Yokogawa, (hereinafter referred to as "our company") has been contributing to the development of society by providing leading-edge products and solutions to the industrial world based on the technologies of measurement, control, and information since its establishment in 1915.
Our company's efforts to solve various customer challenges not only increase the economic value generated by customers, but also contribute to solving social issues that lead to the SDGs, such as energy and resource conservation, reduction of greenhouse gases, and improvement of safety. In August 2017, our company declared the commitment to contribute to the realization of “quality life of all = Well-being,” in its Three goals for Sustainability, and launched the Life Innovation Business (currently the Life Business).
In the Life Business, in 2020, we formulated a business vision “We will lead the world in advancing Bio Industrial Autonomy and contribute to a future embracing global harmony.” We will realize this vision by expanding our business in five business fields, including food production, using new technologies and products under development as a starting point in addition to our existing businesses, and by contributing to the realization of innovation in the entire value chain, from basic research to logistics services (Figure 1).
In the field of Food & Beverage, we support our customers in their efforts to ensure food safety and security and reduce costs, as well as to create comfortable workplaces, by working closely, coping with challenges together and proposing the ideal factories suitable for their operations. In addition to improvement of operation and solution of problems at existing factories, in recent years we have received an increasing number of consultations from the conceptual stage regarding the examination of operation and systematization for new factory construction.
In this article, we would like to describe the proposal of system introduction and the experience of talking with customers of many food factories as a project manager at the time of introduction.
2 What’s Required of Food Manufacturing Industry
We believe that all food manufacturers are dedicated to “Quality improvement” and “Cost reduction” on a daily basis. In order to meet the high quality demanded by consumers, we must strengthen our efforts to ensure safety and security, and thoroughly reduce waste and minimize costs. These two major challenges will continue to exist in the future.
In addition, there is a recent need to address the issue of a declining workforce, as well as the need to improve new lifestyles and work environments affected by COVID-19. “Business continuity” is also required in the food manufacturing industry, which must continue operating under any circumstances (Figure 2).
It is estimated that the working population in 2050, about 30 years from now, will decrease to two-thirds of the current level in Japan, and investment in operations, especially in automated equipment, is on the rise. In recent years, we have been receiving an increasing number of consultations on smart technology, IoT, AI, and DX (Digital Transformation), as there is a strong possibility that a manufacturing operation model that relies on human power will not be viable in the future. It is necessary to advance the preparation to realize the improvement of safety, product quality, and cost competitiveness of production sites by utilizing these technologies. That is the smart factory and the first thing to aim for is “automation”. Automation refers not only to the mechanization of direct operations, but also to the automation of indirect operations.
In the next section, we will introduce some examples of automation in the food manufacturing industry.
3 Key Points of “Automation” to make Production Sites Smart: Manufacturing Sites
We would like to introduce three points of automation directly related to manufacturing: “Automation of manual work,” “Automation of line inspection work,” and “Efforts not to rely on individual skills”.
3.1 Automation of Manual Work
There are many manual operations in the field work, which place a heavy burden on the workers. By promoting mechanization, we can expect to reduce the workload and man-hours and improve production efficiency. However, if we proceed without being aware of the key points of introduction, we may not be able to achieve the expected effect.
At Company A, the packing process had become a bottleneck due to manual process. Therefore, the company used the subsidy to install an automatic packing machine, which doubled the efficiency of the process. However, when the machine was put into operation, workers were required to supply intermediate products, wrapping materials, and pack the finished bags, and the lack of manpower eventually became a bottleneck.
This is a case of introducing the automated machine without considering the “flow of goods” and “flow of work” before and after the process to be automated.
Company B automated the weighing of raw materials in small portions. However, since the weighing values had to be set manually each time, human error occurred and it was not easy to record the results.
This is a case where the company was not aware of the “flow of information” such as the setting data to the automated equipment and the actual data from the equipment.
As you can see from the above examples, “automation of a single piece of equipment” and “automation of a line” are different.
Automation of equipment through mechanization and robotization is an effective means, but automation of a line means the flow of goods and information without human intervention. It is necessary to recognize that promoting mechanization and robotization with partial optimization will only change the bottleneck, and to consider the whole picture.
For example, in the case of Company B mentioned earlier, the establishment of a manufacturing execution system that manages the entire manufacturing process, not just the weighing process, would have enabled the company to reduce the manpower load and man-hours by creating an environment that is aware of the back-and-forth relationship between the automated equipment and the instructions and results.
In this way, total optimization can be achieved by data linkage between the manufacturing execution system and the automated equipment, and the introduction effect can be further enhanced (Figure 4).
3.2 Automation of Line Inspection Work
Major inspection devices such as metal detectors, X-rays, and weight checkers have already been automated, but there are still many sites where inspectors perform visual inspections on the line to detect foreign objects. Variation in accuracy is likely to occur due to the load depending on the skills of a particular inspector and the loss of concentration due to fatigue. Also, as a countermeasure against COVID-19, the number of personnel at the manufacturing site should be reduced as much as possible.
In recent years, camera performance and machine specifications have evolved significantly, and cases of utilizing Deep Learning are also increasing.
At Company C, it became possible to detect small foreign objects that could only be found visually, and to distinguish between patterns and foreign objects on the product.
At Company D, image inspections that cannot be distinguished by size or color alone are judged by multiplying not only two-dimensional data but also three-dimensional data including height.
Conventional image processing, such as rule-based thresholding, has limitations in over-detection and parameter adjustment which normal products are marked as NG. The use of deep learning, in which AI learns rules and patterns based on experience with input data, in the same way that humans learn through experience, will make it possible to automate visual inspections performed by skilled workers.
In cases where there is little difference between non-defective products and foreign matters, or where the color of good products and defective products changes depending on the situation, the material of the object can be determined by using a hyperspectral camera to capture the wavelength. It may be possible to detect foreign matter in vegetables, fruits, dried noodles, etc., to make fried foods uniformly fried and browned, and to determine the degree of damage and freshness of foods themselves.
3.3 Efforts not to Rely on Individual Skills
From the perspective of a declining workforce and skill succession, it is expected to become even more important efforts not to rely on individual skills in the future.
Although personalization occurs in many processes, we will introduce two of them, planning work and standardization of work procedures.
In many cases, planning work is dependent on individuals. A skilled person in charge of planning takes conditions into account that affect quality, such as allergens, in the manufacturing sequence, the reduction in the number of mold changes and cleaning due to product switching, and the optimization of resources, such as equipment and personnel. We often hear that it is difficult for the person in charge with knowledge to take leave.
At Company E, the person in charge used to spend a day planning the week's production but as a result of introducing a production scheduler, the scheduling has been automated and only minor adjustments need to be made manually.
In addition, in food factories where plans are frequently changed, prompt information sharing is essential. By visualizing the work that used to be done by one person using Gantt charts and systemizing the planning rules and sharing them immediately, it is possible not to depend on the personnel. Such efforts, more than introducing a system, provide the best opportunity to identify the constraints between processes and lead to skill succession. In addition, at the manufacturing site, there are frequent checks to ensure that raw materials and supplies are correct, that they have not expired, and that work is being performed in the correct order.
In order not to depend on individual skills, it is important to standardize the work procedures so that everyone can perform the work in the same way, which leads to stabilization of the work quality. Furthermore, it will be possible to make comparisons from the same point of view, which will provide an opportunity to consider improvements. Systematization of on-site work instructions makes it possible to perform work according to the system's work instructions and to check raw materials and materials using barcodes. As a result, double-checking, which used to be done manually, can now be done manually and with the system, resulting in a reduction in the number of workers.
It is necessary to reduce the dependence and the burden on skilled workers in order to improve the situation where the work site cannot run without skilled workers and decisions cannot be made. In addition, standardization of work procedures is effective as a countermeasure against COVID-19 because it can create an environment in which anyone can provide support.
4 Key Points of “Automation” to make Production Sites Smart: Indirect and Management Operations
We would like to introduce the following three points for indirect and management operations: “Building an environment that prevents sudden failures,” “Building an environment that enables paperless manufacturing,” and “Automation of data collection necessary for management”.
4.1 Building Environment that Prevents Sudden Failures
There are many customers who have problems with equipment maintenance. The equipment has the greatest impact on the operating rate, but in many factories, maintenance is done only after a failure has occurred, and they have not taken any steps to analyze the cause of the failure.
To address this issue, we are supporting the transformation of equipment maintenance from corrective maintenance to preventive maintenance.
One of the ways to build an environment where sudden failures do not occur is through automatic analysis and real-time detection of predictive signs using machine learning by on-site controllers. Temperature, current, vibration, etc., are collected as operating data of the equipment, diagnosed in real time in the controller, compared with models to detect signs, and alarms are automatically raised when maintenance is required.
In this system, information collection, analysis, and diagnosis are performed at the edge (near the device or terminal), thereby reducing the time required to detect abnormalities. Only the analysis results are transferred to the host system, eliminating the need for large storage capacity.
Furthermore, using machine learning, the system learns the "usual state" of normal operation, detects "unusual states" that deviate from the usual state, and executes an automatic analysis algorithm that raises an alarm as a predictive sign. Machine learning eliminates the concerns of not being able to obtain data until a failure occurs, not being able to determine the threshold, and not being able to model the failure until it has occurred several times.
This predictive detection often relies on the senses of veteran maintenance workers, and there is an urgent need to pass on their skills as they retire.
It is also important to determine appropriate maintenance standards and inspection items for the transformation to preventive maintenance.
Daily maintenance tasks, such as failure history, inspection records, and maintenance records, are often recorded in individual paper forms or in Excel. By using a system for maintenance management, data can be linked to each equipment master, maintenance records can be registered, and maintenance data can be centrally managed. Notification of predictive failure can also be managed in the same way.
By quantitatively sharing information and conducting failure analysis, it becomes possible to review maintenance standards, and the cycle of “failure analysis → review of maintenance standards → realization of preventive maintenance” can be executed.
By eliminating opportunity losses caused by equipment factors, not only the operation rate can be improved, but quality can also be improved.
In addition, preventive maintenance is effective not only in preventing failures by determining the maintenance cycle, but also in reducing waste by preventing excessive preventive maintenance. If the timing of repairs or adjustments can be determined from home or on a business trip by using predictive detection, it will be possible to understand the situation and immediately give instructions remotely, even in the event of an unexpected call.
4.2 Building Environment that Enables Paperless Manufacturing
Since it is recommended that no paper or media be brought in from outside as a measure against COVID-19, paperless manufacturing is also a key approach.
Operation using paper forms is time-consuming to fill out, and it takes time to circulate the forms. It is also necessary to take measures against transcription and handwriting errors, and it may be difficult to read the forms as they are filled in during work.
By going paperless, data such as work performance can be digitized, greatly reducing the time required for handwriting and circulation. It also eliminates the need to aggregate results and reduces the man-hours required to create daily reports, thereby improving data accuracy and operational efficiency. Above all, the electronic data can be used as materials for "kaizen" (improvement).
There are other benefits of going paperless. At the site where the enterprise system and the paper-based system are in operation, the mesh of the system plan and the site work is different, so people need to break down the work instructions, record, collect, tabulate, and post them. This makes it difficult to perform proper management.
By covering the difference between the management meshes of the enterprise system and the manufacturing site with the manufacturing execution system, it is possible to automatically link the systems and automatically send the actual results by confirming the work at the manufacturing site. In this way, real-time linkage to the enterprise system makes it possible to immediately utilize the site data for management decisions (Figure 6).
In addition, the paperless system reduces the number of man-hours required for filling out and inputting data, as well as the number of man-hours required for circulation, and increases the accuracy of data, leading to more accurate traceability and faster tracking.
4.3 Automation of Data Collection Necessary for Management
We often hear about the problems that line managers face when they cannot grasp the work status and progress without going to the work site, or when members of the indirect operations need to collect forms and make decisions about the data necessary for production planning.
By aggregating the operation data of various on-site devices and centrally managing and monitoring them in real time, it becomes possible to make quick judgments without going to the site. Furthermore, by managing data on work quality and history, it is possible to not only ensure quality but also utilize it for quality assurance.
In addition, if the video data from cameras installed for food defense and other purposes is checked at the same time, it will be possible to understand the situation in more detail and make an immediate decision without being at the site. Since there is no need for people to enter the site, the impact on quality will be minimized.
Such efforts are one way to pursue even greater safety and security. By looking at multiple data in real time, we can correctly grasp the operating status, progress, and work status, and see the variations in quality inspection data. If workers can monitor and control the progress of production in a centralized monitoring room, they can spend more time considering improvements instead of checking at the site. A dramatic increase in productivity may not be a dream.
Since it is difficult to implement centralized management at once in an existing factory, we are helping to realize it in stages.
Due to the impact of COVID-19, there are likely to be restrictions on work and business trip. From the viewpoint of infection prevention, the number of people in the production system should be kept as small as possible, and there may be concerns about stable operation and stable supply because it is not possible to send support across regions. How about setting up a system that allows members with knowledge to remotely monitor and support the operation of production bases from the head office or other factories?
In addition to checking operational data, it is also possible to give advice while viewing camera images of the actual site, send procedures and manuals to the site, add handwriting on the screen in AR (Augmented Reality), and have interactive communication.
There are many people who have been operating the system as a stopgap measure against COVID-19, but from now on, we believe that we should build this kind of environment as new normal with a long-term perspective. In addition, the number of veterans is decreasing, and if they leave for other sites, the work will not be completed. It will be possible to make optimal use of such “scarce resources”.
If such an environment can be established, it will be possible to provide remote support from the head office and partner companies when setting up overseas bases.
5 Key Points of “Automation” to make Production Sites Smart: Headquarters, Divisions, and Management
Lastly, we would like to introduce the key points of automation from the perspective of the headquarters, divisions, and management.
One of the comments from the management was, “We can check the results of products and costs entered in the enterprise system at the factory at the monthly report level, but we cannot grasp the detailed production status. We are worried about whether the site is okay”.
In this new normal era, it is necessary to speed up the decision-making process, so “rules that lead to actions in line with the top management's decisions,” in other words, standardization, is necessary. Speeding up the decision-making process cannot be achieved if it takes too much time to collect data from each workplace or department, so a paperless system and data aggregation are essential. Aggregate, centrally manage, and immediately utilize the analyzed data so that management can make smart decisions and take immediate action in each workplace. It is necessary to create an environment where these can be realized remotely. If we can automatically aggregate the operational data related to production from the systems of each factory, we can access it from anywhere, and we can see that the factory is doing well with our smartphones from home or on a business trip.
By sharing productivity indicators in real time among the various levels of the organization, such as factory workers, maintenance personnel, line managers, and management, it will be possible to immediately grasp whether the factory is keeping up with changes in the business environment, and to give precise instructions if any indicators deviate.
6 Beyond Automation
We have introduced the use of data to improve the efficiency of on-site work, to ensure quality, and to create a system to “support decision making.” However, there is a limit to the improvement activities through human intervention.
So, what will happen after automation? Here, we would like to introduce a future in which automated factories are constantly and automatically improved, that is, they will become autonomous and make decisions automatically.
What is autonomy? In automation, human judgment intervenes between a series of tasks. Safety assurance and operation are performed by people, and unexpected situations are handled by human intervention. Since the number of veterans to judge is decreasing, decisions need to be made automatically without human intervention. In autonomy, on the other hand, safety is guaranteed by the system, operations are performed automatically by machines and robots, the best possible adjustments are always made automatically, and unexpected situations can be handled without human intervention. This type of factory is called an “autonomous factory.”
In our company, we depict this process of transformation from automation to autonomy as “IA2IA (Industrial Automation to Industrial Autonomy).” In order to enable the best adjustment to be performed automatically without human intervention, it is necessary to make advanced use of on-site data and a lot of information. To achieve this, we need to work on DX = Digital Transformation. We would like to introduce an image of autonomy in a production line.
Figure 7 shows an example of autonomous tuning of equipment operating parameters while automatically monitoring operating data and quality data during production.
In the past, a sample inspection was performed, and the adjustment value of the operation parameter of the front-end process was determined by the worker judging the inspection results and the operating status, and then the worker manually changed the setting of the operating parameter. In an autonomous factory, operation data and quality data are constantly monitored, and AI is used to derive parameters for automatic adjustment, which are then automatically fed back to the equipment through system linkage. This advanced use of data will make autonomy possible. It may also become possible to adjust equipment operation methods and operation parameters based on energy predictions, or to automatically change production planning based on pre-detection of failures.
This transformation of business operations using digital data is often referred to as DX = Digital Transformation. At present, many companies are examining their approaches to DX.
DX's efforts are not something that can be done with tools alone. There are many manufacturers of IT, IoT, sensors, and AI, and each has different areas of expertise, so selecting a vendor can be a challenge. When introducing the system, it is necessary to determine what to do by collecting data and what to aim for, and then work on improvement. If this step is skipped, the expected effect cannot be achieved. A grand design drawn in advance is very important, because a fundamental solution cannot be achieved by blindly breaking through individual events. When drawing the grand design, it is important to note that methods such as automated equipment, automated systems, IoT, and AI are not the main players. The main focus should be on total optimization and business reform to ensure the quality of operations and improve productivity. In order to formulate a plan, we recommend that you identify current problems and issues, sort out what the roots are, and depict what you want to be in 10 years.
So far, we have discussed automation, autonomy, and the importance of an overall plan. In order to continue our business, we can't just stand still. We often hear people say, “it’s difficult and I don't know what to start with,” but it would be good to take the first step, even if it is a small one. One of the first steps is to start reviewing the wasteful operations that were revealed by the COVID-19 crisis. In the case study introduced here, the followings would be good examples.
・Try to automate line inspection at one place in a line first.
・Try attaching a sensor to one process, or more specifically, to one location, and record and observe the data first for data collection.
Just like when you get on the scale to start a diet, if you take the first step, you will find some points of interest. It is also important that we do not try to proceed on our own, but rely on those who have knowledge as our partners. In this day and age, we can create something together with our reliable partners rather than doing them on our own.
Our company's slogan is “Co-innovating tomorrow,” and we aim to deepen our relationship of trust with our customers, create unprecedented value together, and grow together with them.
This article was written with data from "The Latest Trends on Food Tech" by CMC Publishing Co., Ltd.
At Yokogawa, we understand that today’s food and beverage companies face unprecedented challenges in climate change, consumer demand, and increased global competitiveness. Overcoming these challenges will require innovative solutions that focus on key areas of production, asset management, and food safety and quality.