Developing Manufacturing 4.0 capabilities does not happen overnight. Lot of groundwork needs to be done in order to create the foundational infrastructure. However, that should not stop you from starting to experiment through pilots to understand what “Smart” capabilities can be embedded in your Manufacturing network.
Irrespective of whether you are a manufacturing entity or a consulting entity looking to develop capabilities in Manufacturing 4.0, the key is to start exploring and building a roadmap accordingly. This article provides you an example of opportunity area that you can explore. Though optimization heuristics are commonly used, I really don’t see these prescriptive algorithms being useful in a Manufacturing 4.0 envoirnment where there will be little to no room for latency. This is why I strongly think that Neural Networks will be an excellent fit for Analytics in Manufacturing 4.0 envoirnment. While this proposed solution is for setup time minimization only, a “Master Neural Network” can be created to optimize other aspects of the manufacturing process.
Some manufacturing entities try to minimize setup time by sequencing the parts with other parts that have common geometry and tool requirements. This form is sequencing is inflexible and though it reduces setup time, it is extremely poor as far as responding to customer demand pattern fluctuations go. With the world getting customer centric, this form of sequencing is becoming unacceptable.
While many attempts have been made to minimize the cost or cycle time of Job shop manufacturing, those addressing the issue of setup waste, while meeting customer on-time demand, are rare.
Job Shop Scheduling (JSM) approach
Many have tried using JSM approaches, particularly the Shortest Processing Time (SRT) heuristics.
SRT Heuristics: Assume that the part number with the lowest total setup time plus machining time is denoted as P1, the second lowest is P2, and so on. Then run the parts in the sequence P1<=P2…..<=Pn. This will result in the lowest average, or mean, delivery cycle time.
This also did not address the customer service issue. While this method may result in a lower mean cycle time, it may also result in say part N in the sequence to be produced last even though it may be required first. Though you can run variations of SRT with additional constraints, where multiple variations of JSM are run.
The Neural Network Approach
We need to find the sequence of producing parts that will replace the wasteful random setups and result in a 50-70% reduction of setup time while still meeting customer delivery dates.
- List of all jobs and part numbers at each Pull group, all of which must be shipped within 30 days.
- A matrix of setup times from any part number to any other part number.
- The setup time data on tools, chucks etc. on the machine that is within 3 hours of finishing its machining task. The 3 hour deadline will allow us to respond to any new job that suddenly appears.
Creating Training Data and why Heuristics won’t work
Theoretically, the solution of this problem is similar to TSP heuristics. So for example there are four parts, A,B,C and D. The logic considers these setup times as “distance between these cities” and hence can leverage the logic of “visiting a number of cities” in a sequence that minimizes the distance, subject to the customer demand requirement constraints.
Remember that there is no mathematical formula that solves this problem. An iterative approximation can be found using Branch and Bound (B&B) . Since we are assuming, 3 hour deadline (in Inputs), B&B may be too time consuming.
So in order to generate the training data that the NN model will see, we need to generate several thousand B&B or TSP Heuristics solutions to random problems ( can be done offline in the cloud using data from the ERP system).
Training the Neural Network
These solutions are then used to train the Neural Network. The training process will adjust the weight of each neuron such that its output on the 4 job numbers is the same as the offline cloud solution. Thus when presented with a similar set of jobs in the future, the Neural Network will calculate the solution in minutes. The trained Neural Network can see every possible sequence of thousands of parts and finds the one that minimizes total setup time based on its training using Branch and Bound examples.
Those of us who frequently solve optimization problems know that it takes a lot of computing time to get to the absolute maximum of the objective function but you can get to 95% very quickly and in most cases, this 95% is sufficient. The same is true of Neural Networks.
Validate through simulation
A final verification for each recommendation can be done by running through a simulation program. Most of the manufacturing operations generally have such models setup already and running them through these simulations will make sure that Neural Network training is up to date.
Bridge between theory and Implementation
While theoretically the Logic seems straightforward, real world has its own challenges. And the major part of that challenge is not the Neural Network algorithm. The first challenge is to understand where you want tio implement this capability. Some of the questions that you need to address here are:
(1) Where will the algorithm sit within the Manufacturing planning systems landscape ? How will the information get consumed ?
(2) Is the data consistency and integrity in place to support training such a model ?
(3) Will the capability required to build such a model need to be developed internally or will an external partner help you in this quest ?
Views my own.