https://sistem.wisnuwardhana.ac.id/index.php/sistem Sistem : Jurnal Ilmu-Ilmu Teknik Fakultas Teknik Universitas Wisnuwardhana en-US SISTEM Jurnal Ilmu Ilmu Teknik 0216-7131 <p>Penulis yang menerbitkan karya ilmiahnya di <strong>Sistem : Jurnal Ilmu - Ilmu Teknik</strong> setuju dengan ketentuan berikut : Bahwa jika manuskrip yang diajukan dan diterima untuk di publikasikan maka hak cipta milik Penulis, Jurnal SISTEM hanya sebagai penerbit artikel&nbsp; dengan karya yang dilisensikan secara bersamaan di bawah Lisensi Internasional Creative Commons Attribution-NonCommercial-ShareAlike 4.0.</p> <p>&nbsp;<a href="http://creativecommons.org/licenses/by-nc-sa/4.0/" rel="license"><img src="https://i.creativecommons.org/l/by-nc-sa/4.0/88x31.png" alt="Creative Commons License"></a><br>This work is licensed under a <a href="http://creativecommons.org/licenses/by-nc-sa/4.0/" rel="license">Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License</a>.</p> https://sistem.wisnuwardhana.ac.id/index.php/sistem/article/view/348 <p>Legacy industrial PID temperature controllers, such as the Autonics TN series (TNH/TNL), are widely deployed to run process cycles in small and medium-scale industrial operations but generally lack native IoT connectivity. While retrofitting such controllers for cloud-based monitoring (the read direction) is comparatively well established, writing process configuration back to the controller reliably, and verifying that the write actually took effect, remains a harder and less-explored problem. This study addresses that gap by proposing and evaluating a register-level write-then-read-back verification protocol for bidirectional deployment of process patterns (PTTN: pattern number, step count, per-step setpoint and duration, and end-state behavior) to a TNH controller over a direct RS-485 serial path (Jalur A), one of two architecturally distinct control paths in the system under study; the other, an ESP32-mediated cloud path (Jalur B), is implemented and compile-verified but not yet validated on physical hardware. Using a Modbus RTU simulator and 100 synthetic patterns ranging from 1 to 20 steps, the protocol achieved 100.00% Pattern Deployment Accuracy (SD = 0.00%), with p50, p90, and p99 end-to-end latencies of 511, 543, and 578 ms, respectively, that did not scale meaningfully with pattern complexity. A subsequent fault-injection experiment (N = 45 trials: 15 baseline plus 30 induced-fault) found that under transient device unresponsiveness, all interrupted trials were reported as client-side failures even though independent read-back confirmed the controller state had been correctly updated, revealing a false-failure risk that could cause naive retry logic to redundantly or unsafely re-send already-successful commands. These simulator-based results establish the correctness and reliability characteristics of the verified deployment logic and motivate, without yet empirically validating, the necessity of the cloud-capable Jalur B path, whose hardware validation remains future work.</p> Rachmad Andri Atmoko Firman Pratama Dewantara Bayu Sutawijaya I Dewa Made Widia Salnan Ratih Asriningtias M. Fajar Adiatmoko Akas Bagus Setiawan ##submission.copyrightStatement## http://creativecommons.org/licenses/by-nc-sa/4.0 2026-08-11 2026-08-11 22 2 1 10 10.37303/sistem.v22i2.348 https://sistem.wisnuwardhana.ac.id/index.php/sistem/article/view/343 <p>Planning the installation of a Solar Power Plant (PLTS) requires weather condition information so that projects can be implemented according to plan. However, monitoring weather conditions and forecasting weather for several days to one month ahead have not been optimally utilized as a basis for decision-making in PLTS project planning. Therefore, this study aims to design and develop a weather prediction device based on the Backpropagation Artificial Neural Network (ANN) algorithm so that weather conditions for one month ahead can be predicted accurately. The sensors used were a BH1750 sensor to measure solar light intensity, a BME280 sensor to measure temperature, humidity, and air pressure, a wind-speed sensor, and a rainfall sensor. A total of 10,080 sensor data records were obtained and stored on a microSD module. The data were then processed as inputs for the Backpropagation Artificial Neural Network algorithm to generate weather predictions. The results show that the combination of the Backpropagation Artificial Neural Network algorithm successfully produced a one-month weather prediction with the lowest Mean Absolute Error (MAE) of 0.218595.</p> Aldo Febriyan As’ad Shidqy Aziz Widi Aribowo Mahendra Widyartono ##submission.copyrightStatement## http://creativecommons.org/licenses/by-nc-sa/4.0 2026-08-20 2026-08-20 22 2 11 19 10.37303/sistem.v22i2.343