agent pi-logo

Digitizing Warehouse for Vehicle Parts Identification: A Revolutionary Transformation Journey of A Transportation and Logistics Company

Customer Overview

Our client is a rapidly expanding company, specializing in comprehensive Integrated Supply Chain Solutions, Global Forwarding Solutions, and Last Mile Solutions. With a global reach, they cater to customers across more than 50 countries, including India, UK, Europe, America, Asia Pacific, and Oceania.

Business Challenges

The client imports vehicle parts from vendors to their warehouse but faces a critical challenge in efficiently identifying these components and promptly detecting any misplaced items.

Although they have an internal subscription-based tool for parts identification, the team recognized the need for an independent tool to customize and unlock significant improvements in their operations and overall quality control efforts.

The company faced following challenges in their vehicle parts identification process:

  • Lack of Reliability: The current tool was slow in part identification and occasionally failed to recognize certain parts, compromising reliability and throughput.
  • Reduced Efficiency: Inefficiency in existing tool led to inconsistencies, errors, and delays in identifying parts, resulting in potential assembly line disruptions and increased costs.
  • Quality Concerns: The company’s ability to efficiently identify misplaced parts lead to quality control issues and customer dissatisfaction.
  • Limited Scalability: The current infrastructure had limitations in terms of scalability, preventing the company from efficiently handling increased data volumes, processing power requirements, or extending to other warehouses.
  • Data Privacy and Security: The current tool faced data security challenges due to its multi-access nature, where various users had differing levels of access to sensitive information. This complexity made it vulnerable to unauthorized access and potential data breaches.

The Client’s Aim: Streamlining Part Identification Process for Optimized Warehouse Management

Our client had a clear set of objectives they wished to accomplish to overcome their part identification challenges. These included:

  • Accelerated Part Identification and Accuracy: They wanted to reduce the time taken to read and identify spare parts, and achieve a time limit of 4 seconds or less. Additionally, they wanted to make sure that the same part is not scanned twice and accurately identify the similar looking automobile parts with minute differences, and different orientation.
  • Material Identification Tag (MIT): The client required solutions on MIT that could identify how many parts to scan and which slot to process first. The system should require operators to follow the first-in-first-out (FIFO) principle, and allow users to work with multiple MITs during a single session without constant login and logout. Such a system would enhance overall operational efficiency, streamline processes, and improve productivity and accuracy.
  • Material Verification: The client required a material verification system to ensure the quality and correctness of parts. The system should validate materials at specified time intervals, scan material IDs, determine SKU capacity, and maintain strict adherence to cycle times. It should allow queued materials to be scanned only after completing the current scanning process, with a little buffer time. Additionally, the system should scan and manage parts stored in TOTE containers, while capturing station numbers for reference.
  • A web based application: They required a web application that could open on mobile and desktop to display results of automobile part identification. This application should work only within the premises (intranet). Additionally, it should allow users to log in to only one system or device at a time.
  • ML Model Deployment on Edge Server: The client wanted to build the vision API model for deployment on Edge to analyze data locally, reduce latency, enable faster real-time decision-making, and ensure data security. For this, once the model was trained and optimized, it was deployed on the client’s Edge server.
  • Label Printing: They needed to make connectivity with downstream application (only one application) on printers to print the label, streamlining the printing process and ensuring accurate and efficient label generation.

Pluto7’s Solution Approach: Enabling Visual Workstation to Digitize Warehouse

Pluto7 helped the client create a vision workstation to digitize their warehouse for parts identification. The vision workstation analyzed multi-angle images of vehicle parts captured by cameras to accurately identify them using predefined patterns or characteristics.

The Vision workstation in the warehouse played a crucial role in automating the parts identification process, reducing manual effort, and improving efficiency. 

With our Decision Intelligence Platform, we enabled accurate and rapid identification of parts, saving time and improving quality checks:

Data Foundation and Vision AI Enablement

We gathered a comprehensive dataset of vehicle parts, including images and associated metadata, to train the AI/ML models. We also collected historical data on actual parts for training the system to detect misplaced parts, laying groundwork to enable vision AI. 

Data Ingestion and Analysis

We created data ingestion pipelines to efficiently and securely transfer data from the customer’s source systems to GCP buckets. Once the data was stored in the GCP buckets, it was further loaded into BigQuery, a fully-managed, serverless data warehouse, allowing efficient querying, analysis, and visualization.

ML Model Development and Deployment on Edge Server

Leveraging AI/ML algorithms, we developed a computer vision model capable of accurately identifying different vehicle parts based on their visual characteristics. The model was trained using deep learning techniques on the collected dataset. The ML solution could

  • Determine the parts by weight
  • Identify with a barcode scanner
  • Identify parts using vision AI 

Extensive testing and validation were conducted to ensure accuracy and reliability. The model was then deployed to on-premise edge servers for offline availability.

Web Application for Part Identification Results

Pluto7 also developed a web application that would display results of part identification. Everytime an image was uploaded to vision workstation, the backend model would identify the part and show the result of Pass or failure of the image identification on the web application accessible on both desktop and mobile devices.

Label Printing

On successful identification of the Part, command will be sent to the printer for printing the labels. They want to make sure that label can be printed only after the parts have been identified and scanned. 

Key Results

Reduced Parts Identification Time: The time taken for parts identification was reduced to 4 seconds or less after image upload, allowing for faster processing and improved overall productivity. 

Accurate Parts Identification: Vision AI solution developed by Pluto7 could identify similar looking automobile parts with minute differences like the mounting hall, and parts with different orientation like left and right orientation.

Increased Misplaced/Missing Part Detection: The system could detect misplaced or missing parts with a high degree of precision, enabling timely interventions and reducing the risk of wrong parts reaching the assembly line.

Minimized Errors: The new system ensures that the same part is not scanned twice, minimizing the chances of misidentified parts and associated errors.

Enhanced data privacy: Part identification model deployment on Edge server enhanced data security by minimizing data exposure to external networks. By keeping sensitive data and computations within the edge server, it reduced the risk of data breaches and unauthorized access. 

Scanning progress tracking: The system gives an indication to complete scanning the pending parts of a GRN (Goods Received Note), ensuring that all parts are accounted for, reducing the risk of missing or misplaced items, and helping maintain accurate inventory records. 

Web Application for Real-Time Results: Users got access to a web based application that worked both on mobile and desktop to see the results of automobile part identification.

Cost Optimization: Streamlining the parts identification process and minimizing errors, reduced expenses and improved efficiency.

By digitizing their vehicle parts identification process using Pluto7’s Konnect Manufacturing Platform and Vision AI solution, our client transformed their operations, achieving improved efficiency, accurate identification, and proactive detection of misplaced/missing parts. They leveraged the transformative impact of AI/ML technologies in manufacturing to drive operational excellence and deliver superior products to their customers.

 

Enable Decision Intelligence Into Every Corner Of Your Product And Operations.

industry image

Industry Manufacturing

Platform Konnect-manufacturing

Challenges

  • Slow, time-consuming, unreliable parts identification process
  • Prone to human error and inefficiency, current processes result in disruptions in the assembly line.
  • Lack of misplaced/missing part detection leading to quality control issues and customer dissatisfaction
  • Limitations in scalability preventing the company from processing power requirements to expanding operations
  • Risk of unauthorized access, loss of data, compliance issues

Results

  • A Vision AI system that captured and analyzed images of the parts to identify missing or misplaced parts
  • Higher throughput and increased reliability by increasing the precision in identifying parts
  • Improved productivity by reducing parts identification time to 4 seconds or less
  • Improved accuracy by making sure same part is not scanned twice, minimizing the chances of misidentified parts and associated errors
  • Reduced risk of wrong parts reaching the assembly line by detecting missing or misplaced parts
  • Enhanced data privacy by keeping sensitive data and computations within the edge server
  • Ensured parts accountability by tracking scanning progress
  • A web based application to see the results of automobile part identification accessible on both desktop and mobile
  • Reduced expenses and improved efficiency by minimizing errors and streamlining the parts identification process

Products Used

  • Google Cloud Storage
  • Google BigQuery
  • Vertex AI
  • Auto ML
  • Vision AI
  • Google App Engine or Cloud run
  • Google Cloud IAP or Firebase Authentication
  • Google cloud trace
  • Google cloud logging
  • Google cloud error reporting
  • Google cloud Firestore